Should Digital Services Taxes Be Creditable?

Reuven S. Avi-Yonah

Reuven S. Avi-Yonah (aviyonah@umich.edu) is the Irwin I. Cohn Professor of Law at the University of Michigan Law School. He thanks Peter Barnes, Kimberly Clausing, Wei Cui, Ed Fox, Ruth Mason, Paul Oosterhuis, and Fadi Shaheen for their helpful comments.

In this installment of Reflections With Reuven Avi-Yonah, Avi-Yonah argues that because digital services taxes are used to offset the taxation impediment that “digital giants” cannot be taxed under permanent establishment rules, they should qualify as an in-lieu-of tax and therefore be creditable.

The impending collapse of pillar 1 of the OECD’s base erosion and profit-shifting 2.0 project means that many countries are likely to join Austria, France, Italy, Spain, and the United Kingdom in enacting digital services taxes. There is a moratorium on new DSTs that has been extended until the end of 2024, but any extension beyond that is unlikely because the elimination of DSTs was premised on pillar 1 coming into effect, and that cannot happen without U.S. ratification of the multilateral tax convention (MLC). It is safe to predict that there will not be 67 votes in the Senate to ratify the MLC regardless of the results of the 2024 election.¹

DSTs are popular for three reasons:

• Fairness — they are perceived as enabling market jurisdictions to tax U.S. digital giants earning billions in profits in these jurisdictions while avoiding local taxes because of limits imposed by tax treaties.

• Efficiency — digital giants are close to being monopolies, and imposing a tax on a monopoly does not significantly change its behavior because in most cases it was already able to charge the maximum price the market would bear before the imposition of the tax.

• Revenue source — they raise significant revenue from nonvoters and therefore satisfy the saying “don’t tax you, don’t tax me, tax the fellow beyond the sea.”²

The United States has fought long and hard against DSTs. It reacted to the first DST, imposed by France, by enacting trade sanctions; and it followed that by threatening trade sanctions against other DSTs, including the one Canada is planning to impose in 2024 despite the moratorium.³ But the United States suspended the sanctions pending the pillar 1 negotiations and allowed the countries that imposed them before 2020 to continue collecting revenue, as long as they promised to credit the DSTs paid against future tax liabilities under pillar 1 — which because of U.S. opposition will not happen.⁴

If additional countries impose DSTs next year, how should the United States respond? The correct response would be to allow a foreign tax credit for DSTs, which would require withdrawing the regulation disallowing these credits.⁵

Under the FTC regulations finalized in 2022, no credit is allowed for foreign taxes that do not satisfy the attribution requirement. Specifically, a tax on services is not creditable if it does not source services “based on where the services are performed, as determined under reasonable principles (which do not include determining the place of performance of the services based on the location of the service recipient).”⁶

A DST is not an income tax because it is imposed on a gross basis and does not allow deductions. That is true of withholding taxes as well, which are creditable as “in lieu of” taxes under section 903. But under the attribution rule, a DST cannot be an in-lieu-of tax because even if it were imposed on a net basis, it is imposed based on the location of the service recipient, which disqualifies it as a creditable tax.

The attribution rule has for now been suspended indefinitely by Notice 2023-80, 2023-52 IRB 1583. But DSTs are still not creditable under section 901 because they are not net-based income taxes or under section 903 if they are also imposed on residents (like the French and U.K. DSTs and the proposed Canadian DST, but unlike the Indian DST).⁷ Does this result make sense?

To answer this question, it is necessary to address what Paul Oosterhuis has recently called the age-old issue of what taxes should be treated as income taxes and therefore creditable under section 901.⁸

As Oosterhuis explains, the traditional distinction for purposes of creditability was between net- and gross-based taxes. He writes that:

I start by setting forth what I believe to be the first of two fundamental principles that should guide our thinking: Net income taxes are economically different than, and should be distinguished from, taxes on revenues. That may be obvious but is worth our focus. While I am no economist, I believe the principle is based on a pretty solid economic foundation: The incidence of net income taxes is different in substantial ways compared to revenue taxes. Economists who contemplated tax incidence in the post-World War I formative days of the income tax generally viewed net income taxes as borne by the earners of income while revenue taxes were borne by purchasers. In today’s world, the economic analysis of tax incidence has become more subtle. But I believe it is still fair to say that most economists see a difference: They view true net income taxes as borne by a mix of investors and labor while revenue taxes are borne in large part by purchasers, assuming at least in the case of a major tax increase or a new revenue tax, the central bank and local law allow for price increases corresponding to the tax increases. A tax credit for foreign net income taxes thus maintains neutrality between purely domestic and cross-border economic activity, which I believe is a fundamental policy goal. A credit for revenue taxes against net income taxes is not consistent with that goal, and indeed can be a windfall to taxpayers who can in fact pass on much of the cost to customers.

Thus, Oosterhuis would not credit foreign taxes on gross revenues like the DST, and he also argues that the treatment of withholding taxes as taxes “in lieu of” a net income tax is not tenable when the rates of the withholding tax are not comparable with the net rate.

Does this distinction between net and gross taxes make sense in the DST context?

It is true that most economists would regard the corporate income tax as falling on shareholders or on employees or both, while a consumption tax that does not allow for a deduction for interest or wages like the VAT would likely be passed on to consumers. But that distinction does not necessarily hold in the case of the targets of the DSTs, which are the U.S. digital giants.⁹

There is considerable evidence that most of the corporate tax in the United States falls on economic rents — that is, above-normal profits that are not subject to market competition. Edward G. Fox has shown that if the corporate tax were modified to allow expensing before 2017, there would be very little difference in the amount of revenue collected, and since expensing means the normal return to capital is exempt, that indicates that most of the tax falls on rents.¹⁰

That is particularly true for the digital giants (Alphabet, Amazon, Meta, Netflix) because they are close to being monopolies, and monopolies earn above-normal returns. That is the reason economists regard taxing them as efficient, because increasing the tax rate would not affect their behavior.

The implication of this argument for the incidence of the corporate tax is that it falls on the shareholders of the corporation, not on employees or consumers. The reason is that a monopolist would mostly increase its prices and reduce its wages to the maximum extent the market will bear before the tax, and therefore the tax will not in most cases be passed on to consumers or employees but rather reduce the profits of the shareholders.

In most cases, the threshold for the DST is set so high that only the U.S. digital giants are subject to it. Because they are close to being monopolists, they will not increase prices in response, and that means that under Oosterhuis’s own analysis the DST should be creditable because, like a net tax, it “maintains neutrality between purely domestic and cross-border economic activity,” and crediting it is not “a windfall to taxpayers who can in fact pass on much of the cost to customers.”

Oosterhuis also explains that “the DSTs do not qualify as in lieu of taxes under section 903 because they are in addition to rather than in lieu of income taxes for residents,” citing the France and U.K. DSTs that “in form” apply to residents as well as nonresidents. That is the IRS position.¹¹

But this is a purely formal argument.¹² The whole premise of the U.S. critique of the French and U.K. DSTs was precisely that because of the threshold they only apply to the U.S. digital giants and are therefore discriminatory.¹³ It is strange for the United States to deny the French and U.K. DSTs creditability because they formally apply to residents, while permitting a credit for the Indian DST because it formally applies only to nonresidents, when in practice all of them only apply to the U.S. digital giants. Under this analysis, the Canadian DST, which has a low threshold and applies also to Canadian companies, would not be creditable because it is actually not discriminatory.¹⁴

Moreover, one needs to consider what is the rationale for the creditability of in-lieu-of taxes. Section 903 was added to the IRC because withholding taxes, which were a common method of taxing the passive income of nonresidents since the beginning of the income tax, could not qualify as income taxes because they did not permit deductions. In the United States, withholding taxes were initially just an enforcement device, and nonresidents were supposed to file a return to claim their deductions and receive a refund. But by 1936, the IRS realized that it could not audit these foreign deductions, and the withholding tax was converted to a final tax.¹⁵

Thus, in-lieu-of taxes are creditable when they are a way of taxing income that would normally be taxable under a net-based income tax but cannot be taxed because of some impediment. This is exactly the reason that DSTs arose: The digital giants were earning billions of dollars from their operations in market jurisdictions but could not be taxed because of the obsolete permanent establishment requirement, which is even older than the international tax regime (it goes back to the 19th-century tax treaties). Thus, DSTs are exactly what in-lieu-of taxes are supposed to be: a gross-based tax enacted because there is substantial income of nonresidents that cannot be taxed under the normal net-based income tax.

Oosterhuis argues that even regular withholding taxes should not be creditable if they do not allow for an election to be taxed on a net basis and if their rate is high compared with the net rate. He is right that the U.S. withholding tax rate of 30 percent is very high when the net corporate rate is only 21 percent and the top individual rate is 37 percent, although few taxpayers actually pay the 30 percent rate. But that is not true for DSTs, because their low rates (typically below 5 percent) are far below the net corporate rates of the taxing jurisdictions and translate to a low effective tax rate given the very high profitability of the digital giants, which is likely to rise even higher with the advent of AI.

To sum up: DSTs should be creditable under section 903 because (a) they normally are borne by the shareholders of the taxpayers, not by consumers or advertisers, and (b) they are a valid in-lieu-of tax for the net income tax that should have been imposed on taxpayers but for the PE limitation. The attribution rule, which seeks to deny them creditability, does not have any statutory basis and should be withdrawn.¹⁶

But would that not result in massive revenue losses to the United States? No, for three reasons.

First, the DST rates are relatively low and are likely to stay so because they are not primarily a revenue source but a solution to the perceived unfairness of not taxing the digital giants. In some cases the taxpayers may be able to shift them to consumers, like Amazon claimed it did for the French DST, and shifting a higher-rate tax would make the voting consumers unhappy. They pay enough VAT as it is.

Second, even if every other country adopted a DST, they each would only tax the digital services provided to their residents, and therefore there would be little duplication. Even if the entire profit of a digital giant outside the United States was subject to a 5 percent gross tax, that would not put a significant dent in its profits (although given that taxing rents is efficient and given that the tax falls on rich shareholders, even a very high level of taxation is justified).¹⁷

Third, the FTC is limited to the U.S. tax rate on foreign-source income, and the U.S. corporate tax rate is lower than the OECD average. A foreign country that tried to increase its DST rate at the expense of the U.S. Treasury would likely hit the limit, and the digital giant could then credibly threaten to withdraw its services from that country rather than suffer double taxation.