Menu
Tax Notes logo

Facebook, the IRS, and the Commensurate With Income Standard

Posted on Dec. 21, 2020
Stephen L. Curtis
Stephen L. Curtis

Stephen L. Curtis is a transfer pricing economist and the president of Cross Border Analytics Inc. He thanks Jeffery Kadet, Diane Ring, Paul Blankfeld, and two anonymous reviewers for their assistance.

In this report, Curtis applies the commensurate with income standard embodied in reg. section 1.482-7(i)(6) to analyze the IRS’s proposed adjustment to Facebook’s 2010 cost-sharing arrangement.

The views expressed in this report are solely the author’s and do not reflect those of any other person or institution.

Copyright 2020 Stephen L. Curtis.
All rights reserved.

I. Introduction

In ongoing transfer pricing litigation, Facebook is challenging the IRS’s adjustment of a buy-in payment under a cost-sharing arrangement (CSA) with an Irish affiliate. The IRS stated that the adjustment was necessary to ensure compliance with the commensurate with income (CWI) standard added to section 482 by the Tax Reform Act of 1986. It is unclear, however, whether the IRS has performed a reg. section 1.482-7(i)(6) periodic adjustment calculation in this case — either before issuing its initial proposed adjustment in 2016 or before its more recently revised adjustment in October 2019.

Applying forensic economic investigation, this report demonstrates that the CWI tool provided by the regulations is relevant to this case and supported by the facts. Using available information, the analysis yields an indicative periodic adjustment calculation that appears to better measure Facebook’s compliance with transfer pricing laws than the IRS’s current approach under reg. section 1.482-7(g). The results suggest that it may be in Facebook’s interest to accept the IRS’s proposed adjustment, and conversely, that it may be in the IRS’s best interest to withdraw that adjustment and instead perform a periodic adjustment to one or more recent open tax years. That can be known definitively only if a periodic adjustment test is performed using complete information.

II. Valuing PCTs

In 2016 the IRS issued Facebook a notice of deficiency and a penalty assessment for undervaluing the buy-in payment of Facebook Ireland Holdings Ltd. under a September 2010 CSA. The original IRS adjustment and a 2019 modification revalued Facebook’s buy-in payment on a net present value (NPV) basis from $6.7 billion to around $20 billion.1 The IRS recharacterized the entire buy-in payment as a platform contribution transaction (PCT). Facebook and its public auditor had valued only the platform technology intellectual property buy-in payment as a PCT and had characterized the user base and marketing IP as non-PCT intangibles. The IRS asserted that the user base, marketing, and technology IP were synergistic and should have been valued as an integrated form of IP that contributes to the intangible development activity as defined by reg. section 1.482-7(c)(1).2 The IRS performed an aggregated valuation of the buy-in payment consistent with guidance in reg. section 1.482-7(g)(2)(iv). Facebook has challenged the IRS’s adjustment in the Tax Court. The trial will continue into 2021.3

The IRS appears to have followed guidance in reg. section 1.482-7(g) for valuing PCTs (which includes buy-in PCT payments). However, another section of the cost-sharing regulations contains separate guidance on the valuation of PCTs: Reg. section 1.482-7(i)(6) gives the IRS a periodic adjustment tool4 to implement the CWI concept, which was added to section 482 by TRA 1986 to address the difficulty of valuing the economic rights associated with various forms of IP.5 The CWI concept provides a statutory basis for future adjustments of the original transfer price of an intangible to reflect the actual income earned from the intangible in the years after its transfer and initial valuation.

In a declaration filed in July 2016, the IRS revenue agent stated that the adjustment to Facebook’s buy-in payment was motivated by the CWI concept:

The IRS is authorized under I.R.C. section 482 to allocate income and deductions among commonly controlled or owned business entities as necessary to prevent tax avoidance or to clearly reflect income. In particular, when rights to intangible property are transferred or licensed between related entities, “the income with respect to such transfer or license shall be commensurate with the income attributable to the intangible.” . . . The information gathered suggested to the IRS examination team that the E&Y approach to valuing Facebook’s transferred intangibles on a stand-alone basis was problematic.6 [Emphasis added.]

It is unclear from the declaration whether the audit team intended to explicitly measure Facebook’s compliance with the CWI standard or was simply reiterating the taxpayer’s obligation to comply with that standard. However, by July 2016, when the IRS issued its adjustment to Facebook’s 2010 tax year, the agency had access to Facebook’s tax returns through the 2014 tax year and therefore could have applied the CWI standard explicitly. Those returns would almost certainly have shown only losses in Facebook Ireland in each of those years (SEC filings reported foreign pretax losses in each year).

Thus, the IRS’s adjustment to Facebook’s buy-in payment was in effect an argument that the foreign losses resulting from the IP transfer should be ignored, although a CWI analysis under reg. section 1.482-7(i)(6) would explicitly include them. When the IRS issued its initial Facebook adjustment, it was defending a similar adjustment asserted against Amazon.com under similar circumstances. The Tax Court would later decide that case against the IRS.7

The highly uncertain nature of a transfer of economic rights associated with potentially valuable IP for exploitation into an unknown market with unknown revenue potential is exactly why, for cost sharing, reg. section 1.482-7(i)(6) provides an explicit and objective “trigger” tool — a periodic adjustment — for measuring compliance with CWI. The periodic adjustment is designed to be applied at any time (regardless of whether it had been applied in an earlier year) and to produce a single-year result on a rolling basis to measure compliance with the pricing of the initial PCT and all later PCTs.

This CWI tool has two parts. The first is an objective trigger mechanism that is purely mechanical and based on existing financial results. It indicates whether the financial results that a CSA participant has reported in the tax periods following the CSA and any PCTs are reasonably consistent with the participant’s investment. That investment is represented by buy-in amounts paid for PCTs and the participant’s cost contributions (CCs) to intangible development costs (IDCs) incurred under the CSA. The second part of the CWI tool, which becomes relevant only when there is an upward trigger for a particular year, is the computation of the periodic adjustments that one or more CSA participants must pay to another participant. Because periodic adjustment calculations are made using the residual profit-split method, there can be some level of subjectivity based on the facts and circumstances.

An important reason for performing the objective trigger calculation is the recognition that contemporaneous PCT valuations are highly subjective and subject to error. They can be based on inaccurate or faulty data or an incomplete understanding of the complex mechanisms that underlie the value of the relevant IP. Estimates also depend on the nature of the transferred IP, which valuation methods should be applied, how they should be applied, whether comparable uncontrolled transactions might exist, and the extent of any adjustments. These are all matters on which the IRS and the taxpayer may disagree, to the point at which courts are called upon to make Solomon-like judgments.

It is clear from the revenue agent’s declaration that the IRS in 2016 rushed to apply an adjustment without all the information it sought.8 However, the IRS should now have all the information it needs to perform at least nine years of annual periodic adjustment calculations through 2019.9 And if the IRS desires, it could impose an adjustment on that basis instead of the 2010 adjustment that it has been pursuing since 2016. As explained and illustrated later, the significant profits earned by Facebook Ireland in the years after the IRS’s 2016 adjustment should support one or more periodic adjustments that could exceed the adjustment that is now being litigated. Further, Facebook would find it much more difficult to challenge this potentially more reliable adjustment.

Facebook and the IRS disagree on several facts affecting the value of the buy-in payment; however, those issues become irrelevant when applying the reg. section 1.482-7(i)(6) periodic adjustment, which more accurately measures and imposes tax on a CWI basis. One such issue is the measurement of projected results. The covered technology IP in 2010 applied mainly to personal computers, or desktop devices. Within a few years of the 2010 buy-in payment, a significant proportion of existing and new Facebook users migrated from computer platforms to smartphones and other mobile platforms, for which Facebook in 2010 did not have a viable technology platform. The IRS believes that the possibility of future revenue from yet-to-be-developed mobile platform technology should have been included in the buy-in payment for existing desktop technology. This same issue arose in Veritas,10 and the court held that the IRS’s inclusion of hypothetical profits for different, nonexistent technology in the buy-in payment for preexisting technology was an overreach. The CWI periodic adjustment, however, measures actual results, so these differences affecting projected results are irrelevant.

To determine the PCT buy-in payment, the IRS applied methods based on the arm’s-length standard that are described in reg. section 1.482-7(g). In theory, reg. section 1.482-7(i)(6) values the same PCTs, but with its CWI focus, the regulation uses different economic data and formulas. These are separate calculations that the IRS may, when it desires, apply to the same taxpayer and the same PCTs. While the former calculates a PCT “price” based on conditions and expectations as of the time of that PCT’s transfer, the latter calculates adjustments to be recognized in future years based on later financial results as reported by the relevant CSA participant and PCT payer.

A natural question is: With the regulations giving the IRS two tools to value the same transaction, each with its own focus, should one or both be applied? It is clear that with the information available to the IRS at the time of its initial 2016 adjustment, application of the two mechanisms produced different results. The 2016 adjustment (although increased in 2019) still reflects the application of the reg. section 1.482-7(g) mechanism. For any guidance on whether and how the two mechanisms might be reconciled, one must look within the cost-sharing regulations and relevant background material.

III. The Periodic Adjustment Rules

As noted, Congress enacted the CWI standard in TRA 1986. The Joint Committee on Taxation’s blue book summarizes some congressional guidance included in the legislative record:

Congress did not intend, however, that the inquiry as to the appropriate compensation for the intangible be limited to the question of whether it was appropriate considering only the facts in existence at the time of the transfer. Congress intended that consideration also be given to the actual profit experience realized as a consequence of the transfer. Thus, Congress intended to require that the payments made for the intangible be adjusted over time to reflect changes in the income attributable to the intangible . . . it will not be sufficient to consider only the evidence of value at the time of the transfer. Adjustments will be required when there are major variations in the annual amounts of revenue attributable to the intangible.

In requiring that payments be commensurate with the income stream . . . the profit or income stream generated by or associated with intangible property is to be given primary weight.11 [Emphasis added.]

The IRS’s ability to apply the CWI standard in a CSA context was explicitly affirmed by the Ninth Circuit in Altera.12 The court stated:

Treasury reasonably interpreted congressional intent in the 1986 amendments as permitting it to dispense with a comparable transaction analysis in the absence of actual comparable transactions. . . . As we have discussed . . . in implementing the commensurate with income amendment, Treasury was moving away from a purely method-based, comparable-transaction view of the arm’s length standard in attempting to achieve tax parity. Treasury’s citation to the amendment, and its legislative history, demonstrates that its position was not inconsistent, and there is no basis under Chenery to invalidate it.13

The preamble to the temporary CSA regulations (reg. section 1.482-7T) promulgated in January 2009 stated:

The Treasury Department and the IRS reaffirm that the CWI principle is consistent, and periodic adjustments are to be administered consistently, with the arm’s length standard. . . .

In determining whether to make any periodic adjustments, the Commissioner considers whether the outcome as adjusted more reliably reflects an arm’s length result under all the relevant facts and circumstances.14

While the preamble language says that the CWI principle is consistent with the arm’s-length standard, reg. section 1.482-7(g)(2)(ix)(A) states that the pricing methods in reg. section 1.482-1 “shall not override the rules provided in paragraph (i)(6) of this section for periodic adjustments by the Commissioner.” Those rules allow the IRS to depart from the pricing rules in reg. section 1.482-1 to perform a periodic adjustment calculation based on CWI concepts, versus concepts based on transactional comparability with uncontrolled transactions, and to conclude that the result is consistent with the arm’s-length principle.

Although guidance in reg. section 1.482-7(g) will not “override” the application of reg. section 1.482-7(i)(6), it is clear that these two rules both target compliance and must in fact work in tandem. A taxpayer must comply with reg. section 1.482-7(g) in setting its PCT pricing, which is subject to IRS review, and the agency can impose reg. section 1.482-7(i)(6) periodic adjustments. If the IRS makes an adjustment under reg. section 1.482-7(g) to a taxpayer’s PCT pricing, the adjusted PCT price will become a factor in applying a periodic adjustment under reg. section 1.482-7(i)(6). Likewise, the higher the adjusted PCT price, the less likely it is that a taxpayer’s realized results will trigger a periodic adjustment. It may be expected that an IRS adjustment under reg. section 1.482-7(g) to increase a taxpayer’s PCT price would not alone cause a periodic trigger under reg. section 1.482-7(i)(6).

Note that reg. section 1.482-7(k)(2)(ii)(J) requires that taxpayers maintain documentation that includes post-implementation data to demonstrate that their PCT pricing “has been applied in a reasonable manner.” Those data would be a part of any periodic adjustment calculations under reg. section 1.482-7(i)(6).

In Facebook’s case, the foreign losses it incurred through 2014 emphasize the uncertainty of the value of the IP as of 2010. A periodic adjustment trigger computation performed using only the information available when the IRS proposed the initial PCT adjustment in 2016 would show that the agency could not impose any periodic adjustment. This is because, under the objective and operating results-oriented test, the actual returns earned within Facebook Ireland were not excessive in the initial years after the 2010 PCT. Rather, Facebook Ireland realized losses in those years, making it all but impossible for the actually experienced return ratio (AERR) to have exceeded 1.5. (See discussion of trigger calculation below.)

The objective trigger tests in reg. section 1.482-7(i)(6) mean that the periodic adjustment is less prone to so-called dueling economists syndrome. Moreover, a periodic adjustment trigger calculation made using a proposed IRS PCT adjustment under reg. section 1.482-7(g) acts as a check on the PCT adjustment. If that adjustment is adequate, the trigger calculation should determine that no periodic adjustment will be allowed under reg. section 1.482-7(i)(6).

To some observers, paragraphs (g) and (i)(6) of reg. section 1.482-7 might appear to overlap; to others, they might appear to be contradictory. However, the two provisions can be reconciled and should work together to achieve the purpose of section 482 and its CWI concept. Because the trigger test of reg. section 1.482-7(i)(6) is purely objective, is based on actual results, and cannot be overridden by traditional transfer pricing methods, it seems that the periodic adjustment rules should normally carry great weight in cases in which results under paragraphs (g) and (i)(6) produce incompatible compliance assessments.

Indeed, the legislative history and court decisions suggest that the periodic adjustment rules under reg. section 1.482-7(i)(6) — which include objective, after-the-fact tests of relative profitability following a PCT — should be given appropriate weight when considering PCT adjustments under reg. section 1.482-7(g). Yet it appears that the IRS has given little or no thought to the guidance that might come from applying the reg. section 1.482-7(i)(6) tests to consider the reasonableness of reg. section 1.482-7(g) adjustments. The IRS’s history of CSA examinations is replete with Tax Court reversals of its attempts to apply the highly subjective, controversial, and error-prone methods in reg. section 1.482-7(g) to value PCT payments with only information available as of the date of the PCT.

It is even more perplexing why the IRS has not computed periodic adjustments, which will often be more objective in nature and based on actual financial results, instead of the fully subjective and difficult-to-defend reg. section 1.482-7(g) rules. A well-documented periodic adjustment approach should be significantly more immune to challenge and would produce a more reliable and sustainable result under the law. It would also be simpler and considerably less resource-intensive than current approaches. So why has the IRS not done this in the Facebook litigation or earlier cases?

A. Reasons to Apply the CWI Test

It is important to note that the CWI test as defined by reg. section 1.482-7(i)(6) is all-encompassing because the inputs include all the expenses and profits of the foreign CSA participant related to IP development, IP exploitation, and PCTs contributed at the inception of the CSA, as well as those developed or contributed later. Therefore, a violation of the CWI standard could arise from almost any activity that relates to a CSA, even if the buy-in PCT was otherwise compliant.

The trigger test is also incontrovertible because there is no estimation or subjective element to it; it is substantially mechanical, and the threshold is fixed.15 The actual results at any point in time (the AERR) incontrovertibly exceed or fall below the fixed periodic return ratio range (PRRR) upper threshold of 1.5.16 The threshold and the inputs are clearly described — there can be no surprises.

Also, as the IRS has made clear, it considers this test consistent with the arm’s-length standard — although this is a moot point because the CWI standard is statutory law. Notably, the CWI periodic adjustment may be applied to all future years; the IRS has the freedom to apply it to any year in which the AERR exceeds the 1.5 PRRR threshold, even if the IRS could have, but did not, apply it to earlier years in which the threshold was also exceeded. Recognizing the long-term nature of this CWI approach within the cost-sharing regulations, the only exception that cuts the IRS off from applying reg. section 1.482-7(i)(6) to all future years is an exception that applies only if the PRRR upper threshold is not met in each of the 10 consecutive years from the date that exploitation begins.*

So why should the IRS apply this test, and why has it not done so in Facebook’s case, or possibly in any other case to date? The first question can be answered with a stylized hypothetical example: Consider a taxpayer that implements a CSA with a foreign affiliate involving a single PCT payment for existing IP in year 1. In this first year of the arrangement, the only income in the foreign affiliate is attributable to the existing IP, but the affiliate expects to begin earning income attributable to newly created IP in year 2, and the income from the existing IP is expected to decline over time. Table 1 shows the estimates of the foreign affiliate’s projected outcomes, which were used to price its buy-in payment.

Table 1. Cost-Sharing Scenario Foreign Affiliate PCT Payment Calculation

Year

1

2

3

4

5

6

7

8

Sales due to existing IP

$100

$90

$81

$73

$66

$59

$53

$48

Total CSA-related sales

$100

$120

$119

$120

$124

$132

$145

$162

Exploitation expenses — cost of goods sold

$50

$60

$59

$60

$56

$60

$65

$73

Exploitation expenses — general and administrative

$15

$18

$18

$18

$19

$20

$22

$24

RAB share of R&D costs

$8

$10

$9

$10

$10

$11

$12

$13

Operating profits

$27

$32

$32

$32

$40

$42

$46

$52

Discount rate

10%

 

 

 

 

 

 

 

Buy-In = NPV of projected operating profits from existing IP

$114

 

 

 

 

 

 

 

Note: Dollars in millions.

Table 2 shows the foreign affiliate’s actual results.

Table 2. Cost-Sharing Scenario Foreign CSA Affiliate Financial Results

Year

1

2

3

4

5

6

7

8

Sales due to existing IP

$100

$92

$85

$78

$72

$66

$61

$56

Total CSA-related sales

$100

$125

$126

$127

$133

$143

$157

$176

Exploitation expenses — cost of goods sold

$50

$63

$63

$57

$60

$64

$71

$79

Exploitation expenses — general and administrative

$15

$17

$18

$20

$22

$24

$27

$29

RAB share of R&D costs

$8

$10

$10

$10

$11

$11

$13

$14

Operating profits

$27

$36

$35

$40

$41

$43

$47

$53

Note: Dollars in millions.

These results differ from the projections. Most notably, the taxpayer earned $322 million in operating profits, versus $304 million that it projected at the start of year 1. (These numbers are the totals of the eight years of operating profits in tables 1 and 2.) The IRS examines each year. In year 1 the IRS tests the buy-in payment using a CUT method, under which a 25 percent royalty rate was found to be arm’s length. The IRS also believes that the existing IP will generate 10 percent more revenue than the taxpayer projected, and it uses this revised estimate to calculate the PCT value. In this example, the PCT payment is deemed compliant because the IRS’s median value of that payment of $110 million shown in Table 3 is slightly less than the actual $114 million PCT payment made (Table 1), and within the interquartile range of CUTs.

Table 3. Cost-Sharing Scenario Examination PCT Calculation Using CUT Method

Year

1

2

3

4

5

6

7

8

CUT (25% royalty on rev for existing IP * 110%)

$28

$25

$22

$20

$18

$16

$15

$13

NPV of royalty income on projected profits

$110

 

 

 

 

 

 

 

Note: Dollars in millions.

In this example, unlike the Facebook litigation (or Veritas or Amazon), the IRS also performs a CWI test under reg. section 1.482-7(i)(6) in each year. In the first seven years, the actual results are considered compliant with the CWI standard because no periodic trigger occurs (that is, the AERR is no higher than 1.5). However, in year 8 this test shows the following as shown in Table 4 (using calculations prescribed by reg. section 1.482-7(i)(6)).

Table 4. Cost-Sharing Scenario Periodic Trigger Test

Year

1

2

3

4

5

6

7

8

Investments (PCT plus IDCs)

$122

$10

$9

$10

$10

$11

$12

$13

NPV of cumulative investments

$157

 

 

 

 

 

 

 

Divisional profits = rev - non-CSA costs

$35

$46

$45

$50

$51

$54

$60

$68

NPV of divisional profits

$262

 

 

 

 

 

 

 

AERR

1.7

 

 

 

 

 

 

 

Note: Dollars in millions.

Because the taxpayer’s AERR exceeds the fixed threshold of 1.5, the foreign affiliate’s aggregate income from the CSA is no longer CWI. Its investments and reasonably anticipated benefit (RAB) share of research and development costs for the CSA activity are too low to allow the increase in profits to remain within the 1.5 range of the PRRR. The affiliate is earning too much income on its CSA investments. In this example, the IRS chooses to apply a year 8 periodic adjustment, shown in Table 5. The $24 million periodic adjustment is the result of compounding forward the year 1 present-value adjustment of $11 million. The periodic adjustment will also be applied to years after year 8. In this example, the initial periodic adjustment is small. However, having triggered the periodic adjustment, the future foreign profits from the CSA activity will be determined according to a residual profit-split method, illustrating how a small violation of the CWI standard can result in much larger adverse consequences.

Table 5. Cost-Sharing Scenario Periodic Adjustment Calculation

Year

1

2

3

4

5

6

7

8

Adj = Residual Profits = Divisional profits - investments - routine return (8 percent * non-CC, non-IDC costs)

$(100)

$20

$18

$23

$23

$24

$27

$31

NPV in year 1 of cumulative adjustments for all years

$11

 

 

 

 

 

 

 

Periodic adjustment

$24

 

 

 

 

 

 

 

Note: Dollars in millions.

This stylized example shows that it is necessary to perform a periodic adjustment calculation under reg. section 1.482-7(i)(6) each year after a PCT, even if the payment for that PCT is deemed compliant according to valuation methods described by reg. section 1.482-7(g) in the year of payment. Likewise, this example shows why the IRS shouldn’t enter into advance pricing agreements that waive the CWI standard (as may be the case with Google Inc. in 2006).17 Projected results will almost never equate to actual results and will sometimes be vastly different. Because of information asymmetry, examination bias, and other factors, both the IRS’s and the taxpayer’s projected results (used as the basis for the PCT value) will almost always be wrong. The CWI standard of reg. section 1.482-7(i)(6) measures how wrong those projections were, according to a fixed standard. If the taxpayer is not monitoring its results for compliance with the CWI standard, and its foreign profit results exceeded the standard, the IRS would not know that unless it performed the CWI test.

B. Reasons for the IRS’s Reluctance

One possible reason the IRS has not applied the CWI test in any documented examinations of CSAs could relate to its continuing faith in its chosen strategy for pricing PCTs in these arrangements. A decision by the Office of Chief Counsel (International) to disregard the Tax Court’s guidance on how to value PCT payments under what is now reg. section 1.482-7(g), and to seek to sustain the IRS’s own method despite repeated rejection by the courts, could be a factor. This is especially true since the Tax Cuts and Jobs Act codified the IRS’s aggregation approach to valuing PCTs. Notably, the periodic adjustment is also an aggregated approach because it measures the results of all combined PCTs in aggregate.

The primary issue in Facebook’s case is identical to that in Veritas: the IRS’s use of an aggregation approach to inflate by some multiple the value of the taxpayer’s existing IP that was contributed to the CSA. The IRS chose to litigate Facebook’s challenge to the PCT adjustment rather than allow it to pursue an IRS Appeals process, which is consistent with the agency’s strategy. Consider the following from the IRS’s December 2010 action on decision in Veritas:

The Court also criticized the Service’s application of an aggregate valuation of the interrelated intangibles and services transactions, what the Court referred to as the “akin to a sale” theory. . . . Again, the opinion suggests the Court believed there was an absence in the factual record of a basis for an aggregate valuation under Treas. Reg. section 1.482-1(f)(2)(i). The Service, however, will continue to apply an aggregate valuation to interrelated transactions related to a CSA where, under the facts and circumstances, such valuation provides the most reliable measure of an arm’s length result.18 [Emphasis added.]

The following quote is from an experienced practitioner’s comments on the action on decision:

Following the loss in the Veritas Software case, the IRS published an [action on decision] in which the IRS disagreed with both the factual findings and the legal conclusions of the Tax Court. IRS officials have indicated that they are looking for other cases to litigate the issues that were lost in Veritas Software. . . . The Tax Court rejected the full range of arguments set forth in the cost-sharing [coordinated issues paper published in 2007]. . . . Thus, the IRS argued all projected income in excess of a routine return that was assumed to arise in the future for an indefinite period was deemed to be attributable to the transferred platform contributions. The Tax Court rebuffed these arguments pointing out that intangible property subject to section 482 was a statutorily defined term that excluded, for example, income attributable to services, income arising from goodwill and going concern value, and income attributable to the efforts of the transferee. . . .

The basic reason for those losses was that the legal theories upon which the IRS adjustments were based are not sound.19 [Emphasis added.]

Notably, none of these controversies exist in the context of a periodic adjustment, which is based on actual results and not conjecture (whether unfounded or not). Regardless, an insistence on prevailing on now-codified transactional approaches to valuing buy-in payments may have caused the IRS to give precedence to its own litigation strategies rather than heed the congressional intent that the CWI standard be applied to base annual adjustments on actual future returns. In Facebook’s case especially, the IRS should reconsider that litigation strategy if (1) the reg. section 1.482-7(i)(6) CWI standard applies and provides more objective and sustainable results, and (2) the factual record suffers the same deficiencies that caused previous courts to reject the IRS’s strategies.

Other reasons may also contribute to the IRS’s reluctance to apply the periodic adjustment rules. In a 2010 paper describing the history of the cost-sharing regulations under reg. section 1.482-7, Yariv Brauner notes that resource and capability constraints on the IRS may be an issue:

Over the years, the IRS has attempted to limit the avoidance potential in the [cost-sharing] regime, while keeping up the arm’s length rhetoric, with limited success. The IRS resorted to particular anti-abuse amendments rather than a substantive restructuring of the regime. Eventually, it is apparent that the IRS does not feel confident that it can effectively scrutinize [CSAs], so it shifted its enforcement focus from the arrangements themselves to the preliminary buy-in transactions. The theory was that correct buy-in valuation will leave little abuse potential for the [CSAs] themselves. Also, in practice the bulk of transfer pricing controversies involved buy-ins and their valuation.20 [Emphasis added.]

The IRS alluded to this as a potential issue in an October 2016 court filing, in which it described its attempts to hire an expert to assist with its examination of Facebook.21 IRS capability constraints are well documented. I have previously investigated how the IRS’s approach to corporate examinations produces a preponderance of errors, with the agency mistakenly challenging a compliant transaction as noncompliant while failing to detect noncompliance.22 The Facebook examination appears to exhibit both errors: mistakenly challenging a compliant buy-in payment, and failing to detect a violation of the CWI standard and its possible causes.

Part of this is attributable to resource, information, and capability asymmetries with taxpayers and their advisers. Moreover, the agency’s deficient organizational design puts pressure on individual examiners to identify and investigate highly complex noncompliance with little supporting detection technology or specialized forensic investigative support. I am aware of one instance of a transfer pricing examiner overseeing 40 examinations simultaneously. This is a “design for failure” problem that explains an almost complete collapse in effective enforcement.

At least in Facebook’s case, the application of the periodic adjustment rules would appear to be less resource-intensive or technically challenging than the more complicated and speculative approach the IRS has undertaken, and it would provide a potentially more reliable (that is, more objective) compliance assessment. The IRS should use this test to not only assess the reasonableness of the proposed adjustment but also to confirm compliance with the CWI standard, as Congress intended.

C. CWI Test Difficult to Challenge

Courts have affirmed Treasury’s implementation of the CWI standard in the context of cost-sharing rules. They have found that Treasury’s rulemaking complied with the Administrative Procedure Act (APA) and that the rules themselves were reasonable and not arbitrary and capricious under the act.23 It would appear difficult to challenge the CWI rules on any procedural grounds; however, there are no known applications of the periodic adjustment rules or any judicial decisions concerning them.

As written, the rules allow the IRS to impose, for a CSA participant, periodic adjustments that reflect all PCTs concerning that participant, even if there is just one PCT that triggers the periodic adjustment. The trigger occurs when that participant has made payments for the particular PCT and records divisional profits that exceed the sum of its CCs and its PCT payments by a pre-defined ratio. In appropriate situations concerning the trigger test or the periodic adjustment itself, one could reasonably maintain that computations should involve only the PCT that triggered the periodic adjustment, and not all PCTs and other factors such as any non-PCT transactions, including those that occur in years after the trigger. In such a case, one could claim, for example, that the value of the divisional profits used to calculate the AERR should be appropriately reduced. Although that claim has some attraction, the broad statutory authority for the CWI standard gives Treasury and the IRS considerable regulatory latitude.

One argument against this line of reasoning is that because cost sharing is a voluntary election with acceptance upfront of all its many conditions, the taxpayer knew what it was getting into and could have instead chosen a different mechanism to develop and transfer IP (such as a license agreement) and simply abided by the arm’s-length standard. Moreover, controlled taxpayers that carefully monitor the results of PCT participants can reasonably ensure that they meet the conditions for the 10-year exception mentioned earlier. Once that exception is met, there can be no periodic adjustments in any later years.24 This exception can be seen as providing taxpayers that elect CSA treatment an escape route from the potential effects of periodic adjustments. Thus, those electing taxpayers are not being treated unfairly.

In summary, the reg. section 1.482-7(i)(6) periodic adjustment approach appears in several ways consistent with the arm’s-length principle. Further, the CWI standard is clearly stated, and compliance is rewarded through the 10-year exception. Both the trigger mechanism and the periodic adjustments are measured according to a substantially mechanical and immutable standard. Moreover, the acceptance of CSA terms and conditions is optional to the taxpayer, which has full and transparent knowledge that the CWI standard will apply. The regulation accomplishes the intent of Congress and has been found by the courts to not be arbitrary or capricious in principle. For these reasons, the CWI standard appears substantially immune to challenge and would certainly be simpler and less resource intensive than the IRS’s current approaches to PCTs.

IV. Facebook: Economic Factors

Focusing on Facebook’s case, the IRS made an adjustment based on economic factors, several of which Facebook contests. However, there are other economic factors that have not been discussed in any of the public materials reviewed for this report. One of them is the substantial difference between the average revenue per U.S. user and that of foreign users. Foreign users generate only a fraction of the advertising revenue per user compared with U.S. users, yet foreign users now represent around 90 percent of all of Facebook’s users.

If Facebook’s total costs globally were allocated according to users (which assumes that Facebook’s total costs benefit all users equally on a per-user basis), its foreign business might be much less profitable.25 This would cast doubt on the IRS’s valuation of approximately $20 billion for the foreign rights to Facebook’s existing IP in September 2010, which represented as much as 70 percent of Facebook’s entire market value of around $29 billion at that time, or its valuation of the loss-making foreign business at more than twice that of the profitable U.S. business, which was earning more than $1 billion a year.26 The divergence in the advertising revenue per daily average user (DAU) in the United States versus foreign locations, shown in Table 6, casts substantial doubt on the magnitude of the IRS’s buy-in payment valuation.

Table 6. Calculation of Facebook Foreign Revenues Per Daily Average User as a Percentage of U.S. Valuea

Year

Foreign Revenue (dollars in millions)
(a)

Foreign DAUs (millions)
(b)

Foreign Revenue Per DAU (dollars)
(c) = a/b

U.S. Revenue (dollars in millions)
(d)

U.S. DAUs (millions)
(e)

U.S. Revenue Per DAU (dollars)
(f) = d/e

Ratio of Foreign to U.S. Revenue Per DAU
(g) = c/f

2010

$751

197

$3.80

$1,223

80

$15.24

25%

2011

$1,644

326

$5.04

$2,067

106

$19.53

26%

2012

$2,511

452

$5.55

$2,578

118

$21.86

25%

2013

$4,259

584

$7.29

$3,613

128

$28.17

26%

2014

$6,817

709

$9.62

$5,649

138

$41.03

23%

2015

$9,415

839

$11.22

$8,513

148

$57.44

20%

2016

$15,059

998

$15.09

$12,579

158

$79.46

19%

2017

$22,919

1,180

$19.42

$17,734

165

$107.75

18%

2018

$31,738

1,318

$24.07

$24,100

166

$145.04

17%

2019

$40,467

1,439

$28.13

$30,230

169

$179.28

16%

aBased on revenues as reported in SEC filings, and estimates of U.S. and foreign daily average users (DAU) based on calculations shown in Appendix.

Another economic factor was the rapid obsolescence of the covered buy-in platform technology, as illustrated in Figure 1. The figure shows that within two years of the CSA buy-in payment, users began abandoning the desktop platform in order to access the Facebook applications using mobile devices, for which Facebook had no native revenue-generating technology in 2010. Users were already migrating to mobile devices by 2010, but it was by no means assured that Facebook could profitably transition to a mobile environment.

Figure 1

The smartphone was still a relatively new phenomenon in 2010. At the time, mobile operating systems on the market included Research in Motion’s BlackBerry operating system (OS), Apple Inc.’s iOS, Nokia’s Symbian OS, Google’s Android OS, and Microsoft’s Windows Phone 7 OS. In the face of a fragmented and nascent market at such an early stage of its existence, Facebook would likely have found it difficult to develop and efficiently operate this many native mobile apps. It was not until approximately 2012 that Facebook was able to begin to generate advertising revenue from its mobile applications — and likely because it had become feasible by then to develop separate native mobile platforms for only the Android OS and iOS. Those two systems had effectively become a duopoly by then, because most hardware manufacturers (aside from Apple) transitioned to Android OS, and the Research in Motion and Microsoft OSs declined in popularity. This made it easier for Facebook to successfully transition to the mobile environment as quickly as it did.

Another economic factor is that Facebook’s foreign operations experienced losses in 2010 and for years afterward. The company’s relevant financial results for the first five years of the CSA are presented in Table 7, based on SEC filings. This table also shows an alternative “before transfer pricing” result by assuming that total costs globally were allocated by territory according to the number of DAUs in those territories — as might be expected if there is an assumption that resources and expenses primarily benefit the Facebook platform and that platform usage is similar across territories on a per-user basis. Notably, the foreign losses are greater after Facebook’s transfer pricing exercise compared with a global allocation of total costs based on users. This is consistent with the large outlays by Facebook Ireland in the early years for its buy-in payments including PCTs.

Table 7. Facebook Financial Results by DAU in Foreign Locations 2010-2014

Year

Reported Foreign Revenue
(a)

Reported Foreign DAUs
(b)

Global Total Cost Per DAU (dollars)
(c)

Allocation of Total Costs by DAU
(d) = b * c

Implied Foreign Pretax Income Before Transfer Pricing
(e) = a - d

Reported Foreign Pretax Income
(f)

2010

$751

197

$3.39

$670

$81

$(19)

2011

$1,644

326

$4.52

$1,476

$168

$(124)

2012

$2,511

452

$7.98

$3,609

$(1,098)

$(568)

2013

$4,259

584

$7.12

$4,155

$104

$(443)

2014

$6,817

709

$8.83

$6,256

$561

$(8)

Totals

$15,982

n/a

n/a

$16,167

$(185)

$(1,162)

Note: Numbers in millions, except where noted.

The combination of all these factors — (1) the IRS adjustment that values the loss-making foreign IP exploitation rights at more than twice the value of Facebook’s profitable U.S. business in 2010; (2) the five years of foreign losses after the buy-in payment transaction; (3) the fractional value per user of the advertising revenue for foreign users compared with U.S. users, which did not cover the average total global costs per user; and (4) the obsolescence of the desktop platform as a revenue driver after 2010 — is important to consider. They suggest that despite strong user growth, the foreign profit opportunity was likely nowhere near the value of Facebook’s existing U.S. business in 2010. The $6.7 billion buy-in payment was about 23 percent of the company’s market value at the time — very close to the proportion of foreign revenue per DAU compared with U.S. users (noting that this revenue did not cover the costs per user). On its face, this buy-in payment does not seem unreasonable.

V. Indicative Periodic Adjustment

Publicly available court filings by Facebook and the IRS, in combination with other sources, contain enough information to perform an indicative periodic adjustment calculation. The IRS does not appear to have performed such a calculation before adjusting Facebook’s 2010 buy-in payment in 2016 or increasing that adjustment in 2019. The IRS would have had six years of tax returns by October 2016, and nine years of returns by early 2020, with which it could have performed this calculation on a rolling basis.

The following exercise relies on some estimates — most notably for the payment schedule for the buy-in payment — that are not reported in the various court documents but would be known to the IRS and Facebook. The purpose of this exercise is to make a reliable estimate of the periodic adjustment calculation that would result if the IRS performed it as advised in this report. Ordinarily, a reliable periodic adjustment estimate using outside information would be difficult because most large multinational taxpayers have more than one CSA, and published financial results lack sufficient information (such as PCT payments) to perform the calculation. However, the multitude of Tax Court filings by Facebook and the IRS provide numerous actual results that appear to allow a reasonable calculation. The information in Table 8 is derived from those various court documents.

Table 8. Information Derived From Facebook and IRS Tax Court Petitions

Variable

Source

Item

Value

NPV of Facebook PCT Payment

Facebook 2016 Petition

(a)

$1,685,644,726

NPV of Facebook User Base IP Payment

Facebook 2016 Petition

(b)

$4,078,192,896

Sum of NPV Value of All IP

Facebook 2016 Petition

(c)

$6,700,000,000

IRS Proposed NPV Value of All IP

Facebook 2016 Petition

(d)

$13,883,630,000

IRS Adjustment to NPV of the Buy-In

Facebook 2016 Petition

(e)

$7,183,630,000

2010 IRS Tax Adjustment

Facebook 2016 Petition

(f)

$1,733,335

2010 IRS Income Adjustment

Facebook 2016 Petition

(g)

$4,952,386

2010 FB Ireland Royalty Payment for Buy-In IP

IRS Oct 2019 Filing

(h)

$99,999,999

2010 FB Ireland Payment for PCT IP

IRS Oct 2019 Filing

(i)

$59,561,111

2010 FB Ireland Payment for User Base License

IRS Oct 2019 Filing

(j)

$37,626,671

2010 FB Ireland Royalty for Marketing IP (1 percent)

IRS Oct 2019 Filing

(k)

$2,812,217

2010 FB Ireland Cost Sharing Payments

IRS Oct 2019 Filing

(l)

$21,160,536

2010 IRS PCT (Royalty) Adjustment

Facebook 2016 Petition

(m)

$84,915,248

2011 IRS PCT (Royalty) Adjustment

Facebook 2018 Petition

(n)

$184,679,049

2012 IRS PCT (Royalty) Adjustment

Facebook 2018 Petition

(o)

$927,104,321

2013 IRS PCT (Royalty) Adjustment

Facebook 2018 Petition

(p)

$983,577,956

2012 R&D Related Stock-Based Compensation

Facebook 2018 Petition

(q)

$60,572,252

2013 R&D Related Stock-Based Compensation

Facebook 2018 Petition

(r)

$756,021,034

2013 Instagram PCT Lump Sum Payment

Facebook 2018 Petition

(s)

$152,000,000

2013 Instagram User Base IP Lump Sum Paid

Facebook 2018 Petition

(t)

$85,000,000

IRS Adjustment to 2013 Instagram PCT (Royalty)

Facebook 2018 Petition

(u)

$830,662,614

Facebook Ireland 2010 RAB Share

Facebook 2016 Petition

(v)

43%

Facebook 2010 Foreign Revenue Percentage

Facebook Form 10-k

(w)

38%

Multiyear Payment Schedule for IP

Facebook 2016 Petition

(x)

Six years

Facebook Expert Upper WACC Rate Applied

Facebook 2020 Pre-trial Memo

(y)

17.28%

Facebook Expert Lower WACC Rate Applied

Facebook 2020 Pre-trial Memo

(z)

16.28%

Average of Facebook Expert High & Low WACC

Calculation = (y + z)/2

(aa)

16.78%

IRS ExPert Median WACC Rate Applied

Facebook 2020 Pre-trial Memo

(bb)

14.4%

Average of FB Expert Ave and IRS WACC

Calculation = (aa + bb)/2

(cc)

15.59%

Taxes Paid for Share-Based Comp in Q2 2019

Facebook Form 10-Q 2020

(dd)

$1,100,000,000

The figures from Table 8 were used to create a hypothesized schedule of PCT payments. This is shown in Table 9, where the actual figures are shaded, and all other figures were estimated according to information provided in the court documents. In particular, the payment schedule is estimated by ensuring that the sum of nominal buy-in payments yields an aggregate result on an NPV basis that matches the NPV of the figures reported by Facebook in court documents. Those documents provide NPV figures for the IRS adjustment and the Facebook PCT and user base IP. They also mention a six-year payment term. Therefore, the PCT payments are estimated as increasing over a six-year payment term by the same percentage increase each year measured from year 1, with the fixed annual percentage increase determined based on a Microsoft Solver routine that equated the sum of the payments as discounted to their aggregate NPV reported in the court filings. This payment schedule calculation is still a hypothesis, of course, because the actual payment schedule was not reported (only the first year’s payment and the NPV of total payments were reported). The model applies a discount rate of 15.6 percent, which is shown in row (cc) of Table 8, which was the average of the discount rates proposed by the IRS and Facebook’s experts.

Table 9. Calculation of Hypothesized PCT Payments Per Court Filings

 Year

Nominal IRS IP (Royalty) Adjustment

NPV of IRS Adjustment

Nominal User Base and Marketing Royalty

NPV User Base IP Payments

Hypothesized Nominal PCT Payment

NPV of PCT Payments (excluding Instagram)

2010

$84,915,248

$84,915,248

$40,438,888

$40,438,888

$59,561,111

$59,561,111

2011

$184,679,049

$159,770,784

$89,287,369

$77,244,890

$100,300,850

$86,772,948

2012

$927,104,321

$693,885,924

$197,142,765

$147,550,374

$168,906,530

$126,417,126

2013

$983,577,956

$636,865,900

$435,282,955

$281,845,347

$284,438,425

$184,173,641

2014

$1,870,549,174

$1,047,823,356

$961,086,507

$538,370,711

$478,994,020

$268,317,523

2015

$3,557,373,559

$1,723,963,847

$2,122,038,695

$1,028,376,113

$806,625,442

$390,904,435

2016

$6,765,342,932

$2,836,404,941

$4,685,372,429

$1,964,366,573

$806,625,442

$569,497,942

Sum

$14,373,542,239

$7,183,630,000

$8,530,649,607

$4,078,192,896

$3,257,182,805

$1,685,644,726

The example in reg. section 1.482-7(i)(6) used for this analysis contains a simplifying assumption that the CSA was entered into at the beginning of the year, and that the first year is year 1. All tables in this report assume that the initial buy-in payments were made at the end of 2010, with this also being the date for the 2010 present value of the calculations, when required. The discounting conventions follow the regulations and examples. Also, separate inputs were needed to perform a periodic adjustment calculation. Those additional inputs are shown in Table 10 under the columns marked “used.” When available, inputs provided in court filings were used. Stock-based compensation (SBC) was based on SEC filings multiplied by RAB share, and this matched the results provided in court filings for specific years. Facebook Ireland revenue and profits were estimated based on calculations using SEC filings and Facebook Ireland’s financial filings (which were reported in euros and converted to dollars at the midyear rate). The last column in Table 10 was used to verify that the $1.1 billion in taxes paid to the IRS in 2019 for SBC approximated the amounts in arrears since 2010. The estimates in Table 10 were derived based on the use of at least two different sources, when possible.

Estimates for the periodic adjustment calculation include CCs and IDCs, both of which depend on Facebook Ireland’s RAB share. (The actual CCs or IDCs were disclosed for only one year in the court filings.) In its 2016 Tax Court petition, Facebook reported that it used a unique RAB share calculation that differs substantially from a more conventional RAB share calculation based on territory revenue. Facebook instead defined the RAB share as follows:

Each party’s RAB share is defined as the ratio of the net present value of the aggregate gross profit of one party divided by the net present value of the aggregate total gross profit of both parties, where gross profit is defined as gross profits in the current fiscal year plus gross profits projected for the two following fiscal years.27

Table 10 calculates the RAB share using this method, but with the gross profits calculated using a revenue-based allocation of total cost of services as reported in SEC filings. The table is based on estimates and is expected to differ at least slightly from statutory (tax) results, which are proprietary to Facebook and the IRS.

Table 10. Supporting Information for Periodic Trigger Calculation (dollars in millions)

 

 

Used

 

 

 

 

 

Used

 

 

Used

 

Used

 

Used

 

(a)

(b)

(c)

(d)

(e)

 

(f)

(g)

(h) = e - g

(i)

(j)

(k) = i * j

(l)

(m) = (h/f) * l

(n)

(o) = l * n

(p)

Year

Discount Term

Discount Factors

Approx. FB Ireland Ltd. Revenues Per Ireland Filings

Total FB Rev. Per SEC Filings

U.S. Rev. Per SEC Filings

Foreign Rev. Per SEC Filings

Est. U.S. And Canada Rev.

Est. FB Ireland CSA-Related Territory Revenues Per SEC Filings

Est. Ireland RAB Percentage Per Calculation In Petition

Total R&D Costs Reported In SEC Filings

Est. FB Ireland RAB Share of R&D Expense

Foreign Pretax Profits Per SEC Filings

Est. Ireland Pretax Profits

R&D Share-Based Comp (SBC) Per SEC Filings

Ireland Share of SBC Based On RAB Percentage

Est. U.S. Taxes on Ire. SBC Payments at Statutory Rate (net of TCJA Repatriation Tax)

FY2010

0

1.00

$304

$1,974

$1,223

$751

$1,358

$218

43%

$144

$21

$(19)

$(6)

$9

$4

$0.8

FY2011

1

0.93

$1,465

$3,711

$2,067

$1,644

$2,294

$1,417

49%

$388

$191

$(124)

$(107)

$114

$56

$11

FY2012

2

0.86

$2,301

$5,089

$2,578

$2,511

$2,862

$2,227

52%

$1,399

$731

$(568)

$(504)

$61

$32

$6

FY2013

3

0.80

$3,957

$7,872

$3,613

$4,259

$4,010

$3,862

52%

$1,415

$738

$(443)

$(402)

$756

$394

$79

FY2014

4

0.74

$6,428

$12,466

$5,649

$6,817

$6,270

$6,196

53%

$2,666

$1,400

$(8)

$(7)

$1,328

$698

$140

FY2015

5

0.68

$8,768

$17,928

$8,513

$9,415

$9,449

$8,479

54%

$4,816

$2,577

$3,392

$3,055

$2,350

$1,257

$251

FY2016

6

0.63

$13,991

$27,639

$12,580

$15,059

$13,964

$13,675

55%

$5,919

$3,239

$6,150

$5,585

$2,494

$1,365

$273

FY2017

7

0.59

$21,142

$40,649

$17,730

$22,919

$19,680

$20,969

56%

$7,754

$4,304

$13,515

$12,365

$2,820

$1,565

$313

FY2018

8

0.54

$29,835

$55,838

$24,100

$31,738

$26,751

$29,087

56%

$10,273

$5,749

$16,561

$16,561

$3,022

$1,691

$101

FY2019

9

0.50

$38,786

$70,697

$30,230

$40,467

$33,555

$37,142

56%

$13,600

$7,657

$19,495

$17,893

$3,488

$1,964

$118

WACC

7.9%

Sum

$126,977

$243,863

$108,283

$135,580

$120,194

$123,270

n/a

$48,374

$26,608

$57,951

$53,050

$16,442

$9,026

$1,294

Tables 8, 9, and 10 were used to create the first calculation described in reg. section 1.482-7(i)(6)(v) for determining whether a periodic trigger exists in the first five years of Facebook’s CSA. Table 11 follows the presentation in Example 1(ii) of the regulation. That calculation does not include any payments for SBC, whose inclusion was required by reg. section 1.482-7(d)(1)(iii) but was excluded by Facebook pending resolution of Altera Corp.’s litigation challenging this rule. Altera was decided in favor of the IRS in June.28 The overall results in each year were “trued up” based on the estimated Facebook Ireland pretax income in that year by adjusting the non-CC/non-IDC costs. (These intercompany charges are typically performed under different transfer pricing rules such as reg. section 1.482-4 or -9, but the IRS may not have examined them, based on information in the cited court documents.) The applicable discount rate (ADR) used was the Facebook weighted average cost of capital (WACC) prevailing in the determination year (2014), in accordance with reg. section 1.482-7(i)(6)(iv)(B).

Table 11. Calculation of Indicative Facebook Periodic Trigger Results for 2010-2014 (dollars in millions)

 

Year

(a)

(b)

(c)

(d)

e1

e2

(e) = e1 + e2

(f) = d + e

(g) = b - c

(h) = g/f

 

Foreign Pretax Profit: Periodic Adjustment Calc vs. SEC Filings

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/ Non-IDC Costs (Est.)

Reported CC/ IDCs (Est.)

Buy-In PCT Paid (Est.)

Additional PCT Paid

Total PCT Paid = e1 + e2

Est. Total Investment Costs

Est. Divisional Profit (Loss)

AERR = PVTP/PVI

AERR < 0.667

Pretax Income Calc = b - c - d - e

Est. FB Ireland Pretax Income Using SEC Filings

FY2010

1

$218

$143

$21

$60

$—

$60

$81

$75

 

 

$(6)

$(6)

FY2011

2

$1,417

$1,232

$191

$100

$—

$100

$292

$185

 

 

$(107)

$(107)

FY2012

3

$2,227

$1,832

$731

$169

$—

$169

$900

$396

 

 

$(504)

$(504)

FY2013

4

$3,862

$3,088

$738

$284

$152

$436

$1,175

$773

 

 

$(402)

$(402)

FY2014

5

$6,196

$4,324

$1,400

$479

$—

$479

$1,879

$1,872

 

 

$(7)

$(7)

NPV thru year 1

$218

$143

$21

PCT portion of buy-in payment

 

 

 

 

Instagram PCT

 

 

 

 

$60

$81

$75

0.932

 

Sum of Pretax Profit

NPV thru year 2

$1,531

$1,284

$199

$153

$351

$246

0.702

 

$(1,025)

$(1,025)

NPV thru year 3

$3,443

$2,857

$826

$298

$1,124

$586

0.522

Yes

NPV thru year 4

$6,515

$5,314

$1,413

$645

$2,058

$1,201

0.584

Yes

 

 

NPV thru year 5

$11,083

$8,501

$2,446

$998

$3,444

$2,582

0.750

 

WACC

7.92%

Based on this exercise, the AERR appears to fall below the PRRR lower bound of 0.667 in at least two of the first five years of the CSA, causing what we will refer to as a “downward” trigger.29 More important, the AERR is already within, and near the lower limit of, the PRRR. If this were to be confirmed using internal tax information available only to the IRS and Facebook, it would mean that the PCTs (including the buy-in PCT) within this period would have been CWI, if not slightly overpriced. Note that a downward periodic trigger that occurs within the first five years, beginning with the first year of substantial exploitation of cost-shared intangibles, would trigger the five-year-period safe harbor under reg. section 1.482-7(i)(6)(vi)(B)(2) and thus prevent the IRS from applying an upward periodic adjustment for five years from the first year of substantial exploitation of the IP. Presumably this would also cast doubt on the reasonableness of an upward adjustment to the initial PCT.

While these calculations exactly follow the regulations, most of the key inputs are estimates based on partial information. This is therefore an “indicative” calculation, designed to provide a reasonable estimate of the outcome of a periodic adjustment calculation if performed by the IRS based on more complete information.

As noted, the IRS proposed a substantial increase to Facebook’s buy-in payment. This included apportioning to Facebook Ireland a share of Facebook’s SBC costs related to its U.S. R&D functions. As also noted, the IRS recharacterized the entire buy-in payment as a PCT. Accordingly, Table 12 was recalculated by moving the user base IP and marketing IP buy-in payment to the PCT column from the non-CC/non-IDC column, increasing the value of the PCTs by the amount of the IRS adjustments, and including SBC. Table 12 is based on the initial IRS adjustment of $7.2 billion proposed in 2016 (and therefore assumes a periodic adjustment determination year of 2016) and does not include the additional $6 billion adjustment it proposed in 2019. The resulting profit figures in the exercise now differ from the reported profit figures because we are no longer measuring Facebook’s reported results and instead are measuring the effect of the IRS’s adjustments, which clearly reduce foreign and Irish profits.

Table 12. Calculation of Indicative Facebook Periodic Trigger Results for 2010-2014 Incorporating IRS Proposed PCT Adjustments (dollars in millions)

 

Year

(a)

(b)

(c)

(d)

e1

e2

e3

e4

(e) = e1 + e2 + e3 + e4

(f) = d + e

(g) = b - c

(h) = g/f

 

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs Excl. UBMI (Est.)

CC/IDCs Including SBC (Est.)

Original Buy-In PCT Payments (Est.)

Move UBMI From Non-CC

IRS Increase to Original PCT

IRS Increase to Instagram PCT

Total New PCT Payments (Est.)

Est. Total Investment Costs (Est.)

Divisional Profit (Loss) (Est.)

AERR = PVTP/PVI

AERR < 0.667

FY2010

1

$218

$102

$25

$60

$40

$85

 

$185

$210

$116

 

 

FY2011

2

$1,417

$1,142

$248

$100

$89

$185

 

$374

$622

$274

 

 

FY2012

3

$2,227

$1,634

$762

$169

$197

$927

 

$1,293

$2,056

$593

 

 

FY2013

4

$3,862

$2,653

$1,133

$436

$435

$984

$831

$2,686

$3,819

$1,208

 

 

FY2014

5

$6,196

$3,362

$2,098

$479

$961

$1,871

 

$3,311

$5,409

$2,833

 

 

FY2015

6

$8,479

$(81)

$3,834

$807

$2,122

$3,557

 

$6,486

$10,320

$8,560

 

 

FY2016

7

$13,675

$(1,192)

$4,603

$1,358

$4,685

$6,765

 

$12,809

$17,412

$14,867

 

 

FY2017

8

$20,969

$4,300

$5,869

$—

$—

$—

 

$—

$5,869

$16,669

 

 

FY2018

9

$29,087

$8,160

$7,441

$—

$—

$—

 

$—

$7,441

$20,927

 

 

FY2019

10

$37,142

$11,592

$9,621

$—

$—

$—

 

$—

$9,621

$25,550

 

 

NPV thru year 1

$218

$102

$25

 

 

 

 

$185

$210

$116

0.551

Yes

NPV thru year 2

$1,444

$1,091

$255

 

 

 

 

$509

$748

$353

0.472

Yes

NPV thru year 3

$3,111

$2,314

$909

 

 

 

 

$1,477

$2,286

$797

0.348

Yes

NPV thru year 4

$5,611

$4,032

$1,810

 

 

 

 

$3,216

$4,759

$1,579

0.332

Yes

NPV thru year 5

$9,082

$5,615

$2,719

 

 

 

 

$5,070

$7,789

$3,929

0.406

Yes

NPV thru year 6

$9,082

$5,876

$4,577

 

 

 

 

$8,213

$12,790

$7,314

0.572

Yes

A check calculation shown in Table 13 was used to measure the output on an NPV basis to the total IRS adjustment reported in the Facebook petition. This check calculation uses the 15.6 percent discount rate because that was the rate used to measure the NPV of the IRS adjustments. This check calculation also excludes the SBC adjustment and the Instagram adjustment because they were not included in the NPV of the IRS adjustments.

Table 13. Quality Control Check of Table 12 Calculations vs. Reported IRS Adjustments

Year

Pretax Income Calc: x = b - c - d - e

y = Est. FB Ireland Pretax Income Using SEC Filings

Calc = y - x - Instagram adj - SBC adj

NPV of Calc

FY2010

$(94)

$(6)

$85

$85

FY2011

$(348)

$(107)

$185

$160

FY2012

$(1,463)

$(504)

$927

$694

FY2013

$(2,610)

$(402)

$984

$637

FY2014

$(2,575)

$(7)

$1,871

$1,048

FY2015

$(1,760)

$3,055

$3,557

$1,724

Sum

$(11,396)

$7,614

$14,374

$7,184

NPV of IRS adjustment (excludes SBC and Instagram adj)

$7,184

WACC

15.59%

 

 

 

These calculations show that the IRS adjustment increases Facebook’s foreign losses, and this causes the AERR to fall below the 0.667 lower limit of the PRRR in every year of the six-year payment term mentioned in the IRS petition — excluding the SBC (which, when added, would further reduce the AERR below the PRRR). The regulation does not recognize this situation as a periodic trigger and simply ignores its implications (that is, the periodic trigger is deemed not to have occurred when the AERR is below the PRRR for purposes of the five-year-period safe harbor, as described in reg. section 1.482-7(i)(6)(vi)(B)(2)).

In any case, the above results showing an AERR below the PRRR in all six years of the PCT payment term strongly suggest that the IRS adjustment itself is inconsistent with the CWI standard. Although it is not required to do so, one would think that the IRS should adhere to the CWI standard as reflected in reg. section 1.482-7(i)(6), just as taxpayers are required to do. The CWI analysis represented by the periodic adjustment calculation is not a hypothesized adjustment as under reg. section 1.482-7(g), in which the IRS and the taxpayer do not agree on the inputs or the results. A periodic adjustment calculation does not have this problem, because it relies on actual results.

VI. Facebook’s Results Since 2014

From 2014 through 2019, Facebook’s results changed dramatically, to the extent that by 2019 Facebook reported approximately 80 percent of its total pretax income in foreign jurisdictions, mostly in low-tax or no-tax jurisdictions.30 Non-U.S. results are shown in Table 14 based on SEC filings, using the same calculations as shown for earlier years.

Table 14. Facebook Financial Results in Foreign Locations vs. Total Cost Apportionment by DAU 2015-2019

Year

Reported Foreign Revenue

Reported Foreign DAUs

Global Total Cost Per DAU ($)

Foreign Share of Total Costs

Implied Foreign Pretax Income Before Transfer Pricing

Reported Foreign Pretax Income

(a)

(b)

(c)

(d) = b * c

(e) = a — d

(f)

2015

$9,415

839

$11.85

$9,946

$(531)

$3,392

2016

$15,059

998

$13.16

$13,128

$1,931

$6,150

2017

$22,919

1,180

$15.21

$17,947

$4,972

$13,515

2018

$31,738

1,318

$20.83

$27,464

$4,274

$16,561

2019

$40,467

1,439

$29.06

$41,810

$(1,343)

$19,495

Totals

$119,598

n/a

n/a

$110,295

$9,303

$59,113

Notes: Dollars in millions, except as noted.

Whereas Facebook reported an aggregate foreign loss between 2010 and 2014, between 2015 and 2019 it reported an aggregate foreign pretax income of $59 billion. Importantly, the results show an almost $50 billion increase in foreign profits compared with the counterfactual result obtained if global total expenses were allocated to the U.S. and foreign affiliates (at cost) based on the number of users. The reported aggregate foreign profit margin between 2015 and 2019 is now 49 percent, compared with only 33 percent in the United States over the same period. Facebook’s financial metrics in 2019 for both U.S. and foreign Facebook locations are shown in Table 15, based on SEC filings.

Table 15. Facebook Comparative U.S. and Foreign Financial Results in 2019

 

Non-U.S.

Non-U.S.

U.S./Non-U.S.

Revenue and income

Revenues

Revenue per DAU ($)

Pretax income

$30,230

$179

$5,317

$40,467

$28

$19,495

75%

637%

27%

Expenses

Total expenses borne

Total expenses per DAU ($)

$25,739

$153

$20,972

$15

123%

1,047%

Profit margins

Margin on costs per DAU

Margin on revenue per DAU

21%

18%

93%

48%

22%

37%

Taxes

Tax provisiona

Tax rate

$2,385

45%

$1,461

7%

163%

598%

Notes: Dollars in millions, except as noted.

aExcludes FTC settlement and share-based compensation.

The results shown in Table 16 add years through 2019 to the earlier periodic adjustment calculations and incorporate the same assumptions and inputs. These results assume no further PCT payments by Facebook Ireland beyond those disclosed in Facebook’s June 2018 Tax Court filing. However, these results do include income in 2019 representing the payment of all SBC in arrears, according to the following comments in the Facebook Form 10-Q for the quarter ended June 30, 2020:

Based on the Altera Ninth Circuit Opinion [June 7, 2019], we recorded a cumulative income tax expense of $1.11 billion in the second quarter of 2019. . . . Since we started to accrue income tax for stock-based compensation cost-sharing in the second quarter of 2019, the denial of the request by the Supreme Court did not have a material impact to our financial results in the second quarter of 2020.31

It is not clear if the $1.11 billion in taxes represented recognition of all SBC payments required by Facebook Ireland, but this is assumed in the tables that follow, which reflect payment of all SBC by Facebook Ireland in 2019. The complete calculations are shown in Appendix 3. The discount rate used for these calculations is Facebook’s 2019 WACC of 9.75 percent as reported in a public financial database GuruFocus.com, which includes a detailed description of its WACC formula, all calculations, and inputs.

Table 16. Calculation of Indicative Facebook Periodic Trigger Results for 2010-2019 (dollars in millions)

Year

(a)

(b)

(c)

d1

d2

(d) = d1 + d2

e1

e2

(e) = e1 + e2

(f) = d + e

(g) = b - c

(h) = g/f

 

 

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs (Est.)

Stock-Based Comp

Reported CC/IDCs (Est.)

New CC/IDCs (Est.)

Buy-In PCT Paid (Est.)

Additional PCT Paid

Total PCT Paid = e1 + e2

Est. Total Investment Costs

Est. Divisional Profit (Loss)

AERR = PVTP/PVI

AERR < 0.667

AERR > 1.5

FY2010

1

$218

$143

$—

$21

$21

$60

$—

$60

$81

$75

 

 

 

FY2011

2

$1,417

$1,232

$—

$191

$191

$95

$—

$95

$287

$185

 

 

 

FY2012

3

$2,227

$1,832

$—

$731

$731

$152

$—

$152

$883

$396

 

 

 

FY2013

4

$3,862

$3,088

$—

$738

$738

$243

$152

$395

$1,134

$773

 

 

 

FY2014

5

$6,196

$4,324

$—

$1,400

$1,400

$389

$—

$389

$1,790

$1,872

 

 

 

FY2015

6

$8,479

$2,041

$—

$2,577

$2,577

$623

$—

$623

$3,199

$6,438

 

 

 

FY2016

7

$13,675

$3,493

$—

$3,239

$3,239

$995

$—

$995

$4,234

$10,182

 

 

 

FY2017

8

$20,969

$4,300

$—

$4,304

$4,304

$—

$—

$—

$4,304

$16,669

 

 

 

FY2018

9

$29,087

$8,160

$—

$5,749

$5,749

$—

$—

$—

$5,749

$20,927

 

 

 

FY2019

10

$37,142

$11,592

$9,026

$7,657

$16,683

$—

$—

$—

$16,683

$25,550

 

 

 

NPV thru year 1

$218

$143

RAB share of R&D-related SBC per SEC filings

 

 

 

 

 

 

 

 

 

Estimates per court docs and SEC filings

 

 

 

 

 

 

 

 

 

$21

Increase by same percentage amount each year for six years from year 0

 

 

 

 

 

 

 

 

 

Instagram PCT

 

 

 

 

 

 

 

 

 

$60

$81

$75

0.932

 

 

NPV thru year 2

$1,509

$1,265

$196

$146

$342

$244

0.713

 

 

NPV thru year 3

$3,358

$2,786

$802

$273

$1,075

$572

0.532

Yes

 

NPV thru year 4

$6,279

$5,122

$1,361

$572

$1,933

$1,157

0.599

Yes

 

NPV thru year 5

$10,549

$8,102

$2,326

$840

$3,166

$2,447

0.773

 

 

NPV thru year 6

$15,873

$9,383

$3,944

$1,231

$5,175

$6,490

1.254

 

 

NPV thru year 7

$23,697

$11,382

$5,797

$1,856

$7,598

$12,316

1.621

 

Yes

NPV thru year 8

$34,629

$13,623

$8,041

$1,856

$9,841

$21,005

2.134

 

Yes

NPV thru year 9

$48,445

$17,499

$10,772

$1,856

$12,572

$30,946

2.461

 

Yes

NPV thru year 10

$64,519

$22,516

$17,992

$1,856

$19,792

$42,003

2.122

 

Yes

This indicative periodic adjustment trigger calculation shows that an upward periodic adjustment may have been triggered around 2017, or possibly earlier. Note that if Facebook has not complied with the documentation requirements of reg. section 1.482-7(k), a lower AERR threshold ratio of 1.25 applies, which, according to the calculations above, could result in a periodic trigger as early as 2015 (noting that these requirements of reg. section 1.482-7(k)(2)(ii)(J)(4) arguably include documentation regarding compliance with the periodic adjustment rules). Importantly, the result in any given year is cumulative, thereby taking into account both underpriced and overpriced PCTs. Therefore, this accounts for the overpriced buy-in PCTs in earlier years. Facebook’s payment of SBC expenses in 2019 is heavily discounted in the calculation and therefore has little impact on the results. If Facebook made undisclosed PCT contributions after the Instagram acquisition, changed the CSA, or added new CSA participants, these indicative results might differ substantially from a periodic adjustment trigger calculation by the IRS with complete information.

One of the variables affecting this result is the discount rate. A higher discount rate will reduce the AERR and the probabilities of a periodic trigger. According to reg. section 1.482-7(i)(6)(iv)(B), the appropriate discount rate is the taxpayer’s WACC, unless the taxpayer convinces the commissioner that a different rate is more reliable:

Publicly traded companies. If the PCT Payor meets the conditions of paragraph (i)(6)(iv)(C) of this section, the ADR is the PCT Payor WACC as of the date of the Trigger PCT. However, if the Commissioner determines, or the controlled participants establish to the satisfaction of the Commissioner, that a discount rate other than the PCT Payor WACC better reflects the degree of risk of the CSA Activity as of such date, the ADR is such other discount rate. (C) Publicly traded. A PCT Payor meets the conditions of this paragraph (i)(6)(iv)(C) if — (1) Stock of the PCT Payor is publicly traded.

One of the main reasons the appropriate discount rate could vary from the WACC would be that the CSA pertained to a project or product line whose probabilities for success were riskier than the company as a whole, and if the business as a whole has a financing structure that reflects both debt and equity, but the risky project may not be suitable for debt financing because of its greater riskiness. This does not appear to apply to Facebook, because (1) its CSA applies to all its business offerings and divisions as a whole, and (2) Facebook did not have any material debt financing in any of the determination years.32 Nevertheless, an exercise was performed to determine if the discount rate might affect the result. Both the 2017 and 2019 models predicted a (weak) periodic trigger in 2016. The results with different discount rates applied are shown in Table 17.

Table 17. Simulation Results for Table 16 With Different Discount Rates

 

2017 Determination Year

2019 Determination Year

Discount rates

10%

No change

No change

20%

No change

No change

25%

No change

Trigger in 2016 eliminated

30%

Trigger in 2016 eliminated

No change

35%

No change

No change

40%

No change

No change

45%

No change

No change

50%

No change

No change

55%

No change

Trigger in 2017 eliminated

60%

Trigger in 2017 eliminated

No change

66%

n/a

Trigger in 2019 eliminated

The predicted periodic trigger in 2017 is not eliminated until the discount rate reaches 60 percent, and the 2019 periodic trigger is not eliminated until this rate reaches 66 percent. These rates would be exceedingly unrealistic. Notably, by 2019 the AERR appears extreme, to a point that it is unclear without additional information whether any adjustments of the size necessary to change this outcome may be possible.

If the IRS were to apply a periodic adjustment in a given year, that year and every subsequent year would be subject to a periodic adjustment that would result in the application of a residual profit split. In the absence of any substantial nonroutine activities in Ireland, this would effectively repatriate most of the Irish profits back into Facebook US. The earlier periodic adjustment occurring in 2016 (or more likely in 2017) is important because, if sustained, all subsequent years would be subject to a periodic adjustment. Importantly, according to reg. section 1.482-7(i)(6)(i), the IRS can apply the periodic trigger in any of the open years to which they might apply, regardless of whether the year was in fact the earliest year exhibiting a periodic trigger:

A periodic adjustment under this paragraph (i)(6) may be made without regard to whether the taxable year of the Trigger PCT or any other PCT remains open for statute of limitations purposes or whether a periodic adjustment has previously been made with respect to any PCT Payment.

The results above raise a question: If the buy-in payment was overpriced, what might cause a periodic trigger in later years? This question, investigated extensively in my recent study with Yaron Lahav,33 appears to be connected to how Facebook allocates its overall (and mostly U.S.-based) operating expenses. The study found that total costs borne by foreign affiliates in aggregate are essentially “right-sized” to the lower foreign revenue per daily average user, compared with a standard cost allocation based on users. As a result, costs per daily average U.S. user exceed costs per foreign user by 622 percent by 2019, and the foreign profits per user exceed U.S. profits by more than 300 percent.

Despite revenue per foreign user measuring only 16 percent of revenue per U.S. user by 2019, the profit margins per foreign user exceed those in the United States. This seems impossible empirically, unless Facebook incurs the majority of its total expenses as a result of users that make up only 10 percent of its total user base. One of the economic aspects of Facebook’s business model — as with any network — is a fixed cost base (that, a fixed number of data centers, servers, and employees) with scalable revenue as new users join the network. It is not clear according to any external evidence why such a large proportion of what are essentially fixed costs mostly benefitting the network as a whole are so tilted toward the much smaller base of U.S. users.

It is clear from SEC filings and other sources that Facebook conducts operations primarily from the United States, and Facebook US manages and operates the business and its internet-based platforms. These personnel, operations, and facilities are used to deliver services to all users and advertisers within and outside the United States (exclusively until 2013, when Facebook built its first data center outside the United States). However, as late as year-end 2019, U.S. long-term assets and personnel were around 80 percent and 70 percent, respectively, and U.S. operational data center capacity was closer to 90 percent or more of the global totals for these resources.34 Facebook, of course, maintains thousands of personnel in foreign countries, but it is understood that the bulk of that personnel is engaged in local marketing and sales support, customer support, administrative activities, site monitoring, and other mostly routine localized functions.

The point is that Facebook’s commercial operations occur on one integrated platform and infrastructure, and it is not intuitive why the total cost of acquiring, hosting, and advertising to the billions of users that generate Facebook’s revenue should differ according to the tax rate of the country where the revenue is booked. It does appear that, on the surface, foreign operations are bearing only a fraction of the operating costs that would be borne if costs were allocated (with appropriate profit elements) according to the number of users versus revenue or other metrics.

The results of this exercise indicate that Facebook may fail to qualify for the 10-year-period safe harbor under reg. section 1.482-7(i)(6)(vi)(B)(1). Based on these calculations, Facebook appears to qualify for the five-year-period safe harbor under reg. section 1.482-7(i)(6)(vi)(B)(2), but that does not protect it from periodic adjustments in any subsequent years.35 The periodic adjustment rules also contain four additional exceptions in reg. section 1.482-7(i)(6)(vi)(A)(1) through (4) that could prevent a periodic adjustment. These exceptions include, for example, a sale of the same IP to a third party, or a result in a year that reverses the periodic trigger in an earlier year. Facebook does not appear to qualify for any of these exceptions, based on information in the court filings.

The indicative periodic adjustment calculation performed here shows that there could be a risk of a substantial adjustment by approximately 2017. The periodic adjustment calculations are shown in appendices 3 and 4 for determination years 2017 (the first determination year in which the models predict a periodic trigger) and 2019 (the most recent determination year), respectively. The calculations in those appendices are based on two important assumptions: (1) that Facebook Ireland performs only routine activities, and (2) that the arm’s-length return for these routine activities is 8 percent of the costs of performing them.36

VII. Additional Approaches

Regardless of what the IRS chooses to do regarding a periodic adjustment, it would be useful to consider other tax mechanisms that might apply to low-substance CSAs. One item to consider would be the economic substance of the arrangement. It has been well documented that the IRS, when examining CSAs, appears to focus on the buy-in payment while often ignoring the economic substance of these transactions — especially the supply chain and exploitation aspects of these arrangements.

Facebook’s CSA, particularly in its earlier years, exhibits elements of a low- or no-substance CSA. That is, the U.S. parent performs virtually all the IP development and IP exploitation activities, and the foreign CSA participant merely books the profits of those activities in return for cost-sharing payments, but with no ability to exploit the IP itself. Often in this type of arrangement the only financial benefits will be tax benefits. Also, the arrangement will be devoid of business purpose because there are few or no activities conducted by the foreign affiliate that materially increase pretax profits apart from tax savings at the net profit level (which is why the CSA participant is almost always located in a tax haven or a low-tax jurisdiction such as Ireland).

Likewise, such an arrangement in which the foreign revenue and profits are factually generated by U.S. activities is likely to create effectively connected income. In these cases, the foreign affiliate will often have a de facto agency in the United States that exercises material control over the risks or functions that generate the foreign revenue and profits.37 A lack of economic substance, or the recognition of ECI, could result in a partial or complete denial of the tax benefits of the CSA, essentially achieving a result similar to a periodic adjustment — something that would effectively corroborate a CWI periodic adjustment.

A. Economic Substance Doctrine

According to its Irish financial filings, Facebook Ireland, at the time Facebook implemented its CSA arrangement, had about 150 employees in Dublin (up from 64 the year before) and no revenue-generating assets (that is, data centers), and the housing and processing of all customer information and advertising operations were performed in the United States. The Irish affiliate relied fully on Facebook US for not only IP development but also for IP exploitation. Virtually all IP exploitation occurred in the United States, with Facebook US performing all commercial business functions, and there was no change in operational facts before or immediately after the arrangement.

Further, before Facebook Ireland’s formation in 2008, Facebook US had launched a “translation app” in more than 100 countries that, in combination with innovations in the performance of its U.S.-based communications network, jump-started immediate exponential user growth in those foreign locations.38 By the time the January 2009 IP license agreement between Facebook US and Facebook Ireland was executed, foreign users already outnumbered U.S. users, and foreign user numbers were increasing far more rapidly — with no efforts from Facebook Ireland, which had just opened its Dublin office and hired its first director of online operations that very month.39 Foreign users also continued to connect directly to Facebook’s U.S.-based infrastructure, with only a disclaimer added to its website visible to foreign users that the services were henceforth being “offered” by Facebook Ireland.40

It appears that Facebook’s overall results would have been unchanged regardless of whether Facebook Ireland existed. It also appears that Facebook’s success overseas was driven by U.S. activities and infrastructure, and definitely not by any activities occurring in Ireland, where there were perhaps only a few dozen employees in 2009 and perhaps around 150 by September 2010, most of whom had been recently hired. (This was still more than Google’s five Irish employees around the time of its CSA transaction in 2003 — a transaction that was the model for the Facebook arrangement and was approved by the IRS with an APA.41)

The Senate Homeland Security and Governmental Affairs Permanent Subcommittee on Investigations, in a 2014 majority staff report, concluded that the IRS should examine the economic substance of transfer pricing arrangements and identify any transactions to which the economic substance doctrine might apply.42 Instead, the IRS continued with its long-standing approach to CSAs of examining only the buy-in payment under rules in reg. section 1.482-7(g).43 It seems clear that this was exactly what Facebook and its advisers expected the IRS to do, and they planned accordingly. The economic substance doctrine was codified in section 7701(o) in March 2010, only months before Facebook’s buy-in payment. It states:

In the case of any transaction to which the economic substance doctrine is relevant, such transaction shall be treated as having economic substance only if — (A) the transaction changes in a meaningful way (apart from Federal income tax effects) the taxpayer’s economic position, and (B) the taxpayer has a substantial purpose (apart from Federal income tax effects) for entering into such transaction.

Interestingly, the pre-2009 cost-sharing regulations would have required threshold levels of economic substance in the Irish affiliate. That requirement was removed in the 2009 regulations.44 Although the pre-2009 regulations specifically allowed a U.S. taxpayer to enter into a CSA with a foreign shell or holding company, conventional wisdom was that this was based on the foreign participant’s ability to check the box for foreign operating subsidiaries that could exploit the transferred IP, which would provide sufficient substance to the arrangement. This is not what is happening in a low-substance CSA. In those arrangements, both IP development and IP exploitation occur in the United States, and there is insufficient substance offshore to exploit the cost-shared IP.

For this reason, both the Google and Facebook CSAs (one approved by the IRS and the other challenged), which were based on the same approach, appear to meet the technical definition of a tax shelter under section 6662(d)(2)(C), according to emails, other sources in the public domain, and a recent decision by the federal district court in Seattle.45 Some commentators recently challenged the assumption that a CSA can be a tax shelter.46 Their primary claim is that a CSA is a “congressionally approved and regulatorily mandated quasi-safe harbor” that merely constitutes a means to “account for a fundamental aspect of ordinary multinational business.” Their article refers to cost sharing as a way to prevent licensing disputes among unrelated parties. This was clearly theoretical; a CSA does not resemble any market-based co-development or joint R&D agreement. In short, no rational company would ever do this with a third party (especially a low-substance version), and there are no third-party examples of CSAs as described by the regulation.

A CSA is most certainly not an arrangement “in the ordinary course of business,” and is so artificial that the statute places guardrails in the form of an explicit requirement that the arrangement contain economic substance and comply with the CWI standard. Reg. section 1.482-7(k)(1)(iv)(A) states: “The provisions will be interpreted by reference to the economic substance of the transaction and the actual conduct of the controlled participants.” (Emphasis added.)

The IRS historically has not enforced these guardrails, and taxpayers have taken advantage of this to effectively implement tax shelters and transactions lacking economic substance disguised as CSAs. The regulations clearly did not intend this type of implementation. For instance, none of the approximately 40 carefully crafted examples in reg. section 1.482-7 show anything resembling a low- or no-substance CSA. In fact, every one of them shows a foreign cost-sharing participant performing exploitation activities with its own resources in its own territory, versus relying on the U.S. parent for both IP development and IP exploitation in the United States, with only a shift in the location of revenue and profits to the low-substance foreign affiliate upon implementation.47

My earlier research shows the ways in which a low-substance CSA (L-S in the following table) differs from a third-party arrangement.48 Those findings are reproduced in Figure 2 with additional explanations.

Figure 2

The CSA regulations contemplate that each CSA participant itself exploits the cost-shared intangibles in its territory or pays an arm’s-length price.49 For example, a putative participant determines its share of IDCs based on its “reasonably anticipated benefits” from its own exploitation of those cost-shared intangibles. If that putative participant lacks an ability to exploit the intangibles itself through its own personnel, facilities, and assets, the agreement should not represent a valid CSA.50 Interestingly, the OECD/G-20 base erosion and profit-sharing project explicitly recognized that a party cannot be a valid participant “if it does not exercise control over the specific risks it assumes.”51

It seems doubtful that recently established Facebook Ireland had any capability to either control the risks being assumed through the putative CSA agreements or exploit the cost-shared intangibles, which factually continued to be exploited almost solely by Facebook US. Even without considering section 7701(o), this nonexploitation by Facebook Ireland is enough for the IRS, should it so desire, to treat Facebook’s (and possibly Google’s) arrangement as not being a valid CSA.

One reason for codifying the economic substance doctrine in section 7701(o) was to provide greater clarity and uniformity in the application of this long-existing doctrine and thus help prevent unintended consequences that can arise in the United States’ highly technical rule-based tax system. As such, this codification was meant to eliminate or reverse tax avoidance transactions that rely on the interaction of technical tax law provisions to produce tax consequences not contemplated by Congress. This should include many of the ubiquitous profit-shifting arrangements conducted by multinationals that have no identifiable business purpose aside from tax avoidance and involve no changes in economic position or functions. These arrangements typically involve paper transactions that shift U.S. profits to a low-tax jurisdiction such as Ireland.

An analysis of the economic substance of the Facebook CSA based on public information strongly suggests that the only benefits to the arrangement were tax benefits. There appear to be no observable nontax benefits, such as material reductions in operating expenses, operational synergies, or functions that might have expanded Facebook’s pretax profits beyond what was already occurring before the arrangement. The tax avoidance motive is supported by at least one internal email as previously noted. The activities undertaken by Facebook Ireland in its territory, as documented by the Irish Data Protection Commission in 2011, were mostly non-revenue-generating and cannot in any manner explain or justify the amounts of profits booked in Ireland. The observable benefits of the arrangement appeared to be tax savings and nothing more.

The IRS Large Business and International Division issued a memorandum in 2011 requiring examiners to assess whether application of the economic substance doctrine is appropriate in a given situation, using a four-step process.52 The LB&I directive listed 17 facts and circumstances that tend to show that application of the economic substance doctrine may be appropriate, including the following:

  • the transaction has no credible business purpose apart from federal tax benefits;

  • the transaction has no meaningful profit potential apart from tax benefits;

  • the transaction is not at arm’s length;

  • the transaction creates no meaningful economic change;

  • the transaction is promoted, developed, or administered by outside advisers; and

  • the transaction results in separation of income between different taxpayers.

It takes only one of the listed factors to trigger a violation of section 7701(o). As many as six (those listed above) may reasonably apply to the Facebook CSA, and likely Google’s CSA used as a template and possibly marketed and sold by the same public accounting firm, based mostly on public information.

One reason why the IRS may be reluctant to examine the economic substance of a CSA is that the LB&I guidance contains an important caveat: If the transaction is a “statutory or regulatory election” (which, presumably, a CSA is), an examiner should not further pursue application of the economic substance doctrine without “specific approval of the examiner’s manager in consultation with local counsel.” First, this consideration would appear to apply only to the extent that the “regulatory election” itself results in a transaction devoid of economic substance. However, U.S. transfer pricing rules in general, and reg. section 1.482-7 in particular, explicitly require economic substance. Further, the requirement to coordinate with IRS counsel is not a prohibition on applying section 7701(o). A CSA with no economic substance violates both reg. section 1.482-7 and section 7701(o) and would fail to qualify for the benefits of a CSA under both the regulation and the statute.

It is unknown if the examination team on the Facebook tax audit consulted with IRS counsel to apply the economic substance doctrine to Facebook’s CSA, or, if it did, what counsel may have advised. What seems clear from the publicly available documents is that the IRS did not address the economic substance doctrine in its examination of Facebook. If that is the case, the only possible inference is that by choosing to examine the buy-in payment, the IRS effectively concluded or assumed that the arrangement was in fact a valid CSA that did not violate section 7701(o) or the economic substance requirements of reg. section 1.482-7.

It seems clear that the IRS should consider initiating or revisiting the issue of economic substance, as requested by Congress,53 as part of its strategy for examining low- or no-substance CSAs implemented by Facebook and other taxpayers. Such a review is especially appropriate for transactions implemented by or with the assistance of the taxpayer’s public auditors, which effectively blocks the transactions from any independent auditor review.

Those situations can lead to breaches in auditor independence. On this point, the promotion, sale, implementation, and self-auditing of a tax shelter by a registered accounting firm on behalf of its audit client is not expressly prohibited by any IRS or SEC regulation. However, this activity is strongly advised against as a best practice by the SEC as part of its auditor independence rules implementing the Sarbanes Oxley Corporate Accountability Act of 2002. Those rules were published May 6, 2003 — approximately two months before Google’s public auditor is widely believed to have facilitated a no-substance cost-sharing tax arrangement54 with an Irish entity that was set up only months earlier and functionally resembled a shell company.55 In January 2009 the same accounting firm, at its audit client Facebook, assisted in implementing a no-substance IP licensing arrangement with a constructive Irish shell company that the taxpayer claimed in its 2016 tax court petition was instrumental to growing Facebook’s foreign user base over the licensing period. Shortly thereafter, in September 2010, the auditor helped convert the “no-substance” licensing transaction into a low-substance CSA.

B. The ECI Rules

Section 864(c) and various sourcing laws and regulations provide the principal rules that allow the United States to directly tax the business income of a foreign corporation when that income is attributable to business activities conducted in the United States by the taxpayer or its agents. That taxable income is ECI. Also, in some instances, specific deductions and credits may be disallowed, and the section 884 branch profits tax may apply.

Under the factual manner in which a low- or no-substance CSA is conducted, there is a high likelihood that the U.S. and foreign affiliates within a group will be conducting joint business activities that create an unintended partnership under the entity classification rules. Importantly, the existence of a partnership not only triggers partnership tax return filing and withholding obligations, it also makes the application of the ECI rules much easier.56

Facebook US and Facebook Ireland conduct a centrally managed joint worldwide business from the United States that is seamless to their vendors, users, advertisers, and other customers and third parties. The same internet-based platform, along with the personnel who manage and maintain it primarily from within the United States, service those vendors, users, advertisers, etc., irrespective of their locations in the world. This conduct of joint business activities with common management and common business functions almost certainly creates a partnership for federal tax purposes under the entity classification rules. With the bulk of the partnership’s digital services being conducted in the United States, some significant portion of the partnership income will be ECI directly taxable in the hands of Facebook Ireland in its capacity as a foreign corporate partner. Because the taxpayer is Facebook Ireland, which is not a member of the Facebook affiliated group that presumably files a consolidated federal tax return, all tax years would remain open unless Form 1120-F, “U.S. Income Tax Return of a Foreign Corporation,” has been filed for each year.57

Note that ECI taxation could be applied to Facebook Ireland concurrently with the ongoing transfer pricing litigation. Depending on the outcome of the litigation, the amount of ECI for each tax year could be adjusted to reflect the final adjudicated income within Facebook Ireland.

C. Disparate IRS Approaches to CSAs

As discussed earlier, according to Facebook sources submitted to the court, the Facebook CSA was based substantially on an earlier Google CSA and involved at least one executive who had participated in implementing both arrangements. Also, the same accounting firm — the public auditor for both companies — assisted with the design and implementation of both the Facebook and possibly Google CSAs. Moreover, the earlier-mentioned email by a Facebook executive, in combination with other sources, commented directly on the Google arrangement in a way that identified it as essentially constituting a tax shelter.

Both arrangements involved (1) little to no economic substance in Ireland when implemented, (2) exclusive exploitation of U.S.-owned IP within the United States for the benefit of the foreign CSA participant, and (3) the recording in Ireland of nonroutine profits of U.S. activities when only routine activities occurred there. The IRS reportedly approved the Google arrangement three years after its implementation with an APA but challenged Facebook’s PCT payments. By these actions, the IRS was implicitly accepting the economic substance of these arrangements on their face.

Forensic economic analyses that include an indicative periodic adjustment calculation show that both results are extreme.58 Google reported only $49 million in foreign losses on $1.5 billion in foreign revenue in the first two years of its CSA. In contrast, Facebook reported more than $1 billion in foreign losses over five years, and from a larger revenue base — $2.5 billion in the first two years of its arrangement. Google has reported more than $130 billion in offshore profits since implementing its CSA — more profits than it reported in the United States. Likewise, Facebook has reported around $58 billion of offshore income since implementing its CSA, over a shorter period than Google, but this income is also more than it reported in the United States over the same period. The difference between the IRS examinations of Google and Facebook could not be starker. Clearly, the IRS is neither applying the same compliance standard to almost identical tax planning arrangements (implemented and overseen by some of the same personnel) nor examining the primary facts that are driving these results.

The IRS has a history of focusing primarily on the level of PCT and buy-in payments when examining CSAs, and no examples have been found in which the IRS applied the CWI-based periodic adjustments in either the 2009 CSA regulations or their predecessors. This appears to be a weakness, because a buy-in payment can omit or underprice critical IP. For instance, in addition to technology IP and user base as IP, Facebook mentioned its network IP (that is, the hardware and software system). Both Facebook and Google possess an installed proprietary communications network of interconnected data centers of massive proportions. In Google’s case, this was described the year after its APA as the “world’s largest supercomputer,” and the company described it as its “secret sauce.”59 The network itself and the valuable customer personal data residing on it were technically not defined by section 936(h)(3)(B), which is the IP referred to in a typical CSA buy-in payment/technology transfer contract (as with Facebook, according to court filings). The value of Google’s network and user data IP was well established by the time of its APA, but the subsequent CSA results do not comport with this value, at least relative to the IRS valuation of similar IP in Facebook, and according to an indicative CWI test as described in the 2009 CSA regulations (which would not apply to Google under successful application of the reg. section 1.482-7(m) transition rules).

Interestingly, Facebook reported in court filings that its user base IP relates to users with which Facebook has development contracts, or that lead specific developer communities. It is unclear if this IP also includes users’ personal data that is exploited by Facebook for targeted advertising revenue — IP that is extremely valuable to the Irish revenue of both Facebook and Google.

Although the disparities in the IRS approaches to these two CSA arrangements were vast, there were also similarities. In neither exam did the IRS appear to scrutinize the economic substance, ECI, or CWI aspects of these arrangements, although the facts clearly would have supported that scrutiny in each case. Similar conclusions are also found in the IRS examinations of CSAs at Apple, Amazon.com, Altera, and Veritas. These exams all exhibit features of errors that create “reverse enforcement” type outcomes, which have been deconstructed in numerous studies by academics and practitioners.60 The IRS’s approach in Facebook appears to be the latest installment.

VIII. Conclusion

This report has investigated the relevance of a periodic adjustment calculation under reg. section 1.482-7(i)(6) as both a check on the IRS’s adjustment to Facebook’s buy-in payment under reg. section 1.482-7(g) and as an explicit test for CWI that should be considered as a more definitive and objective compliance test, which taxpayers are required to meet in any case. The IRS should be held to the same standard.

This report has also explored whether additional economic and regulatory factors are supportive of the IRS’s results. A preliminary calculation based on known but incomplete data, with some estimation, finds that the buy-in payment paid by Facebook Ireland in 2010 was likely overpriced, and that the IRS’s adjustment appears to cause a violation of the CWI standard under reg. section 1.482-7(i)(6).

Several economic factors support that finding, including (1) sustained foreign losses by the Irish affiliate over the IRS examination period, (2) obsolescence of the controlled technology soon after the buy-in transaction, (3) substantially lower advertising revenue per foreign user compared with U.S. users that did not exceed the global average costs per user, and (4) the fact that the IRS’s valuation would have assigned a market value to Facebook’s foreign business more than twice that of its U.S. business (when only the U.S. business was profitable, earning almost $11 billion in the period 2010-2014, when Facebook reported only foreign losses). Indeed, this situation is very similar to the IRS’s challenge of a buy-in payment at Amazon.com in the face of sustained foreign losses, which was overturned by the Tax Court in 2017 and held to be arbitrary and capricious.

Although the periodic adjustment computations and supporting economic data indicate that the buy-in payment was likely not underpriced as the IRS alleges, extending this calculation to later years (which are not at issue in this case) reveals that Facebook’s CSA may have substantial risk of a periodic trigger beginning about 2017, and that by 2019 that risk is much clearer. The periodic adjustments from those triggers could be the bulk of Facebook Ireland’s $59 billion of accumulated profits. If these results are replicated with complete internal tax information, the IRS would be well advised to reverse its risky and likely unsupportable PCT adjustment and consider larger adjustments under the reg. section 1.482-7(i)(6) periodic adjustment/CWI rules that become relevant to later years. Likewise, Facebook may wish to also perform such a calculation itself to determine if it might be better to capitalize on what appears to be a mistake by the IRS and accept a smaller negotiated IRS adjustment now, which could qualify its CSA for the 10-year-period safe harbor. Qualifying for that safe harbor could enable Facebook to eventually recover its settlement payment with sufficiently aggressive transfer pricing, free from any CWI compliance requirement. This would allow it to maintain tax competitiveness with its rival, Google, which got a free pass from the IRS despite an even more extreme result. First, the IRS (and Facebook) should perform a periodic adjustment calculation to confirm whether these opportunities and threats truly exist.

IX. Appendix 1

Facebook DAU calculations derived from Facebook’s Form 10-K filings with the SEC are shown in Table A1.61 Estimates of users are also based on certain estimates.62

X. Appendix 2: Periodic Adjustment Calculations

TRA 1986 added a new concept to section 482, the section that allows the IRS to allocate income, deductions, credits, or allowances between commonly controlled entities to clearly reflect income. The new CWI concept was meant to deal with the difficulty of valuing the economic rights associated with various forms of IP by providing a statutory basis for future adjustments of the original transfer price of an intangible so that it reflects the actual income earned from the intangible in the years after its transfer and initial valuation. The 2009 temporary cost-sharing regulations and the 2011 finalized regulations (reg. section 1.482-7) included an objective trigger mechanism and a periodic adjustment calculation allowing the IRS to easily apply the CWI concept to CSAs. This mechanism is included in reg. section 1.482-7(i)(6).

Under reg. section 1.482-7(g), when one CSA participant makes one or more PCTs, other participants must pay for the PCTs. The valuation is made with knowledge and expectations as of the year in which the PCT is made. If future profits earned by one or more of the other participants exceed a calculated factor, reg. section 1.482-7(i)(6) allows the IRS to make periodic adjustments in those future years that have the effect of correcting the undervaluation under reg. section 1.482-7(g) that occurred in the earlier year of the PCT.

The following sections briefly describe the mechanism that triggers a future-year periodic adjustment, as well as how the periodic adjustment is calculated. They also highlight some exceptions set out in the regulation that allow taxpayers to avoid periodic adjustments in some years.

A. Trigger

Through a trigger mechanism, reg. section 1.482-7(i)(6) enables the IRS to apply a periodic adjustment to a taxpayer’s CSA when specified conditions exist. Namely, a trigger occurs when the AERR associated with PCT payments made by the parties to the arrangement falls outside the PRRR associated with those PCTs. This range is 0.667 to 1.5. (That range is narrowed to between 0.8 and 1.25 if the controlled participants have not substantially complied with documentation requirements in reg. section 1.482-7(k), which provides guidance on contractual requirements.) The AERR is defined as the present value of total profits from the CSA activity by the CSA partner (typically a foreign participant in the CSA that has made PCT payments to a U.S. participant) divided by the present value of CSA-related investments (which is the sum of the CCs and PCT payments from the CSA start date through the end of the adjustment year). When payments are made in both directions (that is, by both or all participants to each other), they are not netted before the calculation.

When the AERR falls out of the PRRR range, this is referred to as a periodic trigger. When a periodic trigger arises, the IRS may consider whether a periodic adjustment is needed that applies the CWI concept to the CSA. That application does not change the original price used in the year when the PCT occurred; it only adjusts income in the year of the determination date, and it is based on all the relevant facts and circumstances known as of the determination date (generally the date of the relevant periodic trigger). The IRS can make a periodic adjustment in any open tax year, regardless of whether the tax year in which the PCT payment occurred is closed and regardless of whether the PCT has already been subject to an earlier periodic adjustment.

Table A1. Calculation of U.S. and Foreign DAU Figures

Year

Q1

Q2

Q3

Q4

Average

U.S. and Canada DAUs (as reported)

2010

82

85

92

99

90

2011

105

117

124

126

118

2012

109

130

132

135

132

2013

139

142

144

147

143

2014

150

152

155

157

154

2015

161

164

167

169

165

2016

173

175

178

180

177

2017

182

183

185

184

184

2018

185

185

185

186

185

2019

186

187

189

190

188

Global DAUs

2010

234

257

293

327

278

2011

372

417

457

483

432

2012

526

552

584

618

570

2013

665

699

728

757

712

2014

802

829

864

890

846

2015

936

968

1,007

1,038

987

2016

1,090

1,128

1,179

1,227

1,156

2017

1,284

1,325

1,368

1,401

1,345

2018

1,449

1,471

1,495

1,523

1,485

2019

1,562

1,587

1,623

1,657

1,607

Estimated U.S. DAUs Based on U.S. vs. Canada Population

2010

74

76

83

89

80

2011

94

105

111

113

106

2012

116

117

118

121

118

2013

125

127

129

132

128

2014

135

136

139

141

138

2015

144

147

150

152

148

2016

155

157

160

161

158

2017

163

164

166

165

165

2018

166

166

166

167

166

2019

167

168

170

170

169

Estimated Foreign DAUs = Global DAUs Minus U.S. DAUs

2010

160

181

210

238

197

2011

278

312

346

370

326

2012

410

435

466

497

452

2013

540

572

599

625

584

2014

667

693

725

749

709

2015

792

821

857

886

839

2016

935

971

1,019

1,066

998

2017

1,121

1,161

1,202

1,236

1,180

2018

1,283

1,305

1,329

1,356

1,318

2019

1,395

1,419

1,453

1,487

1,439

Note: All figures in millions.

B. Calculation

If a periodic trigger is not subject to any of the specified exceptions, it is calculated in the following steps, which are set out in reg. section 1.482-7(i)(6)(v):

  1. Determine the present value, as of the date of the trigger PCT, of the PCT payments using the adjusted residual profit-split method as described in reg. section 1.482-7(i)(6)(v)(B).

  2. Convert the present value of the PCT payments determined in step 1 into a level royalty rate as a percentage of the reasonably anticipated divisional profits or losses over the entire duration of the CSA activity.

  3. Multiply the step 2 royalty rate by the actual divisional profit or loss for tax years preceding and including the adjustment year to yield a stream of contingent payments in each year, and convert these payments to a present value as of the CSA start date. If there is a loss in any particular year, the result will be a negative contingent payment for that year.

  4. Convert any actual PCT payments through the adjustment year to a present value as of the CSA start date and subtract this present value from the present value result determined in step 3. This difference is the amount in present value terms of the participant’s deficiency in PCT payments. Then calculate a nominal amount in the adjustment year that would have a present value as of the CSA start date equal to this deficiency in PCT payments. This nominal amount is the periodic adjustment to be recognized by the recipient of the PCT payments in the adjustment year.

Note that step 4 actually includes three steps. In the appendices that follow, step 4 is separated into steps 4a, 4b, and 4c for clarity.

After a periodic adjustment has been calculated for a particular year, two further steps under the regulations explain how to apply the periodic adjustment calculation to subsequent years:

  1. Apply the step 2 royalty rate to the actual divisional profit or loss for each tax year after the adjustment year, through the determination date, to calculate amounts for each year. Then subtract from each year’s amount any actual PCT payment made for the same year. The differences are the periodic adjustment for each respective tax year.

  2. Apply the step 2 royalty rate to the actual divisional profit or loss for each tax year after the year that includes the determination date. This yields the periodic adjustment for each such tax year.

A step-by-step illustration of a CSA calculation (including tables) is provided in reg. section 1.482-7(i)(6)(vii), Example 1. That example was the basis for the calculations shown in the following appendices, each of which provides a calculation that assumes that the year at issue is the year of the original determination date. Before developing those tables, the discounting procedures were checked against the IRS examples (because Example 1 includes a simplifying assumption in which all financial flows are at the beginning of the year, which is unusual). As mentioned earlier, for a public company such as Facebook, the ADR is required to be the company’s WACC, unless a different discount rate can be shown to be more reliable.

C. Exceptions

There are four exceptions to a periodic trigger, and two situations in which a periodic trigger is deemed to not have occurred. The four exceptions are described in reg. section 1.482-7(i)(6)(vi)(A)(1)-(4), and the two situations are described in reg. section 1.482-7(i)(6)(vi)(B)(1)-(2).

The first exception involves a PCT in which the platform contribution is the same contribution that was previously transferred in an unrelated transaction under similar conditions and pricing, and that transaction is used as the basis for pricing using the CUT method described in reg. section 1.482-7(g)(3). The second exception applies when the periodic trigger was attributable to extraordinary events beyond the control of the controlled participants that could not reasonably have been anticipated at the time of the trigger PCT.

The third exception applies when the periodic trigger would not have occurred had the PCT payer’s divisional profits or losses that were used to calculate the periodic trigger excluded profits or losses attributable to the PCT payer’s (1) routine contributions to its exploitation of cost-shared intangibles, (2) nonroutine contributions to the CSA activity; (3) operating CCs, and (4) routine platform contributions. The fourth exception applies through a complicated calculation for situations in which a periodic trigger would not have occurred had the periodic trigger calculation included information relating to one or more years after the adjustment, as described in the third exception. Any projected amounts in the calculation are determined based on all information available as of the determination date.

No periodic triggers are deemed to occur in (1) all years after the first 10 years of the CSA activity (beginning with the first tax year in which there is substantial exploitation of cost-shared intangibles resulting from the CSA) if no periodic trigger was generated in any of those first 10 years, and (2) the first five years (beginning with the first tax year in which there is substantial exploitation of cost-shared intangibles resulting from the CSA) if in any of those first five years the AERR falls below the lower bound of the PRRR (that is, a downward trigger).

XI. Appendix 3: Periodic Adjustment Tables (2017)

The tables in this appendix follow the calculation in reg. section 1.482-7(i)(6)(vii), Example 1, applying a periodic trigger calculation for 2017. They were constructed to be as close as possible to the tables shown in the regulatory example. According to the regulations, all calculations for the determination year must use the taxpayer’s WACC applicable to that year. Facebook’s WACC for 2017 is reported in a financial website as 5.83 percent. All the following tables in this appendix use that WACC.

A. Step 0

The AERR calculation is made using information from Facebook court filings; SEC filings; public information, with some assumptions about the schedule of buy-in payments; and other information and assumptions, excluding the IRS’s adjustment to Facebook’s buy-in payment.

Table A3.1. Calculation of Indicative Facebook Periodic Trigger Results in 2017 Determination Year (dollars in millions)

 

Year

(a)

(b)

(c)

d1

d2

(d) = d1 + d2

e1

e2

(e) = e1 + e2

(f) = d + e

(g) = b - c

(h) = g/f

 

 

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs (Est.)

Stock-Based Comp

Original CC/IDCs (Est.)

New CC/IDCs (Est.)

Buy-In PCT Paid (Est.)

Additional PCT Paid

Total PCT Paid = e1 + e2

Est. Total Investment Costs

Est. Divisional Profit (Loss)

AERR = PVTP/PVI

AERR < 0.667

AERR > 1.5

FY2010

1

$218

$143

$—

$21

$21

$60

$—

$60

$81

$75

 

 

 

FY2011

2

$1,417

$1,232

$—

$191

$191

$92

$—

$92

$283

$185

 

 

 

FY2012

3

$2,227

$1,832

$—

$731

$731

$142

$—

$142

$872

$396

 

 

 

FY2013

4

$3,862

$3,088

$—

$738

$738

$218

$152

$370

$1,109

$773

 

 

 

FY2014

5

$6,196

$4,324

$—

$1,400

$1,400

$337

$—

$337

$1,737

$1,872

 

 

 

FY2015

6

$8,479

$2,041

$—

$2,577

$2,577

$519

$—

$519

$3,096

$6,438

 

 

 

FY2016

7

$13,675

$3,493

$—

$3,239

$3,239

$800

$—

$800

$4,039

$10,182

 

 

 

FY2017

8

$20,969

$4,300

$—

$4,304

$4,304

$—

$—

$—

$4,304

$16,669

 

 

 

NPV thru year 1

$218

$143

RAB share of R&D-related SBC per SEC filings

Estimates per court docs and SEC filings

$21

Increase by same percentage amount each year for six years from year 0

Instagram PCT

$60

$81

$75

0.932

 

 

NPV thru year 2

$1,557

$1,307

$202

$146

$348

$250

0.717

 

 

NPV thru year 3

$3,545

$2,942

$855

$273

$1,127

$603

0.535

Yes

 

NPV thru year 4

$6,803

$5,548

$1,477

$585

$2,063

$1,256

0.609

Yes

 

NPV thru year 5

$11,742

$8,994

$2,594

$853

$3,447

$2,748

0.797

 

 

NPV thru year 6

$18,129

$10,531

$4,535

$1,244

$5,779

$7,598

1.315

 

 

NPV thru year 7

$27,863

$13,018

$6,840

$1,847

$8,654

$14,845

1.715

 

Yes

NPV thru year 8

$41,966

$15,910

$9,735

$1,847

$11,549

$26,056

2.256

 

Yes

B. Step 1

Table A3.2 calculates the residual profit for each year, based on the difference between the foreign affiliate revenue and the non-CSA-related expenses, as marked up by the routine return. This difference is the nonroutine return. The nonroutine return is additive by year, and the sum of each year and prior years is determined and converted to an NPV as of the CSA start date. Importantly, the calculations in this appendix and Appendix 4 are based on two assumptions, as mentioned earlier: (1) that Facebook Ireland performs only routine activities, and (2) that the arm’s-length return for these routine activities is 8 percent of the costs of performing them.

Table A3.2. Step 1 Residual Profit Calculation for 2017 Determination Year (dollars in millions)

 

Year

(a)

(b)

(c)

(d)

(e)

(f)

(g) = d - e - f

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs

Divisional Profit and Loss

CC/IDCs

Routine Return %

Routine Return (MUTC) = (c) * Routine Return

Residual Profit

FY2010

1

$218

$143

$75

$21

8%

$11

$43

FY2011

2

$1,417

$1,232

$185

$191

8%

$99

$(105)

FY2012

3

$2,227

$1,832

$396

$731

8%

$147

$(481)

FY2013

4

$3,862

$3,088

$773

$738

8%

$247

$(212)

FY2014

5

$6,196

$4,324

$1,872

$1,400

8%

$346

$126

FY2015

6

$8,479

$2,041

$6,438

$2,577

8%

$163

$3,698

FY2016

7

$13,675

$3,493

$10,182

$3,239

8%

$279

$6,664

FY2017

8

$20,969

$4,300

$16,669

$4,304

8%

$344

$12,021

Cumulative PV thru year 8 as of start date

$41,966

$15,910

$26,056

$9,735

 

$1,273

$15,048

C. Step 2

Table A3.3 is used to calculate the residual return to the exploitation of the covered IP. The routine return markup is applied to the non-CC costs. That return and the CC costs are then subtracted from the reported divisional profit and loss. This table needs to be recalculated for each determination year, using the discount rate (WACC) applicable to that year.

Table A3.3. Step 2 Royalty Rate Calculation for 2017 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d)

(e)

-

(f)

(g) = d - e - f

(h) = g/d

Year No.

Foreign Sales

Non-CC/Non-IDC Costs

Divisional Profit and Loss

CC/IDCs

Routine Return Percentage

Routine Return Percentage * Non-CC Costs

Residual Profit

Royalty Calc

FY2010

1

$218

$143

$75

$21

8%

$11

$43

 

FY2011

2

$1,417

$1,232

$185

$191

8%

$99

$(105)

 

FY2012

3

$2,227

$1,832

$396

$731

8%

$147

$(481)

 

FY2013

4

$3,862

$3,088

$773

$738

8%

$247

$(212)

 

FY2014

5

$6,196

$4,324

$1,872

$1,400

8%

$346

$126

 

FY2015

6

$8,479

$2,041

$6,438

$2,577

8%

$163

$3,698

 

FY2016

7

$13,675

$3,493

$10,182

$3,239

8%

$279

$6,664

 

FY2017

8

$20,969

$4,300

$16,669

$4,304

8%

$344

$12,021

 

PV thru year 8 as of year 1

$41,966

$15,910

$26,056

$9,735

 

$1,273

$15,048

58%

D. Step 3

Table A3.4 calculates the nominal value of the royalty payment representing the residual profit as calculated in Table A3.3 on a rolling basis, from the beginning of the CSA to the determination year. Each determination year must be separately calculated using the discount rate for that year. Table A3.4 calculates the nominal royalty in the determination year, based on the sum of the NPV of the payments in each year and the years before it.

Table A3.4. Step 3 Royalty Payment Calculation for 2017 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d) = b * c

Year No.

Divisional Profit and Loss

Royalty Rate in Determination Year (2017)

Nominal Royalty Due Under Adjusted RPSM

FY2010

1

$75

58%

$43

FY2011

2

$185

58%

$107

FY2012

3

$396

58%

$229

FY2013

4

$773

58%

$446

FY2014

5

$1,872

58%

$1,081

FY2015

6

$6,438

58%

$3,718

FY2016

7

$10,182

58%

$5,880

FY2017

8

$16,669

58%

$9,627

E. Steps 4A, 4B, and 4C

Step 4A uses the table below to calculate the periodic adjustment by subtracting the nominal PCT payments from the nominal royalty payments calculated in Table A3.4. Step 4B converts these results to an NPV payment in the first year of the CSA. Step 4C takes the step 4B amount and, through compounding, converts it to a nominal payment in the determination year. All three steps are combined in Table A3.5. This result is the amount of the periodic adjustment in the determination year. This amount, which is shaded and bolded, represents an income adjustment to the U.S. tax return.

Table A3.5. Step 4 Periodic Adjustment Calculation for 2017 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d) = b * c

(e)

(f) = d - e

Year No.

Divisional Profit and Loss

Royalty Rate in Determination Year (2017 Result Shown)

Nominal Royalty Due Under Adjusted RPSM

Nominal Payments Made

Additional PCT Owed

FY2010

1

$75

58%

$43

$60

$(16)

FY2011

2

$185

58%

$107

$92

$15

FY2012

3

$396

58%

$229

$142

$87

FY2013

4

$773

58%

$446

$370

$76

FY2014

5

$1,872

58%

$1,081

$337

$745

FY2015

6

$6,438

58%

$3,718

$519

$3,199

FY2016

7

$10,182

58%

$5,880

$800

$5,080

FY2017

8

$16,669

58%

$9,627

$—

$9,627

PV of Periodic Adjustment in Determination Year 2017

 

 

$19,677

XII. Appendix 4: Periodic Adjustment Tables (2019)

The tables in this appendix follow the calculation in reg. section 1.482-7(i)(6)(vii), Example 1, applying a periodic trigger calculation for 2019. The tables were constructed to be as close to the tables shown in the regulatory example as possible. According to the regulations, all calculations for the determination year must use the taxpayer’s WACC applicable to that year. Facebook’s WACC for 2019 is reported on a financial website as 9.75 percent. All the following tables use this WACC.

A. Step 0

The AERR calculation is made using information from Facebook court filings; SEC filings; public information, with some assumptions about the schedule of buy-in payments; and other information and assumptions, excluding the IRS’s adjustment to Facebook’s buy-in payment.

Table A4.1. Calculation of Indicative Facebook Periodic Trigger Results in 2019 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

d1

d2

(d) = d1 + d2

e1

e2

(e) = e1 + e2

(f) = d + e

(g) = b - c

(h) = g/f

 

 

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs (Est.)

Stock-Based Comp

Reported CC/IDCs (Est.)

New CC/IDCs (Est.)

Buy-In PCT Paid (Est.)

Additional PCT Paid

Total PCT Paid = e1 + e2

Est. Total Investment Costs

Est. Divisional Profit (Loss)

AERR = PVTP/PVI

AERR < 0.667

AERR > 1.5

FY2010

1

$218

$143

$—

$21

$21

$60

$—

$60

$81

$75

 

 

 

FY2011

2

$1,417

$1,232

$—

$191

$191

$95

$—

$95

$287

$185

 

 

 

FY2012

3

$2,227

$1,832

$—

$731

$731

$152

$—

$152

$883

$396

 

 

 

FY2013

4

$3,862

$3,088

$—

$738

$738

$243

$152

$395

$1,134

$773

 

 

 

FY2014

5

$6,196

$4,324

$—

$1,400

$1,400

$389

$—

$389

$1,790

$1,872

 

 

 

FY2015

6

$8,479

$2,041

$—

$2,577

$2,577

$623

$—

$623

$3,199

$6,438

 

 

 

FY2016

7

$13,675

$3,493

$—

$3,239

$3,239

$995

$—

$995

$4,234

$10,182

 

 

 

FY2017

8

$20,969

$4,300

$—

$4,304

$4,304

$—

$—

$—

$4,304

$16,669

 

 

 

FY2018

9

$29,087

$8,160

$—

$5,749

$5,749

$—

$—

$—

$5,749

$20,927

 

 

 

FY2019

10

$37,142

$11,592

$9,026

$7,657

$16,683

$—

$—

$—

$16,683

$25,550

 

 

 

NPV thru year 1

$218

$143

RAB share of R&D-related SBC per SEC filings

Estimates per court docs and SEC filings

$21

Increase by same percentage amount each year for six years from year 0

Instagram PCT

$60

$81

$75

0.932

 

 

NPV thru year 2

$1,509

$1,265

$196

$146

$342

$244

0.713

 

 

NPV thru year 3

$3,358

$2,786

$802

$273

$1,075

$572

0.532

Yes

 

NPV thru year 4

$6,279

$5,122

$1,361

$572

$1,933

$1,157

0.599

Yes

 

NPV thru year 5

$10,549

$8,102

$2,326

$840

$3,166

$2,447

0.773

 

 

NPV thru year 6

$15,873

$9,383

$3,944

$1,231

$5,175

$6,490

1.254

 

 

NPV thru year 7

$23,697

$11,382

$5,797

$1,856

$7,598

$12,316

1.621

 

Yes

NPV thru year 8

$34,629

$13,623

$8,041

$1,856

$9,841

$21,005

2.134

 

Yes

NPV thru year 9

$48,445

$17,499

$10,772

$1,856

$12,572

$30,946

2.461

 

Yes

NPV thru year 10

$64,519

$22,516

$17,992

$1,856

$19,792

$42,003

2.122

 

Yes

B. Step 1

Table A4.2 calculates the residual profit for each year, based on the difference between the foreign affiliate revenue and the non-CSA-related expenses, as marked up by the routine return. This difference is the nonroutine return. The nonroutine return is additive by year, and the sum of each year and prior years is determined and converted to an NPV as of the CSA start date.

Table A4.2. Step 1 Residual Profit Calculation for 2019 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d)

(e)

(f)

(g) = d - e - f

Year No.

Est. Facebook Ireland Territory CSA-Related Sales

Non-CC/Non-IDC Costs

Divisional Profit and Loss

CC/IDCs

Routine Return Percentage

Routine Return (MUTC) = (c) * Routine Return

Residual Profit

FY2010

1

$218

$143

$75

$21

8%

$11

$43

FY2011

2

$1,417

$1,232

$185

$191

8%

$99

$(105)

FY2012

3

$2,227

$1,832

$396

$731

8%

$147

$(481)

FY2013

4

$3,862

$3,088

$773

$738

8%

$247

$(212)

FY2014

5

$6,196

$4,324

$1,872

$1,400

8%

$346

$126

FY2015

6

$8,479

$2,041

$6,438

$2,577

8%

$163

$3,698

FY2016

7

$13,675

$3,493

$10,182

$3,239

8%

$279

$6,664

FY2017

8

$20,969

$4,300

$16,669

$4,304

8%

$344

$12,021

FY2018

9

$29,087

$8,160

$20,927

$5,749

8%

$653

$14,525

FY2019

10

$37,142

$11,592

$25,550

$16,683

8%

$927

$7,940

Cumulative PV thru year 10 as of start date

$64,519

$22,516

$42,003

$17,992

 

$1,801

$22,210

C. Step 2

Table A4.3 is used to calculate the residual return to the exploitation of the covered IP. The routine return markup is applied to the non-CC costs, and this return and the CC costs are subtracted from the reported divisional profit and loss. This table needs to be recalculated for each determination year, using the discount rate (WACC) that is applicable to that year.

Table A4.3. Step 2 Royalty Rate Calculation for 2019 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d)

(e)

(f)

(g) = d - e - f

(h) = g/d

Year No.

Foreign Sales

Non-CC/Non-IDC Costs

Divisional Profit and Loss

CC/IDCs

Routine Return Percentage

Routine Return Percentage * Non-CC Costs

Residual Profit

Royalty Calc

FY2010

1

$218

$143

$75

$21

8%

$11

$43

 

FY2011

2

$1,417

$1,232

$185

$191

8%

$99

$(105)

 

FY2012

3

$2,227

$1,832

$396

$731

8%

$147

$(481)

 

FY2013

4

$3,862

$3,088

$773

$738

8%

$247

$(212)

 

FY2014

5

$6,196

$4,324

$1,872

$1,400

8%

$346

$126

 

FY2015

6

$8,479

$2,041

$6,438

$2,577

8%

$163

$3,698

 

FY2016

7

$13,675

$3,493

$10,182

$3,239

8%

$279

$6,664

 

FY2017

8

$20,969

$4,300

$16,669

$4,304

8%

$344

$12,021

 

FY2018

9

$29,087

$8,160

$20,927

$5,749

8%

$653

$14,525

 

FY2019

10

$37,142

$11,592

$25,550

$16,683

8%

$927

$7,940

 

PV thru year 10 as of year 1

$64,519

$22,516

$42,003

$17,992

 

$1,801

$22,210

53%

D. Step 3

The tables under step 3 calculate the nominal value of the royalty payment representing the residual profit as calculated in Table A4.3 on a rolling basis from the beginning of the CSA to the determination year. Each determination year must be separately calculated using the discount rate for that year. The table calculates the nominal royalty in the determination year, based on the sum of the NPV of the payments in each year and the years before it.

Table A4.4. Step 3 Royalty Payment Calculation for 2019 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d) = b * c

Year No.

Divisional Profit and Loss

Royalty Rate in Determination Year (2019)

Nominal Royalty Due Under Adjusted RPSM

FY2010

1

$75

53%

$40

FY2011

2

$185

53%

$98

FY2012

3

$396

53%

$209

FY2013

4

$773

53%

$409

FY2014

5

$1,872

53%

$990

FY2015

6

$6,438

53%

$3,404

FY2016

7

$10,182

53%

$5,384

FY2017

8

$16,669

53%

$8,814

FY2018

9

$20,927

53%

$11,066

FY2019

10

$25,550

53%

$13,510

E. Steps 4A, 4B, and 4C

Step 4A uses Table A4.5 to calculate the periodic adjustment by subtracting the nominal PCT payments from the nominal royalty payments calculated in Table A4.4. Step 4B converts these results to an NPV payment in the first year of the CSA. Step 4C takes the Step 4B amount and, through compounding, converts it to a nominal payment in the determination year. This result is the amount of the periodic adjustment in the determination year. This amount, which is shaded and bolded, represents an income adjustment to the U.S. tax return.

Table A4.5. Step 4 Periodic Adjustment Calculation for 2019 Determination Year (dollars in millions)

Year

(a)

(b)

(c)

(d) = b * c

(e)

(f) = d - e

Year No.

Divisional Profit and Loss

Royalty Rate in Determination Year (2019 Result Shown)

Nominal Royalty Due Under Adjusted RPSM

Nominal Payments Made

Additional PCT Owed

FY2010

1

$75

53%

$40

$60

$(20)

FY2011

2

$185

53%

$98

$95

$3

FY2012

3

$396

53%

$209

$152

$57

FY2013

4

$773

53%

$409

$395

$13

FY2014

5

$1,872

53%

$990

$389

$601

FY2015

6

$6,438

53%

$3,404

$623

$2,782

FY2016

7

$10,182

53%

$5,384

$995

$4,389

FY2017

8

$16,669

53%

$8,814

$—

$8,814

FY2018

9

$20,927

53%

$11,066

$—

$11,066

FY2019

10

$25,550

53%

$13,510

$—

$13,510

PV of Periodic Adjustment in Determination Year 2019

 

 

$47,159

FOOTNOTES

1 Reg. section 1.482-7(i)(6)(vi)(B)(1).

2 The IRS’s intention to increase its adjustment was disclosed in Respondent’s Statement of Position, Facebook v. Commissioner, No. 21959-16 (T.C. Oct. 18, 2019). Other filings in the Tax Court case are publicly available. See, e.g., Facebook’s 2016 petition, infra note 2; and the Stone declaration, infra note 6. There is related litigation in federal district court. See Facebook’s Opposition to Amended Petition to Enforce Internal Revenue Service Summonses, United States v. Facebook Inc., No. 3:16-cv-03777 (N.D. Cal. Oct. 11, 2016). I have also obtained Facebook’s petition in a separate Tax Court case contesting the IRS’s adjustments for 2011 and 2013: Facebook Inc. v. Commissioner, No. 012738-18 (T.C. June 29, 2018).

3 Petition, Facebook Inc. v. Commissioner, No. 21959-16 (T.C. Oct. 11, 2016). Facebook’s CSA was implemented when temporary cost-sharing regulations under reg. section 1.482-7T (T.D. 9569) were in effect. Those regulations became final December 16, 2011 (T.D. 9630). This report generally refers to the final regulations, except when its provisions differ from those of the temporary regulations. In those instances, the temporary regulations are referenced.

4 The first trial session began in February, and a second session has been scheduled for 2021, according to Facebook’s second-quarter 2020 Form 10-Q.

5 The first sentence of reg. section 1.482-7(i)(6) states that subject to the exceptions in reg. section 1.482-7(i)(6)(vi), “the Commissioner may make periodic adjustments for an open taxable year.” (Emphasis added.) The regulation does not require the taxpayer to make this adjustment; rather, it is a tool for the IRS to apply.

6 While reg. section 1.482-7(i)(6) applies the CWI concept to CSA situations, reg. section 1.482-4(f)(2) applies that concept to other related-party transactions involving transferred intangibles.

7 Declaration of Nina Wu Stone in Support of Petition to Enforce Internal Revenue Service Summonses at para. 21, United States v. Facebook Inc., No. 3:16-cv-03777 (N.D. Cal. July 6, 2016).

8 Amazon.com Inc. v. Commissioner, 148 T.C. 108 (2017). I analyzed the case using forensic economic modeling techniques, which demonstrated that the IRS’s adjustment was unsound and likely violated the CWI and arm’s-length standards. For comparison, I analyzed a CSA that was implemented by an Amazon.com competitor and approved by the IRS. According to the same forensic models, the approved CSA was aggressively noncompliant. I found that the IRS may have a bias toward challenging economically compliant arrangements, but that it often fails to detect or challenge what appear to be — based on forensic economic modeling — noncompliant arrangements. See Stephen L. Curtis, “Forensic Approaches to Transfer Pricing Compliance and Enforcement,” 8 J. Forensic & Investigative Acct. 359 (2016).

9 Interestingly, the taxpayer’s apparent recalcitrance was a partial cause for this. The revenue agent comments: “Due to budgetary constraints, the IRS team could not commence the lengthy expert retention process until after October 1, 2015. Because of this, the IRS team asked Facebook to extend the statute of limitations, to afford the team time to confer with experts and follow up with any additional requests for information to Facebook. Facebook, however, refused to afford the IRS team any further extensions of time, unless the IRS agreed to unacceptable conditions.” Stone declaration, supra note 6, at para. 27. In July 2016, just before the expiration of the statute of limitations and when the IRS still had not received requested information, the agency issued Facebook a statutory notice of deficiency.

10 It is unknown what, if any, recent-year Facebook information the IRS has already requested. In any event, the IRS should obtain as soon as possible, and consider, taxpayer information through 2019 in order to calculate the periodic adjustment(s) that this report believes would result under reg. section 1.482-7(i)(6).

11 Veritas Software Corp. v. Commissioner, 133 T.C. 297 (2009).

12 JCT, “General Explanation of the Tax Reform Act of 1986,” JCS-10-87, at 1016 (1987).

13 Altera Corp. v. Commissioner, 926 F.3d 1061 (9th Cir. 2019). The Supreme Court denied Altera Corp.’s petition for certiorari June 22, 2020, ending Altera’s challenge to the requirement to share stock-based compensation expenses in CSAs and affirming the Ninth Circuit’s decision.

14 Id. at 1083.

16 The calculation of the AERR is purely mechanical and immutable, based on recorded financial results. The calculation of the periodic adjustment can also be relatively mechanical when the controlled foreign affiliate conducts and owns no nonroutine (non-benchmarkable) functions or assets. The periodic adjustment calculation relies on a residual profit-split approach, which would normally be prohibited under reg. section 1.482-7(g) when the controlled foreign affiliate has no nonroutine contributions. However, reg. section 1.482-7(i)(6)(v)(B)(3) waives this restriction. Therefore, in low-substance CSAs exhibiting only routine functions by “tiny” numbers of people (as described by Facebook’s chief operating officer in an email dated October 1, 2008, submitted to the Tax Court by the IRS), the calculation becomes substantially mechanical.

17 The reg. section 1.482-7(i)(6) mechanism is described in more detail in Appendix 2.

18 Since the 2009 CSA regulations were finalized, the IRS issued Rev. Proc. 2015-41, 2015-35 IRB 263, which includes the following language implying that the IRS will agree in an advance pricing agreement to waive compliance with the periodic adjustment rule of reg. section 1.482-7(i)(6) and thus with the CWI standard: “If a covered issue is a platform contribution transaction, the APA may provide that such transaction will not be treated as a Trigger PCT within the meaning of [reg. section] 1.482-7(i)(6)(i) for purposes of making periodic adjustments, during or after the APA term, under [reg. section] 1.482-7(i)(6).” It seems unbelievable that the IRS might agree to terms that emasculate the congressionally mandated CWI standard.

19 AOD 2010-05 (footnote omitted).

20 Michael F. Patton, “Will the New IRS Transfer Pricing Director Bring Sanity to the IRS’s Transfer Pricing Enforcement and Litigation Efforts?” 40 Tax Mgmt. Int’l J. 537 (2011).

21 Brauner, “Cost Sharing and the Acrobatics of Arm’s Length Taxation,” 38 Intertax 554 (2010).

23 Curtis, supra note 7; and Curtis and Yaron Lahav, “Forensic Approaches to Transfer Pricing Enforcement Could Restore Billions in Lost U.S. Federal and State Tax Losses: A Case Study Approach,” 12 J. Forensic & Investigative Acct. 285 (2020).

24 See Altera, 926 F.3d 1061.

25 Facebook’s CSA has now surpassed this 10-year period. It is unclear whether Facebook or the IRS has monitored these results over that period to assess whether Facebook has met these safe harbor requirements. There is no mention of a periodic adjustment calculation in the cited court filings, or whether Facebook has complied with the purported documentation requirements for demonstrating compliance with the periodic adjustment rules.

26 See Curtis and Lahav, supra note 23, at Table 7.

27 TechCrunch, “Facebook Valuation Over Time” (2011).

28 Facebook petition, supra note 2, at para. 5.b.6.

29 Altera, 926 F.3d 1061.

30 See appendices 2 through 4 for a complete description of the calculations and inputs.

31 In 2019 Facebook paid a fine of $5 billion to the Federal Trade Commission, which, per SEC filings, reduced solely U.S. profits. This fine related to a violation of user privacy laws perpetrated by a company located in Facebook Ireland's CSA territory, and was non-deductible. Facebook reflected this as a decrease to U.S. pretax income, but with federal taxes computed on U.S. pretax income before paying the fine.

32 Facebook Inc. Form 10-Q for the quarter ended June 30, 2020, at 41.

33 Facebook had a $1.5 billion term loan outstanding between October 25, 2012, and August 2013 to fund a portion of the withholding tax liability resulting from the vesting and settlement of restricted stock units.

34 Curtis and Lahav, supra note 23.

35 Estimates of the proportion of long-term assets are based on SEC filings. Estimates of personnel ratios and operational data center capacity are based on Curtis and Lahav, supra note 23.

36 As written in the 10-year safe harbor rule, a downward periodic trigger in the first 10 years would also cause the taxpayer to run afoul of the 10-year safe harbor threshold. This seems to be an unintended result because a periodic adjustment should be required only when the PCT payer has experienced excessive earnings. The regulations do not otherwise distinguish between upward and downward periodic triggers, except for the five-year-period safe harbor rule in which a downward periodic result triggers the taxpayer-beneficial result.

37 These assumptions are supported by the 2011 audit of Facebook Ireland by the Irish Data Protection Commission, which reported that most of the employees represented entry-level positions providing routine administrative activities, such as site monitoring, recordkeeping, and customer service. Further, as mentioned, Facebook’s $5 billion fine paid to the FTC in 2019 reduced solely U.S. profits (supra note 31), and Facebook Ireland does not appear based on this to truly bear financial risks attributable to its activities, territory, and customers, which would further support only routine returns.

38 This is investigated extensively in Curtis and Lahav, supra note 23.

39 See Miguel Helft, “Facebook Makes Headway Around the World,” The New York Times, July 7, 2010; and David Kirkpatrick, The Facebook Effect: The Inside Story of the Company That Is Connecting the World (2010); and Cade Metz, “Meet the Data Brains Behind the Rise of Facebook,” Wired, Feb. 4, 2013.

40 Jemima Kiss, “Facebook Ireland Chief: Tax Breaks, 100 New Staff and a ‘Reputation for Driving Revenue,’” The Guardian, Dec. 7, 2010.

41 This was disclosed in the Irish Data Protection Commission audit report in 2011.

42 See Pamela Newenham, ed., Silicon Docks: The Rise of Dublin as a Global Tech Hub (2014).

43 Senate Permanent Subcommittee on Investigations, “Caterpillar’s Offshore Tax Strategy: Majority Staff Report,” at 6 (Aug. 28, 2014).

44 In contrast to this Senate request, the IRS has issued directives designed to prevent examiners from asserting the economic substance doctrine in examinations of corporations. See Lee A. Sheppard, “How Did We Lose the Economic Substance Doctrine?” Tax Notes Federal, Nov. 9, 2020, p. 893.

45 Pre-2009 reg. section 1.482-7A(a)(1) stated that the costs of IP development would be shared according to “their individual exploitation of the interests in the intangibles assigned to them under the arrangement.” (Emphasis added.) This sentence does not appear in the 2009 regulations, although the regulations refer multiple times to the “exploitation rights” in various examples, which show the foreign participants exploiting the IP themselves, and none of them resemble low- or no-substance CSAs in which the Irish affiliate performs or controls few or none of the IP exploitation functions that generated the revenue and profits being recorded there. The regulations nevertheless require economic substance.

46 This conclusion is supported by the October 2008 internal Facebook email, supra note 15, in which the COO refers to the Google CSA she helped implement as a basis for the Facebook arrangement, describing the transactions as tax sheltering arrangements that involving only “very small” and “tiny” numbers of employees and providing “tax breaks . . . to run international revenues through." This is supported by additional documents, including a 2019 decision by the Administrative Court of Appeals in France, in which the company maintained that Google’s search and advertising services in France were provided directly by Google US, and that Google Ireland Limited was not involved in these services, and had no control over them (Judgement No. 17PA03065, The Administrative Court of Appeal of Paris (9th Chamber), April 25, 2019). The email also notes that the arrangement was identical to the Google arrangement that the same executive helped arrange five years earlier. Likewise, Newenham (supra note 42) documents that Industrial Development Agency Ireland put tax avoidance “front and center” of its extensive lobbying efforts to attract Google to Dublin. See also United States v. Microsoft Corp., No. 2:15-cv-00102 (W.D. Wash. 2020); and associated analysis in Curtis and Lahav, supra note 23.

47 Tyler M. Johnson, John Hildy, and John W. Horne, “Cost Sharing Is a Tax Shelter Now. Wait, What?” Tax Notes Federal, Sept. 18, 2020, p. 2193.

48 Reg. section 1.482-7(b)(4)(ii) appears to require that the physical exploitation activities occur in the CSA participant’s own territory. It states: “Each controlled participant will be assigned the perpetual and exclusive right to exploit the cost shared intangibles . . . in its territories. Thus, compensation will be required if other members of the controlled group exploit the cost shared intangibles in such territory.” (Emphasis added.) Facebook’s generation of foreign advertising revenue from U.S.-based data centers (exclusively before 2013, and substantially thereafter) would appear to be inconsistent with this requirement.

49 Curtis, supra note 7.

50 Supra note 45.

51 See reg. section 1.482-7(a)(4)(ii).

52 See OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations, para. 8.14ff.

54 Senate Permanent Subcommittee on Investigations, supra note 43.

55 For instance, in June 2013 the British House of Commons held a hearing with representatives for both Google and Ernst & Young side by side soliciting information on EY’s involvement in developing Google’s CSA.

56 This characterization reflects that when the CSA was executed, the Irish affiliate had only around five employees, who were hired approximately three months earlier and managed by a U.S. executive. Newenham, supra note 42. The affiliate also lacked any material operations with which to physically earn the nonroutine, residual IP exploitation profits that were attributed to it after execution of the CSA. For transfer pricing purposes, such exploitation generally requires sufficient brains and bodies to physically make business decisions, manage commercial and other risks, and perform nonroutine revenue-generating functions. Indeed, the absence of any nonroutine contributions by Irish cost-sharing participants is often cited as a reason for rejecting the residual profit-split method of reg. section 1.482-7(g)(ii) to test the appropriateness of Irish profits under the arm’s-length standard. That regulation states: “The relative value of each controlled participant’s contribution . . . must be determined in a manner that reflects the functions performed, risks assumed, and resources employed.” (Emphasis added.) Because there are no functions or physical resources used, financial risk-bearing is often cited as the only contribution supporting the residual profits earned in Ireland (as Facebook claimed in its 2016 petition regarding the 2009 IP license agreement). But even here this depends on the IRS ignoring the requirements of reg. section 1.482-1(d)(3)(iii)(B)(3), under which risk-bearing is respected only to the extent that the entity also controls the underlying risks. Because these Irish affiliates have no control over the U.S. parent’s exploitation activities putatively attributed to them, this risk-bearing cannot be the basis for any residual profits. Combined with the presence of only tax benefits, these no-substance CSAs would appear to be textbook violations of the economic substance doctrine, even if the U.S. transfer pricing regulations do not explicitly prohibit entering into tax shelters effectuated by CSAs with quasi-shell companies.

57 See Jeffery M. Kadet, “Attacking Profit Shifting: The Approach Everyone Forgets,” Tax Notes, July 13, 2015, p. 193; Kadet and David L. Koontz, “Profit-Shifting Structures and Unexpected Partnership Status,” Tax Notes, Apr. 18, 2016, p. 335; and Kadet and Koontz, “Effects of the New Sourcing Rule: ECI and Profit Shifting,” Tax Notes, May 21, 2018, p. 1119 (see, in particular, Section II).

58 Section 6501(c)(3) keeps the year open if no return has been filed. Even if a protective return has been filed that claims no U.S. trade or business or ECI, the blatant profit-shifting structure suggests that section 6501(c)(1) or (2) would apply to prevent any early year from closing.

59 See Curtis and Lahav, supra note 23.

60 See “Inside the Googleplex,” The Economist, Aug. 30, 2007.

61 See Kadet, supra note 57; Kadet and Koontz, “Unexpected Partnership Status,” supra note 57; Kadet and Koontz, “Effects of the New Sourcing Rule,” supra note 57; H. David Rosenblum, “Kumquat: The U.S. International Tax Issues,” Tax Notes Int’l, June 25, 2018, p. 1521 (Kumquat referring to Apple); Sheppard, “What About Cupertino?” Tax Notes Federal, July 27, 2020, p. 565; Curtis, supra note 7; and Curtis and Lahav, supra note 23.

62 Facebook aggregates DAUs for the United States and Canada but reports third-party revenue by U.S. and foreign jurisdictions. These figures are for the Facebook app and do not include users on Instagram, WhatsApp, or other products. WhatsApp had an estimated 1.5 billion monthly active users in 2019, but only 4.5 percent were in the United States, while Instagram had an estimated 1 billion monthly active users as of 2018, of whom 12 percent were located in the United States. See J. Clement, “Instagram — Statistics and Facts,” Statista, May 14, 2020; and “Number of Daily WhatsApp Status Users From 1st Quarter 2017 to 1st Quarter 2019,” Statista (2020).

63 Facebook SEC filings report both DAUs and monthly average users (MAUs), and that as many as 11 percent of MAUs may be duplicate accounts, and that most of those duplicate accounts are located in developing countries. Facebook does not provide estimates of duplicate accounts associated with DAUs. Further, these estimates are based on algorithms and assumptions about user location and are subject to error (for instance, usage through a proxy server in a different country from the user could result in the user being assigned to a different country or region). I believe that the use of DAUs versus MAUs, and the use of U.S. versus consolidated foreign groupings, may reduce the effect of these measurement issues. However, this cannot be guaranteed.

END FOOTNOTES

Copy RID