Yesterday we completed the NYU Tax Policy Colloquium’s (and my) 24th year of operations with a paper on a hot topic, Wei Cui’s The Digital Services Tax: A Conceptual Defense. The paper makes creative and important contributions regarding how best to think about a new fiscal instrument that has been attracting a lot of attention, most of it hostile. Digital services taxes (DSTs) are a rich topic that I will probably be thinking about some more in the near future. E.g., I’m scheduled to give an international tax policy talk in Singapore next January, and my lecture topic will likely have something to do with DSTs, whether as themselves the prime topic or as an important subordinate / illustrative piece thereof.
There are some actual DSTs out there, in varying stages of being adopted or proposed. There are also related instruments that appear not to be quite the same – for example, Singapore recently enacted what appears at an initial glance to be an extension of their VAT/GST to digital services, requiring certain foreign suppliers to register and participate. That’s not quite the same thing, and indeed might be relatively uncontroversial as a tax policy matter.
In this post, I’ll lay out some preliminary structure that might aid one in understanding Cui’s arguments and the DST debate more generally. For reasons of overall length, I’ll continue the discussion in a follow-up, Part 2 post.
1) What is a DST for the paper’s purposes?
a) Conceptual overview – In principle, whether or not perfectly in practice, a DST within the paper’s conceptual defense is a gross receipts tax on the use of a digital platform to earn location-specific rents in a given country, via an (at least) two-sided business model, and featuring both non-rival deployment within the country and zero marginal cost. The paper recognizes that some of these features will never be completely true, but, as noted below, if they are sufficiently true they might be highly policy-relevant.
b) Whom might one have in mind as companies potentially subject to such a tax? – Before unpacking this a bit more, let’s give a sense of the territory, by noting some well-known companies that might be subject to a DST – although, in fact, many of them are deliberately exempted by existing versions.
Consider Facebook, Google, Amazon Marketplace, AirBnB, Uber and Lyft, Microsoft and Android, Apple, Netflix, PayPal, Spotify, LinkedIn, EBay, Expedia, YouTube, and Seamless and GrubHub if they went international. All are potentially within scope at least as to some of their operations, if one conceptualizes the DST as Cui does.
To be clear, here, many of the above companies have been deliberately left outside the scope of existing DST proposals. In some cases, the DST proponents have had express rationales for the exclusions that reflect their viewing the instrument’s purposes differently than the paper does. For example, they may focus on whether content or valuable data is being provided by local customers, or instead is being supplied from outside the jurisdiction (a la Netflix and YouTube). This reflects rationalizing the DST differently than the paper does.
It will be illuminating below to compare a couple of these companies with one that presumably would NOT be subject to any DST, yet that shares some (though not all) of the key economic features noted above. The company I have in mind for this purpose is Starbucks, although I’ll also briefly mention American Express and Visa.
2) Illustrating and rationalizing each of the above DST’s main features
a) Gross receipts tax – Suppose the taxing jurisdiction is the UK, and that the taxpayer is either Facebook or AirBnB. The tax as envisioned by Cui would fall on the company’s gross revenues from exploiting the UK market. For Facebook, this would be advertising revenues that Facebook earns from third party advertisers (wherever located) that are targeted to UK consumers. For AirBnb, the tax would fall on fees paid to it by guests (wherever located) who used it to book UK accommodations. (In other words, this would be the excess of what they paid AirBnB over what it then remitted to the host.)
b) Digital platform – According to a standard online definition of digital platforms, they are “online businesses that facilitate commercial interactions between at least two different groups, with one typically being suppliers and the other consumers.” This obviously fits most of the companies that I listed above as potentially subject to the DST, although for some it either applies only to some of the company’s operations, or is debatable. For example, Netflix presumably isn’t intermediating between two groups as to its own content, but (along with Spotify) could be seen as doing so for commercially released stuff that it allows one to stream. Apple isn’t intermediating when it sells iPhones, but is doing so as between app developers and iPhone owners. PayPal and Venmo are playing a similar market role to American Express and Visa, but the latter aren’t using a digital platform in the same way as the former. Starbucks has physical stores and really isn’t using a digital platform at all (unless I am missing something), although it’s true that the stores may attract some customers via free wireless.
Who cares, for tax design purposes, whether or not one uses a digital platform? It doesn’t or at least shouldn’t, directly matter at all. It’s just one way of doing business, and there are others. But the use of a digital platform may be highly correlated with various features that do matter.
Just to name one for now, although more will emerge below, digital platforms can make it easy for a company to reach a given country’s consumers (or suppliers, as in the AirBnB case) without having the sort of physical presence that can give rise to a “permanent establishment” (PE) for tax purposes. Both tax treaties, and numerous countries’ domestic income tax laws, treat the existence of a PE (or of something similar, such as a “U.S. trade or business” under U.S. income tax law) as a precondition to taxing non-resident companies on the profits that they earn from interacting with one’s resident consumers and/or suppliers.
Suppose one thinks that a PE really oughtn’t to be necessary in order for a given country reasonably to tax non-resident companies on the profits that they earn from interacting with one’s residents. It’s rather an obsolete concept, from the days when interaction without physical presence was costly and cumbersome, and when nonresident businesses might find it difficult to understand, or costly to comply with, a given country’s tax regime unless their people were there. Then companies with digital platforms would be at least among those that one wanted to tax and that found it easy to avoid a PE. But seemingly the most direct response would be to eliminate the PE requirement, not focus directly on the presence or absence of a digital platform.
To illustrate, suppose that Starbucks used franchising to place stores in a given country such as the UK, rather than operating its own store. Now, I presume that the main reason they don’t do this is that it simply isn’t the business model they prefer to maximize profits from their brand. As explained by Ronald Coase’s theory of the firm, they have apparently concluded that direct control over employees is the best way to get the behavior and incentives that they want. But if they DID use franchising and independent agents, and if this worked as well from a business standpoint as what they actually do, then they might be able to avoid having PEs outside the United States. Yet they would not be a penny less globally profitable than they actually are (again, under the assumption that doing business this way worked just fine).
In that scenario, the UK would of course be taxing Starbucks’ UK franchisees on the profits that they earned from the UK stores. But let me offer a guess: these profits wouldn’t be very impressive. Indeed, at a first approximation they’d resemble those currently earned by Starbucks’ UK employees (adjusting for the allowance of normal returns to any capital that they supplied or entrepreneurial risk that they bore). Or perhaps a better comparison is to the folks running rival coffee establishments that couldn’t take advantage of the Starbucks brand. After all. why would Starbucks give away to them any more than it had to, given their wage rates and opportunity costs that had nothing to do with the value of the Starbucks brand?
Among other things, this point arguably puts a rather low ceiling on the amount of profit that would end up in Starbucks UK operations for corporate income tax purposes, under its existing commercial structure, if arm’s length transfer pricing principles were neutrally and “accurately” deployed with respect to its UK PEs. Starbucks may be earning huge global profits, but not by reason of what’s happening in the UK in particular. But this doesn’t tell us whether either “we” from the standpoint of global welfare, or the UK from the standpoint of its residents’ self-interest, ought to support the UK’s getting its hands on what one reasonably deemed to be the UK piece of Starbucks’ global hyper-success.
c) Location-specific rents – Suppose one takes here for granted the point, admittedly requiring further analysis, that a host of highly successful global companies, such as all those that I named above with digital platforms, but also the likes of Starbucks, are earning extra-normal returns that we can properly call rents, within standard economic parlance. And let’s ignore for now the issue of whether, while looking like rents ex post, they are better thought of as quasi-rents that, from an ex ante perspective, merely reflected an ordinary, competitive, risk-adjusted return to intermingled labor effort and capital investment.
From at least the ex post perspective, taxing rents is efficient and in theory will not lead to efficiency costs. For example, if I can costlessly pocket $1 billion (e.g., by reason of getting extra-normal returns on investments for which I only required the normal return), then taxing even 99% of it away won’t deter me from still pocketing it all so that I can keep the residue.
If either Facebook or Starbucks is properly viewed asd earning rents around the world, then those it is earning in any particular place might be called rents specific to that location. For example, we might say that Facebook has UK rents from selling advertising that is targeted at UK users (as identified by their IP addresses or whatever). And Starbucks has UK rents from the profits it derives, given the value of its trademark etc., from operating stores in the UK. (Or, one might instead or also say, from selling to UK individuals anywhere in the world, if this fact automatically emerged in the course of a normal customer transaction.)
d) Non-rival deployment – Crucial to the idea of calling Facebook’s or Starbucks’ UK rents location-specific is the view that its IP can be simultaneously deployed everywhere. A given car, or for that matter a given espresso shot using particular Starbucks beans, can only be sold in one place. But Facebook’s platform can be offered and used around the world in non-rival fashion. The same holds for Starbucks’ brand value and basic organizational scheme. Thus, taking as given that the two companies are earning rents, they needn’t ask themselves where they should try to earn these rents. The answer is everywhere. There is no opportunity cost to entering the UK market – in the pure case, their doing so has no impact on their also doing so everyplace else. So if the global rents that they can earn include some that we have defined as location-specific to the UK, then of course they will choose to do so, (at least within the model’s contours) even if these rents are highly taxed.
e) Zero marginal cost – A further and related key assumption, in terms of making an appealing theoretical case for something like a DST, is that the firm has zero marginal costs of operating in a given jurisdiction. If Facebook already has a digital platform at the ready, then one could imagine its costing the company zero to allow UK users to access the platform, and outside advertisers to pay for targeted access to these users. In practice, this of course is not likely to be literally true. For example, Facebook may need to be able to handle more traffic by reason of its being in the UK, and may need to address U.K. regulatory issues, along with users who speak in English dialect (rather than, say, American or Australian).
Zero marginal cost matters because, where found, it takes away the sting from DSTs’ being a gross receipts tax. Gross revenue equals net revenue if there are no marginal costs. So one avoids the problems with gross turnover taxes that made their decades-ago replacement by VATs a huge step forward for various European and other countries. And even if there are some marginal costs, so that the gross revenue facing DST exceeds true net revenue, corporate income tax accounting efforts to measure the latter, especially for a given location within a company’s global operations, may conceivably lead to answers that are actually further from the mark. So there may be a case for gross revenue taxes on digital platforms if we believe that net revenues from its deployment in a given country are both small enough in truth, and tax-manipulable enough in practice.
What about Starbucks? Well, a tax on gross revenues from its UK operations would be normatively unappealing, given that it clearly does have substantial marginal operating costs, such as from paying rent, wages, and for supplies. So its not using a digital platform has a lot to do with why placing it under a pure DST-style regime would not make sense. Yet there is an element of its operations that overlaps with the DST rationale. After all, analogously to Facebook, it might be viewed as earning location-specific rents in the UK from non-rival deployment there of its global brand value.
Suppose that the UK corporate income tax was inevitably going to under-tax both Facebook and Starbucks, relative to a UK-only firm that had no rents. I am here using “under-tax” to refer both to the unique profit-shifting opportunities that multinational firms often have – although the concept of “profit-shifting” implies a judgment as to “true” UK-source income – and to the rationale for imposing higher taxes on rents than on normal returns. Also, I am ignoring here the considerable flexibility that an instrument which we chose to label as the “corporate income tax” might in practice actually have – at least, if not overly constrained by income tax treaties.
Then applying a DST to Facebook, but nothing new to Starbucks, might combine the virtue of raising revenue efficiently from the former, and better aligning its taxation to that of the UK-only firm, with the vice of newly tax-disfavoring it relative to the latter. This is a tradeoff that DSTs will almost inevitably offer, even where the theoretical conditions for favoring them are pretty well met, if the underlying concept couldn’t somehow be suitably generalized and adapted to settings outside that of digital platforms.
f) Two-sided business model – There is an extensive literature (not currently all that familiar to me) concerning two-sided business models, and referenced by Cui’s paper. A key issue here is that the intermediary, such as a digital platform, doesn’t have to charge an independent arm’s length price to each side of the arrangement. It can subsidize one side and recover by charging the other side, because (a) it only cares about its net from the two together, and (b) things are suitably adjusted as between the other two parties in their indirect dealings with each other.
The Facebook example is: We get to use the platform for free because Facebook is delivering us to advertisers (or others through whom it monetizes the data it has acquired about us).
This of course isn’t a newfangled product of modern digital platform technology. TV and radio stations that broadcast for free to non-subscribers but that are funded by commercials are doing a version of the same thing.
The AirBnB example is: the company offers value to hosts who sign up, but doesn’t directly charge them anything, because it can collect fees on the consumer side (i.e., the gross amounts guests remit to it, minus what it then pays forward to the host).
While I’m planning to read up more on the literature concerning two-sided business models, my impression is this feature doesn’t play a first-order role in the DST debate. Rather, it matters here mainly via its effect on the application of conventional corporate income taxes to platform companies.
In the Facebook example, the UK can’t glom onto the fees paid by users to Facebook, as a ground either for taxing those fees or using them to establish a broader PE from which it can stake its claim to other revenues or profits that are attributable to Facebook’s having UK users.
In the AirBnb example, the fact that the company gets nothing directly from UK hosts, as distinct from the hypothetical case (with identical net cash flows) in which guests paid hosts directly and the hosts then remitted a piece to the company – or, for that matter, from that where hosts paid it sign-up and subscription fees – similarly at least complicates things for the UK tax authorities, if they would have liked to get revenues from the company, say through the corporate income tax.
Two-sided business platforms therefore strike me as potentially important to the motivation for imposing DSTs, but less so as to how one might think about DSTs in fundamental economic terms.
Okay, that’s plenty long enough, if not too long, for a single blog post. It offers grist for both sides with respect to DSTs’ good and bad efficiency properties. I’ll shortly (or when I can) post a follow-up that looks to the equity issues – meaning here inter-nation equity, not say vertical distribution – aka, who gets the money and how should we think about that issue.