Has Cost Sharing Outlived Its Usefulness?

Reuven S. Avi-Yonah

Reuven S. Avi-Yonah is the Irwin I. Cohn Professor of Law at the University of Michigan Law School. He thanks Elise Bean, Kim Clausing, Steve Curtis, Ryan Finley, Robert Goulder, Mindy Herzfeld, Jeff Kadet, and Marty Sullivan for helpful comments.

In this installment of Reflections With Reuven Avi-Yonah, Avi-Yonah explains why he thinks cost-sharing arrangements could be eliminated.

In her thoughtful column on cost sharing, Mindy Herzfeld explains the history of cost-sharing arrangements (CSAs) going all the way back to 1966, and then asks several questions, including: “Should the existing CSA regs be scrapped in favor of another regime, or simply eliminated as a choice for taxpayers?”¹

In my opinion, cost-sharing regulations should be eliminated. They have become a way for large multinational enterprises to avoid the requirement under section 482 that transfers of intangibles to foreign affiliates produce a super royalty that is commensurate with the income attributable to the intangibles. Further, the rationale for adopting them has proven to be invalid.²

To understand the problem with CSAs, a good place to start is Amazon v. Commissioner.³

The case involved a CSA entered into in 2005 between Amazon and its Luxembourg subsidiary (Luxco). Under the CSA, Amazon granted Luxco the right to use certain preexisting intangible assets in Europe, including the intangibles required to operate Amazon’s European website business. The CSA required Luxco to make an upfront “buy-in payment” to compensate Amazon for the value of the intangible assets it developed that were to be transferred to Luxco.⁴ Thereafter Luxco was required to make annual cost-sharing payments to compensate Amazon for ongoing intangible development costs (IDCs), to the extent those IDCs benefited Luxco.⁵ As consideration for the transfer of preexisting intangibles, Luxco made a $254.5 million buy-in payment to Amazon.

Applying a discounted cash flow method to the expected cash flows from the European business, the IRS determined a buy-in payment of $3.6 billion, later reduced to $3.468 billion. In Tax Court, Amazon contended that the IRS’s discounted cash flow method is substantially similar to that rejected by the Tax Court in Veritas Software Corp.⁶ Amazon further contended that the IRS’s determinations were arbitrary, capricious, and unreasonable and that the comparable uncontrolled transaction method was the best method to calculate the requisite buy-in payment.

Amazon used a multistep allocation system to allocate costs from its cost centers to IDCs.⁷ While accepting Amazon’s allocation method in many respects, the IRS determined that 100 percent of the costs captured in one important cost center (Technology and Content) must be allocated to IDCs. Amazon contended that the IRS’s determination to allocate to IDCs 100 percent of the Technology and Content costs was inconsistent with the regulations.

The Tax Court, in a 207-page opinion, held that the IRS’s determination of the buy-in payment was arbitrary, capricious, and unreasonable. It further held that Amazon’s CUT method, with appropriate upward adjustments in numerous respects, was the best method to determine the requisite buy-in payment. The court also held that the IRS abused its discretion in determining that 100 percent of Technology and Content costs constitute IDCs, and that Amazon’s cost-allocation method, with certain adjustments, supplied a reasonable basis for allocating costs to IDCs.

The IRS has been more successful in transfer pricing litigation since Amazon was decided in 2017, but none of its victories have involved a challenge to the transfer of intangibles under a CSA.⁸ In fact, CSAs were a major contributing factor to the ability of U.S.-based MNEs to amass over $2.5 trillion of profits in offshore low-tax jurisdictions before 2017. The tax planning behind this achievement has been the focus of numerous Congressional investigations, especially by the Senate Permanent Subcommittee on Investigations under the leadership of former Sen. Carl Levin.⁹

The subcommittee’s investigations revealed a consistent pattern of tax planning among U.S. MNEs. The companies develop intangibles in the United States but shift profits from the intangibles to low-tax jurisdictions offshore using CSAs. The data assembled by the subcommittee show that there was no correlation between the location of these profits and the actual activities undertaken by the MNEs in locations like Ireland, Singapore, or Puerto Rico.

For example, consider the following subcommittee data for Microsoft. In 2011 Microsoft earned $8.93 billion in Ireland and paid an effective Irish tax rate of 5.76 percent. It had 1,050 employees in Ireland, translating to over $8 million earned by each employee. In Singapore, Microsoft earned $2.48 billion at an effective tax rate of 2.74 percent, with its 687 employees each earning over $3 million. Most impressively, in Puerto Rico, Microsoft earned $4.015 billion (mostly for sales to the United States), paid taxes at an effective rate of 1.02 percent, and its 177 employees each supposedly generated over $22 million in income.¹⁰ Kimberly Clausing’s research has shown that this pattern of shifting profits offshore has continued after 2017 because of the disparity between the domestic tax rate and the global intangible low-taxed income rate and the lack of a per-country limit in GILTI.¹¹

In 1986 after a series of high-profile cases involving transfers of intangibles developed in the United States to low-tax jurisdictions, Congress amended code section 482 for the first time since 1924 by adding the “super royalty rule.”¹² This rule was supposed to require the transferee to pay royalties to the U.S. transferor that would increase over time and negate any profit-shifting potential from the transfer. But the subcommittee found that U.S. MNEs have still been able to shift trillions of dollars of profits, attributable largely to U.S. developed intangibles, to low-tax jurisdictions. The explanation is mostly the cost-sharing rule.¹³

Cost sharing is a regime introduced by Treasury and the IRS in 1966, which became even more important from the early 1990s following the adoption of the super-royalty rule. Under cost sharing, the U.S. parent enters into a CSA with a controlled foreign corporation to share the costs of developing an intangible. Importantly, nothing actually happens in the CFC: The entire development takes place in the United States. The CFC contributes a portion of the costs (for example, 80 percent) by simply receiving a contribution from its parent and paying it back. If the development is successful, the CFC is then entitled to 80 percent of the profits from the intangible, without any concern about potential application of IRC section 482 and its regulations (including the super-royalty rule).

The idea behind cost sharing was that the MNE cannot know whether the development will be successful, and risks losing 80 percent of the research and experimentation deduction if it is unsuccessful because the deduction is shifted to the CFC, which does not have U.S.-source income. This, it was thought, would inhibit MNEs from taking too aggressive a position in their CSAs.

But this idea is deeply flawed for two reasons. First, successful intangibles result in profits that far outweigh the costs of development. Thus, it makes sense for a multinational that develops an intangible for $100 million with a profit potential of $1 billion to risk losing $80 million in deductions, if there is a good chance that this will shield $800 million or more in profits from current U.S. taxation. The data assembled in the Microsoft investigation, like the $15 billion earned by Microsoft in 2011 in Ireland, Singapore, and Puerto Rico, show the immense sums that can be earned from the successful exploitation of intangibles.

Second, for cost sharing to work, the MNE cannot know whether the intangible will be successful or not. But MNEs are in the best position to know that, and it is very hard for the IRS to second-guess MNE knowledge or lack thereof. MNEs typically enter into CSAs only when they know the intangible will be profitable, and while in theory this requires a buy-in payment by the CFC that leads to results similar to the super-royalty rule, in practice this becomes a valuation issue. As Amazon shows, the IRS has not been successful in litigation over buy-in payments.

Thus, in my opinion, cost sharing has been an expensive mistake. It enables MNEs to shift profits from intangibles developed in the United States offshore without incurring any serious risk of losing the R&E deductions.

While Amazon was litigated under the old cost-sharing regulations, there is no reason to assume a different result would occur under the new regulations. Amazon is a very profitable company that is taking no risk in developing a new version of its intangible property. Under these circumstances, cost sharing should not apply. Rather than try to litigate this issue again and again, the IRS should simply revoke cost sharing.

Herzfeld writes that:

Proponents of CSAs argue that they reflect the reality of most large multinationals, which is that numerous jurisdictions contribute to the development of intangibles that are used by many facets of a business, across multiple locations. Believers in the need for cost-sharing regs argue that CSAs are little more than a way of providing simplified accounting of costs across group members and subsequent division of ownership consistent with the arm’s-length principle, avoiding the need for complex licensing agreements and cross-border royalty payments that would be subject to withholding taxes.¹⁴

But this defense of cost sharing ignores the fact that generally the CFC contributes nothing to the development of the intangible except the cash that it received from its parent, and therefore the CFC’s jurisdiction does not in fact “contribute to the development of intangibles.” Nor is the solution to require that the CFC conduct actual research, because that will encourage U.S. MNEs to move R&E outside the United States, and R&E comes with positive externalities, so it is better to keep it here.¹⁵

Stanley Surrey had many positive contributions to U.S. tax law, including the concept of a base-broadening, rate-cutting tax reform that underlay the Tax Reform Act of 1986, the tax expenditure budget, and the idea that international tax rules should prevent both double taxation and double nontaxation, whose effect can still be seen in the Tax Cuts and Jobs Act and pillar 2.¹⁶ But he also made his share of mistakes, including the adoption of capital export neutrality as the rationale for subpart F in 1961, the rejection of formulary apportionment, and the retention of the arm’s-length standard after the House bill adopted formulary apportionment in 1962, and the adoption of CSAs in 1966.¹⁷ Since this last error is only embodied in a regulation and is not included in either the code or U.S. tax treaties, it can be reversed by eliminating CSAs.

In a recent column, Ryan Finley takes the opposite point of view.¹⁸ He makes three points, but I do not find them persuasive.

First, Finley writes that:

Before even attempting to assess whether Treasury and the IRS should try to put an end to cost sharing, it’s necessary to examine whether they could if they wanted to. Does section 482 provide the statutory authority necessary to support an effective prohibition on cost sharing? If the legislative histories of the Tax Reform Act of 1986 and the Tax Cuts and Jobs Act are any indication, probably not.

It seems to me that this is the wrong question. The right question is whether section 482 requires Treasury and the IRS not to put an end to cost sharing. I do not see how anything in the statutory text before 1986 or as amended in 1986 and in 2017 requires cost sharing to be retained. The fact that the legislative history accepts cost sharing does not mean that there is an obligation on Treasury and the IRS to keep it in place. Nor does the fact that cost sharing has been in the regulations in some form since 1966 give taxpayers the right to assume it will continue forever. As long as the withdrawal of cost sharing is not retroactive (does not apply to CSAs entered into before the withdrawal), there is no valid reliance argument against withdrawing it.¹⁹

A good analogy for this situation is the check-the-box regulation, adopted in 1997.²⁰ In 1998 Treasury realized that extending check-the-box rules to foreign corporations was a mistake, and it tried to fix the error in Notice 98-11, 1998 C.B. 433. The MNEs complained to Congress, and Treasury was forced to withdraw the notice, but Treasury adopted it as proposed regulations with an effective date five years after finalization. I was told that had Al Gore won the 2000 election, the regulations would have been finalized. But nobody suggested then or thereafter that Treasury and the IRS could not withdraw check-the-box regulation, which is why the Republican majority in Congress codified it in October 2006 as section 954(c)(6) when it was widely and correctly expected that the Democrats would soon capture Congress and the White House and would be able to repeal the regulation.²¹

Second, Finley argues that:

Withdrawing the cost-sharing regulations and eliminating any tacit regulatory acceptance of CSAs wouldn’t actually end cost sharing. MNEs generally enjoy the freedom to arrange their operations and structure their intercompany transactions however they wish. That means they don’t really need reg. section 1.482-7’s permission to co-develop intangibles with foreign affiliates or arrange for foreign affiliates to cofund intangible development activities carried out in the United States.

CSAs are forbidden now? Fine, then these codevelopment or cofunding arrangements aren’t CSAs anymore. Now they’re simultaneous assignments of territorial intangible development and exploitation rights subject to reg. section 1.482-4, and (potentially bilateral) contract research and development service agreements subject to reg. section 1.482-9.

Sure, the administrative particulars and the exact calculation of the parties’ payment obligations would differ. But the essential features of the arrangement could remain the same. Whether it’s called a platform contribution or just an assignment of rights, the original owner could still make its valuable intangible property available for use in further joint development. The same parties could also arrange for the joint development itself, or at least the joint funding of that development, in a manner that roughly aligns with the treatment of cost contributions under reg. section 1.482-7.

The key tax compliance issue would still be the pricing of the initial intangible transfer or contribution, and nominal extinction of CSAs wouldn’t change that. Withdrawing reg. section 1.482-7 would simply mean that the only remaining regulatory instruments to address the issue would be provisions of reg. section 1.482-4, which signally failed the IRS in Veritas and Amazon.

I am not persuaded. If CSAs did not exist, then courts would be free to apply the commensurate with income rule in the statute to any transfer of an intangible out of the United States, and that would significantly improve the tax outcome.²² I do not see why courts would accept the allocation of any profit to a subsidiary whose only contribution to the development of an intangible is money that was contributed to it by its U.S. parent, which is the current result under the “investor model” of the CSA regulations. No income should be assigned to a foreign subsidiary that does nothing.

I am reminded of the United States Gypsum case, in which the gypsum was owned by the Canadian subsidiary while on the dock in Canada and by the U.S. parent while on the ship sailing to the United States, but while it was falling from the dock to the ship, it was owned by a low-taxed foreign subsidiary. As the court said:

As illustrated in the above chart, Export owned the gypsum rock only for a brief moment while it fell from Canadian’s dock into the hold of the waiting vessel — or possibly until the vessel was fully loaded. All of Export’s dealings relating to the purchase and resale of crude gypsum rock were within the corporate family. The extent of its participation in the entire transaction is set forth in the post trial brief of the government:

“Export did not mine the gypsum (or own the mines), deliver the gypsum from the quarry to the dock, or load the gypsum from the dock into the vessel: CG (Canadian) did all of that. Export did not own or charter the vessels which carried the gypsum rock to the United States, order delivery of the gypsum by CG (Canadian) at dockside in Canada, or notify CG (Canadian) of the vessel’s arrival: Panama Gypsum Company and Gypsum Transportation Company did all of that. Export did not process the gypsum rock, manufacture it into finished products, or sell those finished products: USG did all of that. And Galileo taught us from his tower in Pisa that gravity — not Export — caused the gypsum rock to fall from the conveyor on CG’s (Canadian’s) dock to the stowage of the Gypsum Prince” . . .

In its purchase and resale of crude gypsum, what risk did Export bear? What service did it perform other than to shift profits and thereby lower taxes? My answer to both questions is none.²³ [Emphasis in original.]

Similarly, a subsidiary that only contributes cash that was given to it by its parent to the development of an intangible has not performed any function that entitles it to a share of the profits.

Finally, Finley writes that:

Even if Treasury and the IRS have the legal authority to withdraw reg. section 1.482-7 in the effort to end cost sharing, it makes little sense for them to use it. The likely harm that withdrawing the cost-sharing regulations would cause would almost certainly outweigh any abstract benefit associated with abolishing the implied acceptance of CSAs. Whether the cost-sharing regulations strengthen or impede the IRS’s transfer pricing enforcement abilities isn’t a question that can be settled at the level of generalities.

I agree that just withdrawing the CSA regulation by itself would not help, but if the IRS also started to enforce its statutory commensurate with income authority, a better result could be obtained.²⁴ If that does not help, the United States should consider going along with the OECD and adopting formulary apportionment under pillar 1, because for the largest MNEs, it ensures that some of the profits will be taxed by the United States.