Does data even have any value?

This determination, of course, requires some common acceptance that raw data, before any processing by the company, does in fact have some inherent value. This would seem logical, just as any primary commodity has some value before it is transformed into a final product (for example, crude oil, which has value in its raw form upon extraction) before any processing.¹⁵

Many digitalized businesses have, however, claimed that such data is only valuable once it has been analyzed and processed—that is, that the real value comes from the application of intangible assets, such as algorithms and coding, which are used to interpret data and provide the revenue-generating service. The latter reasoning has been used to undermine the assertion that a portion of the profits of highly-digitalized businesses can be attributed to users, and to instead claim that all value creation is done at the level of the firm that manipulates the data. While it may well be the case that a large portion of the value realized from data is generated in the processing stages, it may equally be argued that without any user data to process there would be no value realized at all.

Moreover, user data is a nonrival but excludable “club” good—similar to other inputs such as “know-how” or “ideas” that feature in endogenous growth models (Romer, 1990)—which benefits only those businesses that collect, buy, trade, and process it.¹⁶ The company would not be incentivized to collect user data—by providing services for free—unless there were some value to doing so. The fact that companies maintain proprietorship over the data that they collect on their users is one facet that allows companies to secure market power and extract (and preserve) rents. If all user data were to be made publicly available, the value derived from this excludability property—and the competitive advantage it confers—would disappear.¹⁷

Determinants of data value

With this in mind, what are some of the attributes of data that determine its value? From the perspective of the firm, user data is a very heterogeneous good and—and therefore its value—is likely to depend on the type of digital service being provided, as well as the characteristics of the data, such as quality, utility, and the availability of substitutes. For example:

• The value of user data will depend on its vintage, quality, and sensitivity (for example, data value may change over time in response to security issues, litigation, or legal regulations which affects how it can be used).

• The value of user data will vary from one individual to another based on their economic profile, e.g., their affluence (purchasing power), propensity to spend, and consumption habits, and therefore the extent to which their engagement with a platform can be monetized (HMT, 2018). In this way, the average value of data will also vary by country based on the distribution of its users and its economic size, and so on.

• The value of user data can change around key life events, such as marriage or childbearing. These events can present lucrative advertising opportunities.

• The value of user data can depend on the intensity of a user’s engagement with a service. Some users are much more active and generate significantly larger data volumes than others—increasingly accurate user profiles may make them more or less valuable to a company.

• The more users a business has, the more valuable the aggregate dataset collected for the purposes of inferring trends and preferences. That is, the value of data is likely to increase more than proportionately with the volume collected—this non-linearity is driven by network externalities that are discussed a later in this section. This will also vary by firm, as well as by the amount of data already in the firm’s possession.

As we can see, given that users can generate data of differing quantity and quality, valuing data can be complex and highly context dependent. Moreover, since raw user data is not widely traded on markets but often stays within the firm that has produced it, it is hard to value, because its economic usefulness depends on the individual capacity of their acquiring firm to distill relevant information from it and subsequently to use this information to meet customer needs. Companies can also combine publicly available and proprietary data to create unique data sets for sale or use.

Conversely, from the perspective of the user, how they value their data will depend on their preferences regarding data privacy, as well as the value they place on the service which they are trading their data for—both the initial service as well as the customization of that service which may be derived from previous user activity, for example, better targeted advertisements. Just as for firms, these values will vary by individual, and will depend on the degree of trust in the firm’s integrity regarding data usage. For those that value privacy increasingly highly, the net benefit of using digital services will be lower. In addition, the perceived profits accruing to the businesses that harness this data could factor into the user’s willingness to engage in the digital barter.

Imputing the value of data

At present, there is very limited information on how data is or could be valued. Even though private markets exist in which data brokerage business sell information about consumer purchase habits, little is known about the value of such sales. It is possible to get some idea of the value of user data—much like intangibles—from past mergers and acquisitions, as well as bankruptcies. Examples include the sale of RadioShack’s data in February 2015 following its bankruptcy; Microsoft’s acquisition in December 2016 of LinkedIn (approximately US$ 260 per user); and Facebook’s acquisition of Instagram in April 2012 (approximately US$ 20 per user) and WhatsApp in February 2014 (approximately US$ 42 per user).

Small-scale efforts have emerged to create a user-centric personal data economy that empowers individuals to retain control over their data and in some cases monetize it themselves (for example, Datacoup and Meeco). The converse approach is the “pay for privacy” model, where consumers pay an additional fee to prevent their data from being collected and mined for advertising purposes (Elvy, 2017). Although not yet widely used, these efforts also provide some initial indications of how users value their own personal data.

The issue at hand is therefore how to value a commodity for which there are no market prices or established benchmarks. Open and transparent markets in which standardized units of user data can be traded have failed to develop so far, preventing businesses and governments from quantifying data inputs to production. Institutions, regulations, and transparency are therefore needed to build economy-wide (minimum) rules and establish benchmark values for different types of data. Then, just as for oil, adjustments can be made to determine country-specific values for user data.

In the absence of such benchmarks, policymakers might look to more observable proxies of user data value. Some proposals have looked to the final revenues of companies from certain activities. However, unlike natural resources, data is not a physical commodity that can be traced from the point of extraction to a specific point of consumption. Rather it will likely be blended, processed, and analyzed for use in delivering a wide range of goods and services in a range of different countries. In some cases, profits may be realized from data (through its use in product improvement or refining machine learning processes) many years after it has been collected. In these cases, it may be almost impossible to determine ex-post which portion of the profits are attributable to the initial user data, even though it provided a critical input. The nonlinear value of data poses an additional challenge in establishing a method to “netback” from these final revenues to the initial information used as inputs to digital services.

One exception may be the case of advertising, where data is used as an input in determining the placement of person-specific (and, therefore, location-specific) advertisements.¹⁸ In principle, each advertisement viewed by a particular user has an associated value, determined by auction according to the willingness to pay by online retailers, which would presumably correlate closely with the relative value of the initial data provided by that user to the business. Of course, the blending of data from different sources, possibly across different countries, when determining how to best target advertisements would mean that the value of the advertisement is only a proxy for the value of that particular user’s data.¹⁹

More generally, if revenues are used to derive a proxy for user value, what would be necessary then is an agreement on the size and scope of netback deductions, reflecting any post-data collection processing, in order to determine the value of the initial information used as inputs to these services. In the case of extractive industries, countries often use the net smelter return concept, taking the international market price of the final refined metal product as a benchmark and making deductions for treatment and refining costs to derive the value of the mineral at the point of extraction. We will return to this analogy in more detail in Section IV.

User value from network externalities

Beyond the use of personal data in production, the other much-cited source of user-generated value is the role that users play in creating valuable networks for digitalized businesses. The success of digital platforms relies on the interaction between both buyers and sellers, which gives rise to strong complementarities—notably, network and information externalities—where the value in transactions increases for both groups as the numbers on each side increase (Armstrong, 2006; Caillaud and Jullien 2003; Ellison and Fudenberg 2003; Evans 2003; Rochet and Tirole 2003; Rysman 2009).

Of course, such network effects have existed well before digitalization (Katz and Shapiro, 1985, 1994; Liebowitz and Margolis, 1994).²⁰ For example, the underlying idea of a ride-sharing application is that each individual user adds to the pool of users that make up the network and thereby increases the overall value of the service. This concept applies to a number of both digital and non-digital networks. In a non-digital setting, a farmer’s market is more valuable to buyers and sellers alike, if there are more farmers and more shoppers participating. Newspapers and television networks are another widely-referenced example where network externalities exist between the market of readers and advertisers. Collier (2018) and Collier and Venables (2018) also highlight how the externalities associated with urban agglomeration produce rents that accrue to a small group of individuals and businesses (the “urban surplus”).²¹

Digitalization has exponentially increased both the scale and the speed with which network effects ratchet up. Users posting on social media provide content that can quickly attract other subscribers, and the more users engage with one another on the platform, the more content they create and so on (OECD, 2018). And the more users there are on the platform, the more the firm will invest in its development. There are also information externalities: users on retail platforms provide feedback through reviews that influence other users of the platform in their consumption choices. While users might be acting in their own interest, their actions both attract other users and reveal more and more commercially-valuable information about themselves, creating increasingly valuable opportunities for marketing and product improvement. Other digitized markets also display such scale effects, such as in the financial services sector (for example, credit card services) and in the insurance market where liquidity and heterogeneity of participants is particularly important.

Given the existence of network externalities in both digital and nondigital marketplaces, they would not appear to justify ringfencing digitalized services and taxing them differently. However, some governments feel that they are a source of rents for the businesses that capture growing user bases. Cui (2018) also notes the location-specificity of network effects (both direct and indirect) as the source of a location-specific rents. Moreover, the policy implications are unclear. On the one hand, the network externalities highlighted so far appear to be positive—they generate increasing value for users that are part of the network. From the perspective of optimal taxation, a positive externality in a one-sided market would in fact lead to the under-provision of the good or service, so the government should instead subsidize it to increase its production. On the other hand, for two-sided markets, Kind and others (2008, 2010) show instead that increasing (ad valorem) tax rates—rather than subsidies—can increase output and enhance welfare.

Even if policymakers could rationalize the taxation of rents, the network externalities that generate them can be just as challenging as user data to evaluate. Cui (2018) provides a theoretical description of how user value (encompassing both data provision and network effects) could be measured. Such a method requires measurement of the demand curves for a range of products and services, and measurement of the changes in such demand curves upon the introduction of online reviews and customized advertising, from which one can calculate the increase in producer surplus arising from such changes. As we proceed to examine how a system which allows countries to tax user generated value might operate, we leave aside the notion of user generated network externalities, and instead focus on value generated from user data.

Taxation or Regulation?

More generally, one might question whether tax is the right tool to address the issue of seemingly excessive rents generated through network externalities. Markets where network effects exist without any interoperability between providers typically tend towards monopoly, in the absence of any intervention.²² As a result, there has been a rapid consolidation of power among certain digitalized businesses—such as internet search and social media firms—which have been able to develop some of the largest user bases thanks to these positively reinforcing network effects. Facebook, for example, has over 2 billion users as of 2019, which is about 70 percent of the world’s population that has access to internet.²³ This market concentration has reinforced the ability of only a handful of digital platforms to monopolize the aggregation and analysis of large amounts of personal data.

Indeed, when markets tend to natural monopoly, taxation is not typically the optimal policy response—regulation is typically the first best. The approaches can vary from breaking up monopolies into smaller units (e.g., AT&T’s Bell system in 1982) to price regulation—some of which can create their own distortions (Newbery, 2000). At present, few of the recent wave of digitalized businesses have been subject to any form of anti-trust regulation in the U.S. The previous notable case was that of Microsoft in 2001, in which it was asserted that the company was using its control over the personal computer market to force out competing operating systems and browsers. Google could be considered a potentially similar case, given its dominance of search and, therefore, advertising services—notably, the European Commission has already fined it for manipulated search results (EUR 2.42 billion in 2017) and for bundling activities (EUR 4.34 billion in 2018).²⁴

In the case of Amazon, some have argued that the company has become a utility—given its distribution infrastructure, which is used by many other businesses—and is engaging in anticompetitive actions by pricing below cost to eliminate competitors or force their acquisition by sale (Khan, 2017). While this has not triggered reaction thus far, as the impact is admittedly to the benefit of the consumer, this “antitrust paradox” is reducing competition across the retail sector. Price regulation could be a natural response by regulators to such tactics.

What is ultimately important to note is that the economic features of platform-based digitalized businesses can incentivize rapid growth over profits in the short term to secure market dominance (Annex I). This tends to make anti-competitive behavior, such as predatory pricing and the large-scale acquisition of competitors, a rational strategy. Furthermore, for those sorts of digital platforms where network effects lead to substantial market share through access to excludable user data, and where breaking up companies into smaller (more regional) units does not necessarily solve the problem, price controls—where applicable—are the more efficient tool. Ultimately what works for one type of digitalized business might not be appropriate for others, making a business-by-business (or sector-by-sector) approach necessary. For example, specific regulation bills—akin to the U.S. 1996 Telecommunications Act—that govern a wide range of activities, from data collection and privacy to ethical content, might be necessary for those companies that provide certain types of universal services, e.g., messaging and social media.

As Grinberg (2018) observes, the nature of these markets suggests that tax and regulation need to be appropriately distinguished in the digital economy debate. With ever-increasing market share, rents from monopoly power over information are likely to increase. By resorting to taxation as a means for redistributing the rents being generated by user-derived network externalities, the question is whether tax policy is being used to treat a symptom, whereas regulation might be more appropriate to address the underlying cause.

The Production and Ownership of User Data

Establishing a right to tax the income (or rents) derived from user-generated value relies on carefully disentangling the relationships between the user, the digitalized business, and data being traded between them. Typically, under rights conferred through national legislation and contractual agreement, the owner of any factor of production is directly remunerated by a business for its use and that income is then subject to tax in the jurisdiction in which the factor is located. However, with largely undefined property rights over personal data, and in the absence of legislation governing the collection of or access to data, ownership over user data is implicitly vested in the companies that collect it, with no legal obligation to compensate the individuals that have generated this data—beyond the digital service for which this data may have been exchanged. Many companies also feel justified in owning the data as it is their intellectual knowhow and technology that enabled the collection, storage, and processing in the first instance. Furthermore, once collected, non-rival user data can be used repeatedly by the business and even sold to other businesses. Without any rights or control, users are unable to extract the full return on their data.

Ultimately, there is a dilemma over who should have ownership or control over user data, either explicitly through legal property rights or implicitly through compensation for its use. Does the data belong to the user, or the firm that collected it? Should it be held in trust by the government on behalf of its citizens? If these issues can be defined more clearly, it might then be possible to determine the basis on which the user (or government) can be compensated. Recent reforms have gone in this direction, with greater attention paid to regulating the privacy of users and control over their personal data, the stringency of which has implications for both ownership and value, and, therefore, the possibility for compensation of users.²⁶

If we take the view that the underlying subject of the data, the user, is entitled to compensation for its use, then the nature of such compensation will depend on the ownership or control of the data once it has been collected. If the data is owned by the company, then the relationship between user and business could be analogous to that of an employee and employer, where the former has been actively engaged by the latter and whose time and inputs should be contracted and compensated, but who do not own their “work product”—in this case, the user data.²⁷ The level of compensation could in this case reflect the present value of the data, which may, of course, be used repeatedly by the firm over time. Alternatively, if the ownership or control over data remains with the user, then there would need to be mechanisms in place to allow businesses to access and remunerate users accordingly, perhaps either on a per-use basis or over a certain time period. From a taxation perspective, if the user is compensated directly for primary sale of or access to their data, this remuneration would be deducted at the corporate level, and the user would simply be taxed under the personal income tax schedule—in a manner similar to a self-employed business that declares its income earned.

A related issue is whether users should be individually compensated or whether the government should be compensated on their behalf. Indeed, the nonstandard and person-specific nature of data—together with the ability of individuals to move location, change nationality, etc.—might suggest that individuals should directly be compensated for access to or collection of their data. Unlike a sub-surface natural resource that is generally acknowledged to be a collective national asset, it could be argued that user data is an inherently personal asset that lends itself more readily to the notion of individual sovereignty.

However, for individuals to be able to exercise and enforce these rights requires not only a strong legal system, but also additional shifts in technology. For example, for users to have full ownership and control over their data, they would need to be able to take it with them between digital platforms, retrieving and sharing their data as they move from one digital service provider to another, or at least monitoring its use. Therefore, the nature of system and software interoperability and data portability would need to evolve.

Moreover, the nature of the exchange of user data for digital services might raise practical questions regarding the feasibility of compensating each user for their data or access to the data. As noted earlier, digital barters are not large, infrequent transactions between businesses but instead are small, high-frequency “micro” transactions between businesses and consumers across multiple separate online platforms. For example, social media platforms and search engines record billions of “likes” and queries each day, respectively. The value of the data being traded in these “microbarters” is likely to vary significantly, and the current lack of a transparent market or an agreed standardized measure for the value of such data further exacerbates the difficulty in determining and reporting a verifiable person-specific financial value for each exchange.

A more practicable alternative might be achieved if custody over citizens’ data were assigned to the government as a proxy. For example, in the case of natural resources, governments typically hold the resources in trust on behalf of their citizens and collect payment for extraction in the form of government revenues. In the case of digital services, data on the behavior and preferences of a country’s citizens would be seen as a collective national asset with compensation for its use payable to the government. Indeed, a government is better placed to exert collective power on behalf of individuals who cannot capture their rents by levying a tariff (for example, an export tax) on the extraction of data by foreign digital service providers (Hufbauer and Lu, 2018; IMF, 2019). This arrangement would apply whether data ownership is transferred to the company upon collection, or retained by the government, although the arrangement chosen (i.e., time-bound access or indefinite ownership) would have implications for the level of compensation. And indeed, while user data is intrinsically tied to an individual, the nature of data collection and the fact that the nonlinearly-increasing value from user data is often derived—not only from aggregating data related to a large number of individuals but also from multiple points in time—also suggests the practical need to centralize ownership so as to coordinate access and compensation.

If such an arrangement were developed, companies that wish to collect user information would then need to compensate the jurisdiction in which the information was collected in accordance with agreed rules, through the corporate income tax system or through other tax instruments such as a payment of a royalty, a concept which will be explored further in Section IV. Such an arrangement still does not obviate the need for clear valuation rules around the data collected but does allow for the possibility of using country-level proxies or formulary methods for determining user value, which will be explored further in this section.

Reappraising the Permanent Establishment Concept

For governments to have the right to tax the returns associated with user data through the corporate income tax system, there needs to be an explicit legal basis on which to recognize the economic activity of data collection in the local jurisdiction. The traditional basis for a government to exert taxing rights over a foreign multinational enterprise doing business and deriving profits from their activity in their jurisdiction is the concept of “permanent establishment”. It is this concept that needs to be reappraised to facilitate source-based taxing rights over value generated from user data.

A permanent establishment exists when a multinational enterprise—highly-digitalized or not— has nexus based on “significant economic presence” in a country such that their economic integration with the local economy crosses some threshold. Threshold tests are typically specified in domestic legislation and/or double tax treaties, and have been designed to determine the degree to which a foreign company is integrated locally. Typically, the test specifies that a permanent establishment exists if the multinational enterprise has a fixed place of business within the target country or if there is a dependent agent that regularly exercises the authority to conclude contracts locally on its behalf.

The collection of data from users—regardless of whether it is in exchange for a digital service or not—is typically conducted remotely by the business without any physical presence in the country of the user and therefore does not create a permanent establishment under current definitions. While large online retailers typically operate storage and distribution centers in market jurisdictions, these are typically exempted in treaty definitions of permanent establishment (OECD, 2017b, Article 5, para 4). Local affiliates are also commonly structured to have no ownership interest in intangible assets, perform no development, enhancement, maintenance, protection, and exploitation functions, and do not assume any risks. Accordingly, only a modest return may be allocated to these “limited risk distributors”.

While activities such as local data collection or warehousing were previously considered to be of a merely auxiliary nature—typically contributing only marginally to business profits and therefore neglected for the purposes of profit allocation or justifying only low profit attributions—they now form core elements of digitalized business models. In other words, previously simple routine functions for which it was generally acceptable to allocate only a small share of the overall business profits have arguably become key activities for many firms, with a pivotal role for data analytics.

However, this surge in cross-border business-to-consumer sales as a result of digitalization has taken us into a grey area. Traditional exporters of goods and services have never been subject to permanent establishment for tax purposes in importing countries. This could be due to the fact that the majority of exports were typically from business to business, especially in the case of intermediate goods. However, the scale of growth of peer-to-peer exports, as well as exports of retail and intermediation services by highly-digitalized businesses, is straining the traditional permanent establishment concept. As businesses increasingly digitalize, the user in the importing country generates valuable inputs for an increasing number of exporters. By ignoring these issues, a consequence of a tax system that relies solely on physical presence to assert taxing rights is to create variation in the cost of capital between businesses exporting to and those producing domestically for the same market.

Any solution that calls for a different allocation of profits beyond what the current system enables will need to be supported by new or expanded taxing right. As has also been acknowledged under the OECD’s (2019b) Pillar One proposal (the “unified approach”), the current definition of permanent establishment would need to be modified to establish the right for user-jurisdictions to tax multinational enterprises that collect data on their citizens. Until then, a multinational enterprise can sidestep the nexus issue by operating remotely, or establishing local affiliates that are not entitled to an appropriate share of the group’s profit.

Many have already put forward ways to modify the permanent establishment concept. For example, online retailers, warehouses, and distribution networks developed to cater to users (buyers and sellers) were previously excluded from the definition of the permanent establishment in the existing version of the OECD Model Treaty (OECD 2017b). However, Action 7 of the OECD/G20 Base Erosion and Profit Shifting (BEPS) Project and Article 12 of the Multilateral Instrument provide a basis for an extensive application of the permanent establishment concept to account for the fact that these functions are no longer merely auxiliary, but instead form a strategically decisive component of the business model. Some countries have also already taken steps to change the permanent establishment threshold tests in domestic legislation to include them (Table 1).

^{Table 1.}

Country Approaches for Significant Digital Presence

^{Table 1.}

Country Approaches for Significant Digital Presence

Country	Digital Permanent Establishment Definition
European Commission (Proposed)	A digital platform will be deemed to have a taxable ‘digital presence’ or a virtual permanent establishment in a Member State if it fulfils one of the following criteria in a taxable year: >EUR 7 million in annual revenues; >100,000 users; >3000 business contracts for digital services.
India	In April 2019, a significant economic presence test was proposed, where (i) domestic transactions in respect of goods/services/data exceed a threshold for the previous year, or (ii) where there is a systematic or continuous soliciting of domestic business activity or users by digital means. A foreign entity meeting either of these tests will have a permanent establishment, regardless of whether they have any physical presence. However, this rule will not be effective in cases where the foreign entity is located in a country with which there is a tax treaty. As of February 2020, the application of this test has been deferred to April 2021.
Israel	In April 2016, a significant economic presence test was introduced which asks: (i) whether there are a significant amount of contracts with domestic customers for internet activity; (ii) whether there are a significant amount of domestic customers using the digital service; (iii) whether the online service is adapted for use by domestic customers (for example by using local language and currency), (iv) what is the level of internet activity by domestic customers, and (v) whether there is a correlation between use by residents and payments to foreign entities.
Nigeria	In January 2020, the Finance Bill signed into law the principle of significant economic presence to the basis of taxation of non-resident companies operating in the digital services and e-commerce sectors.
Slovakia	In March 2018, the definition of a fixed place of business was expanded for foreign online platforms offering transport or accommodation services in the country.

View Table

^{Table 1.}

Country Approaches for Significant Digital Presence

Country	Digital Permanent Establishment Definition
European Commission (Proposed)	A digital platform will be deemed to have a taxable ‘digital presence’ or a virtual permanent establishment in a Member State if it fulfils one of the following criteria in a taxable year: >EUR 7 million in annual revenues; >100,000 users; >3000 business contracts for digital services.
India	In April 2019, a significant economic presence test was proposed, where (i) domestic transactions in respect of goods/services/data exceed a threshold for the previous year, or (ii) where there is a systematic or continuous soliciting of domestic business activity or users by digital means. A foreign entity meeting either of these tests will have a permanent establishment, regardless of whether they have any physical presence. However, this rule will not be effective in cases where the foreign entity is located in a country with which there is a tax treaty. As of February 2020, the application of this test has been deferred to April 2021.
Israel	In April 2016, a significant economic presence test was introduced which asks: (i) whether there are a significant amount of contracts with domestic customers for internet activity; (ii) whether there are a significant amount of domestic customers using the digital service; (iii) whether the online service is adapted for use by domestic customers (for example by using local language and currency), (iv) what is the level of internet activity by domestic customers, and (v) whether there is a correlation between use by residents and payments to foreign entities.
Nigeria	In January 2020, the Finance Bill signed into law the principle of significant economic presence to the basis of taxation of non-resident companies operating in the digital services and e-commerce sectors.
Slovakia	In March 2018, the definition of a fixed place of business was expanded for foreign online platforms offering transport or accommodation services in the country.

Separate Accounting

Many criticize arm’s length pricing for a number of reasons. First, it runs into practical implementation difficulties when balance sheets comprise significant amounts of unique, highly-valuable yet hard-to-value intangible assets, as is the case for highly-digitalized companies. It is relatively easier to discover a reasonable arm’s-length price of a barrel of crude oil for which there is a spot market—that is, some form of comparable asset. It is less clear what the comparable is for unique intellectual property, such as patents and trademarks. For digitalized businesses, no matter how assiduously one performs “functional analyses” designed to identify “uncontrolled comparables” that are reasonably similar to transactions undertaken between members of multinational groups, they are hard to find.

Second, multinational enterprises frequently use this system to reduce the amount of tax they pay by diverting profits offshore through contrived structural arrangements between related parties, such as intellectual property transfers and related royalty arrangements, marketing/distribution/procurement hubs, and offshoring services. As a result, profits are often not being taxed even once in existing source or residence jurisdictions thanks to the use of transfer mispricing and no-or-only-nominal tax jurisdictions, violating the single tax principle.³⁵

Third, separate accounting systems treat each affiliate of a multinational enterprise as a distinct entity with its own costs and incomes. However, allocating income and expenses across countries is not only complex because of the lack of comparables, but also conceptually unsatisfactory, given that worldwide income is generated by interactions between affiliates across countries. Multinational enterprises exist in large part because these interactions generate more income than separate domestic firms interacting at arm’s length. Requiring firms to allocate this additional income among domestic tax bases is necessarily artificial and arbitrary, therefore, because it would disappear if the related entities operated at arm’s length.

In the case of user data collection, the separate accounting approach would involve establishing the associated income and costs for the multinational enterprise, as if the data collection agency were a separate enterprise. However, as digitalization leads to the pooling of data across countries, separate accounting adds additional layers of complexity on top of the aforementioned valuation issues (Section II.C). Opportunities for transfer mispricing would likely increase even further, given the difficulty of determining precisely how much of the enterprise’s income is attributable to data collected by affiliates in one country versus another.

In sum, the application of arm’s length pricing is not only increasingly difficult but unrealistic as intangible assets come to dominate company activity and value. The irony is that multinational enterprises arose precisely to avoid the inefficiencies that arise when unrelated companies must transact with one another at arm’s length. Frustrations with their abuse of the arm’s length principle and the growing shortcomings of current nexus requirements for permanent establishment highlight the inadequacies of this approach.

Formulary Apportionment

The formulary system disregards all of these legal distinctions and simply looks at a company as a large, single unit. Based on factors such as sales, payroll, and assets, along with threshold tests, the system divides a business’ income among various taxing jurisdictions in which it is conducting activity. Intercompany transactions are therefore ignored and income-shifting curtailed, reducing profit allocation disparities between high- and low-tax countries.

Indeed, the debate over some form of formulary apportionment of profit has been ongoing for some time (see, for example, Altshuler and Grubert, 2010; Avi-Yonah and others, 2009; Avi-Yonah, 2010; Graetz and Doud, 2013). Such a realignment of taxing rights is likely to lead to a redistribution of profits across countries through the reallocation of the tax base from low to high-tax jurisdictions. It could also potentially alter the size of the pie available for distribution, due to differences in loss treatment between separate accounting and formulary apportionment, which consolidates losses at the global level before apportionment (Fuest and others, 2007; Cobham and Loretz, 2014). On the merits of formulary apportionment as an alternative to arm’s length pricing, we refer the reader to other pieces (see, for example, Altshuler and Grubert, 2010, de Mooij and others, 2019).

There are a number of strengths of applying a formulary apportionment system to accommodate increasing digitalization. Firstly, as the nature of business changes, the formula can be augmented (or reduced) to account for those factors that are most critical for determining a business’s connection to a location—and the base for its tax liability. Formulae often vary by sector, such as in Canada, where special factors and weights apply to insurance, banking, and transportation, and in the United States³⁶, where Alaska uses an origin-based sales factor for extractive industries (de Mooij and others, 2019).

Secondly, formulary apportionment does not need to apply to all profits. To the extent that some countries demand that a certain portion of profits must remain within their jurisdiction, apportionment can then be applied to “residual” profits (see next section). Again, in the case where market countries demand a share of profits given the contributions of their users, profits can be first split, with only a certain fraction available for apportionment using some sort of user-based factor.

User Data as a Factor for Formulary Apportionment

This section proceeds to focus on how a formulary apportionment system could effectively account for value created by digitalized businesses from user data. To prevent any unnecessary ringfencing of specific digital services, we must design an expanded formula for all businesses to account for new factors that have become relevant because of ongoing digitalization in every sector. And we must consider how these factors should be weighted relative to the others (i.e., sales, assets, payroll).

The specification of the formula can have important implications for a number of reasons. For example, the inclusion of certain factors, such as assets and employment, creates an implicit tax on them, discouraging their accumulation in high-tax locations.³⁷ Furthermore, not all of these factors might be relevant for gradually digitalizing businesses. Payroll is one such case, which might find itself increasingly irrelevant: given the ability to sell remotely, there can be virtually zero labor presence in certain markets and therefore zero apportionment.

The principle that taxes should be applied “where value is created” fails to provide any definitive guidance on relevant factors (Devereux and Vella, 2018). As we have discussed, there are many activities that create value for digitalized businesses on the demand or supply side. Yet not all of these activities are directly generating profit. The chosen formula can potentially combine a number of location-specific factors from both sides of the market. If measurable, these could include, for example:

• From the demand side: sales, population (total or by age); mobile phone and internet penetration (as a proxy for usage of digital services). For example, internet usage is near universal amongst young adults (<50 years), college graduates, and those from high-income households.³⁸

• From the supply side: assets (total, tangible, intangible), employees (or payroll), users (volume, value of data they provide).

To the extent that user data itself can be measured and valued, it can potentially feature in the formula. The latest draft proposal for the EU-wide Common Consolidated Corporate Tax Base includes a fourth data factor that reflects the collection and use of personal data of online platforms and services users.³⁹

One proxy for user value would be the volume of users, which can be measured in many different ways. For example, a country’s population could represent the potential maximum volume of users in each jurisdiction; alternatively, we could consider volume as that part of the population that has access to mobile telephony and the internet. However, using population as a proxy for the volume of users introduces a number of possible complexities.

First, the population of a country is not static—it can change over the course of the year for many reasons. Moreover, the user population will include foreign nationals that are resident in the country as well as its own citizens and exclude local nationals living abroad. In other words, migration flows alter the population of a country throughout the year, especially in the case of seasonal flows such as tourism, which can in many cases swell the numbers in a country at any one time—e.g., small island tourist destinations.

Second, even with accurate statistics which can be used to gauge the flows and stock of users in each country, the units used to measure “volume” need to be defined very clearly to reflect the intensity of usage by each member of the population, e.g., the number of active users in the population⁴⁰, the number of clicks during a fixed period of time, or the number of completed purchases by users in a country. In April 2019, India proposed assigning different weights to different categories of digital businesses depending on the level of user-intensity—namely, 10 percent to the users for those business models involving low or medium user-intensity and 20 percent those business models involving high user-intensity. However, the determination of intensity remains undefined.⁴¹

Moreover, it is not necessarily the case that each user is equally valuable both within and across countries. Certain income and demographic profiles would matter more for some businesses than others. For example, users from countries with higher per capita incomes would be more valuable for retail businesses in terms of level of activity and purchasing power, and suggesting that those economies should have a larger apportionment of profits.

The choice of factors can have important implications for the distribution of profits (Figure 3). Large, populous economies that are both large source and destination economies, such as the United States and China, would secure a larger share of profits under formulary apportionment with greater sensitivity to users (proxied by the size of the population). However, other destination countries with large markets would gain substantially, while small-source countries would lose. To the extent that formulary apportionment is applied universally, tradables sectors with global supply chains would be affected disproportionately.

Distribution of Alternative (user-related) Factors for 20 Most Populous Economies, 2018

(Multiples of U.S. quantities; United States=1)

Citation: IMF Working Papers 2020, 076; 10.5089/9781513545974.001.A001

Sources: IMF, World Economic Outlook database; International Telecommunications Union; UN Department for Economic and Social Affairs, Population Division, and authors’ calculations.Note: RHS=right-hand side

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Distribution of Alternative (user-related) Factors for 20 Most Populous Economies, 2018

(Multiples of U.S. quantities; United States=1)

Citation: IMF Working Papers 2020, 076; 10.5089/9781513545974.001.A001

Sources: IMF, World Economic Outlook database; International Telecommunications Union; UN Department for Economic and Social Affairs, Population Division, and authors’ calculations.Note: RHS=right-hand side

• Download Figure
• Download figure as PowerPoint slide

Distribution of Alternative (user-related) Factors for 20 Most Populous Economies, 2018

(Multiples of U.S. quantities; United States=1)

Citation: IMF Working Papers 2020, 076; 10.5089/9781513545974.001.A001

Sources: IMF, World Economic Outlook database; International Telecommunications Union; UN Department for Economic and Social Affairs, Population Division, and authors’ calculations.Note: RHS=right-hand side

• Download Figure
• Download figure as PowerPoint slide

In addition, longer-term demographic trends that mean that the share of the youngest in the population—that is, those that typically use technology and digital services more frequently and intensively—is set to expand in certain parts of the world while contracting in others. In this case, user based factors would imply a gradual reallocation profits to these economies—notably in Africa and Asia—where demand for digital services is likely to be greater.

Residual Profit Allocation

Alternative forms of apportionment have also been proposed that allocate only “residual” profit by formula (Devereux and other, 2019, Avi-Yonah and others, 2009; IMF, 2019).⁴² Such forms require distinguishing between “routine” and “non-routine” returns to investment. The routine return can be calculated in a number of ways: for example, it could be associated with the economic concept of normal profits, calculated as some markup on costs incurred by the businesses in each country or a fixed minimum return on tangible assets (e.g., similar to the United States provision for Global Intangible Low-Taxed Income). In doing so, it can also still allow for partial use of the existing arm’s length pricing approach.

The key notion is that residual profits are those harder to attribute to specific jurisdictions, that is, the profits of a business after the activities of service providers have been awarded an arm’s length return. Once calculated, these profits can be allocated using an agreed-upon formula. This allocation of residual profits, while arguably adding a new layer of complexity, eliminates the existing complexity in measuring and valuing intangible assets. While still administratively complex, such a hybrid solution may be more politically palatable and expedient than comprehensive formulary apportionment, as it more closely resembles the current system.

One application of this method in a digital setting—though not part of the current OECD proposal for a unified approach—would be to attribute a portion of these residual profits to the value created by the activities of users, using quantitative/qualitative information or through a simple percentage. This portion of profits could then be allocated between the jurisdictions in which the business has users, based on an agreed allocation factor, as discussed above.

Taxing User Value: Lessons from the Extractive Industries

The approach to taxation in the extractive sectors has evolved over time, in line with a shifting balance of power between host governments and investors. Hogan and Goldsworthy (2012) note that, prior to World War II, governments typically granted concessions to investors to explore for and extract natural resources in exchange for a relatively low burden of initial bonuses, royalties, and land rental fees. Decolonization after World War II and establishment of the permanent sovereignty principle⁴³ drove the desire of a number of resource-rich countries to receive a larger share of the resource rents, at which point fiscal regimes began to involve the increased use of state participation, income taxes, ad-valorem royalties, and other revenue instruments.⁴⁴

One might conclude that similar trends are emerging with respect to the ‘extraction’ of personal data. With largely undefined property rights and legislation over personal data, its collection and use have gone largely uncompensated, except indirectly through the taxation, if any, of overall returns under the current corporate tax system. However, governments are now exerting greater sovereignty over their citizens’ data, and seeking a ‘fairer’ global distribution of tax revenues, both through the corporate tax system and through the application of new instruments, as will be discussed in this section.

The key objective of extractive industry fiscal regimes is to maximize compensation to host governments, while maintaining the investor’s incentives to undertake exploration and extraction, given their respective risk profiles. While an objectively ‘fair’ division of resource rents remains contentious, international norms have developed over time. Regimes typically comprise a combination of production and profit-based instruments⁴⁵ to make up a combined ‘resource charge’, and thus involve an inherent trade-off between the efficiency and neutrality of the regime, as well as the timing and risk associated with revenues.

One could conceive of a similar approach to the taxation of value from user data. The design of any data-specific fiscal regime or resource charge (perhaps more appropriately an ‘access charge’, depending on whether the individual or government retains data ownership) must appropriately balance a sufficient return to the investor with the need for a ‘fair’ level of government revenue as compensation for extraction or access to the national data asset. Given the range of companies involved in the data economy and the variability in their profitability (Annex I), a degree of progressivity may be desirable, in order to capture large rents if they arise. It may also be desirable to ensure a minimum charge for all data collection activity regardless of rents, which governments would need to moderate to avoid deterring the development and growth of data-intensive companies—like natural resource companies, a high degree of risk typically characterizes the initial start-up phase, particularly given the high fixed (and sunk) costs incurred for data collection and analysis systems to be developed.

While there are many possible alternative tax instruments—or combinations of instruments—that can compensate countries for the collection and use of personal data, the focus here will be on the design of a production-based instrument, that is, a royalty on data collection. Annex IV highlights additional alternatives for taxing user value, drawn from the experiences of the extractive industries.

Designing a Royalty on User-Generated Value

Royalties are payments that companies must make to governments for permission to engage in certain activities. They are commonly charged for the right to use intellectual property, such as copyrights, patents, and trademarks, and in the extractive industries for the extraction of natural resources, and typically complement corporate income taxes. Royalties are also a useful—if nonetheless imperfect—substitute when direct taxation of rents is difficult, particularly for sectors heavily reliant on hard-to-value assets, and where administrative capacity is weak, and where rent taxation is therefore vulnerable to cost-based profit shifting (Boadway and Keen, 2015). Furthermore, by relying on revenues, they are simpler to administer than profit-based taxes, reducing the need to monitor costs. Finally, royalties sidestep the practical legal challenges associated with adjusting direct income taxes, since they do not require modification of existing permanent establishment rules or changes to income tax treaties.

Their application as a charge for the use of consumer data is therefore logical if not also somewhat ironic, since intra-company royalties are already used by multinational enterprises themselves to shift profits across jurisdictions. In the natural resource sectors, they have proven attractive to governments on the grounds that they secure some revenues as soon as production begins. Indeed, IMF advice to resource-rich member countries, particularly low-income countries, has been to charge a modest royalty alongside the statutory corporate income tax, with an additional rent-capture mechanism for particularly profitable projects (IMF, 2012).

In terms of design, a broadly-applicable royalty is likely to be desirable given the wide range of digitalized businesses collecting personal data. As the range of digital services continues to expand, it is important to prevent the proliferation of multiple royalty rates and bases across countries and activities which can generate both inefficiency and non-neutrality. This will also simplify administration and limit the scope for (purposefully) misclassifying taxable activities. Moreover, as we will show later in this section, under certain simplifying conditions, if all countries apply a consistent approach to determining user-based royalties, this would be broadly equivalent—in revenue terms—to a system of user-based formulary apportionment of profits.

There are a number of options for the design of the royalty base. A simple ad-valorem royalty based on the value of the flow of data collected from a country and used by a company in a given year would be the ideal design—this is the most common specification used in the extractives sector. Alternatively, if the firm retains ownership over the data, rather than accessing it on a use basis (as discussed in Section III), countries may wish to tax the stock of data held on a periodic basis, justified by the fact that companies are repeatedly generating value from this data.⁴⁶ However, both methods require both/either standardized data valuation methods and/or the establishment of a spot market for data.

In the absence of clear valuation methods, as long as the volume of data (stock or flow) collected from a country can be measured, a simpler approach would be to charge a specific royalty on this base. This can take the form of a defined volume-based charge on a unit of data collection, such as a pre-specified monetary charge per terabyte. This approach is often used for bulk, low-value minerals such as aggregate and sand. However, it would require some careful design and monitoring by governments to ensure that the tax is periodically adjusted to reflect evolving data compression technologies.

Without valuation methods and volume measurement capabilities, an alternative two-step approximation for user-generated value can be considered. The first step would be to estimate the portion of a company’s final revenues from digital services that were generated by a particular country’s users. Once these country-specific revenues have been determined, the second step is to allow for a (standardized) netback deduction from these country-specific revenues, to account for value added from post-data collection processing and analysis.⁴⁷ These steps provide a proxy for the value of a country’s user data, which can then serve as the base of the royalty. Indeed, in the extractive industries, countries often simplify valuation by using internationally-quoted benchmark prices for finished minerals, restricting netback costs to those that are easy to measure, or standardizing allowable deductions in contracts or legislation.

The first step requires a method for apportioning a company’s user-generated revenues by location to approximate the relative value of the user base in different countries. However, this can be difficult because as explored in Section II, data is unlike other physical commodities, which can be traced from their point of extraction to their point of use. Moreover, its nonrival nature means that the same portion of data may be blended, processed, and used simultaneously in the production of multiple goods and services—in some cases even many years after it has been collected. Section III reviewed some of the country characteristics that can be used to inform such apportionment, for example, population, purchasing power, and sales revenue (in the case of online retailers).

Without coordination amongst countries, there is a risk that revenue attribution methods may vary within and across both countries and the digital services being targeted, introducing the opportunity for multiple (inconsistent) approaches. For example, the apportionment method for advertising revenues may differ from that used for user data sales, online sales of goods, and intermediation revenues. Even within certain activities, different keys could be used by different countries (e.g., value-based versus volume-based keys). Further complications might arise from competing claims over the revenues from cross-border transactions and sales of multinational data sets.

For the second step, the netback deduction would serve as an approximation to derive a value for the underlying personal data. This could be defined company-by-company—given different data refinement processes and cost structures—or in legislation at broader (sub)industry levels to ease administration. A case could be made for standardization of this netback across countries, assuming that multinationals are collecting data of similar quality from users worldwide and centralizing their processing operations, implying a common cost structure.

Alongside the royalty base, the choice of the rate is also important. The netback deduction and royalty rate are two sides of the same coin, since a country can calibrate the effective tax burden by adjusting either the netback or the rate of the royalty. In the extractive industries, for example, some countries disregard costs altogether in defining their royalty base, instead valuing output at the benchmark price for its mineral content and lowering the royalty rate to compensate.

Rates should be set at an appropriately modest level, as they are applied to revenues rather than profits.⁴⁸ Where royalties are charged on a gross—rather than a net—basis, they no longer only tax pure economic rents and can distort the activities they are targeting.⁴⁹ This is analogous to the disincentive effects of royalties on exploration activity in the extractive industries. Thus, those countries with lower value consumer markets (e.g., with fewer active users or lower purchasing power) might set rates lower so as not to disincentivize data extraction by nonresident digitalized businesses or deter market entry altogether.

In setting the rate, policymakers should also be sensitive to the risks of excessive pass-through to consumers. While the location-specific nature of the asset means that companies cannot themselves relocate to avoid payment of the tax, their relative market power may allow them to pass the tax on to consumers. This is likely much easier than in the resource sectors where for most countries commodity prices in export markets are exogenously determined.

At the same time, this risk should not be overstated. If the marginal cost of providing the taxed service is low, then the royalty acts like a tax on the firm’s quasi-rents: rents that are exclusive of costs sunk in establishing the business (IMF, 2019). The primary impact may therefore not be on current pricing but on future investment. However, if the digital service is itself used as a business input, then the royalty may introduce production inefficiencies. This would reduce profitability for businesses and may raise prices for consumer, though the magnitude of such inefficiencies are unclear.

Cui and Hashimzade (2019) show that when the marginal cost is non-zero, the incidence of a tax on platform revenue will fall on both the platform and the advertisers/producers, but the effect on consumers is ambiguous. Moreover, as the authors argue, countries may well view some cost-passthrough to domestic users as a reasonable price to pay for capturing some of the platform rent. However, incidence effects are complex in two-sided markets. Firms may aim to shift some burden to the untaxed side: a tax on advertising creates an incentive to raise the price charged (or reduce the subsidy provided) to users; the price of advertising services may even fall. On tax incidence in two sided-markets, see Bourreau and others (2016) and Kind and others (2008 and 2010).

Businesses at different points in their life cycle might require different tax treatment. For this reason, a case could be made to introduce thresholds and rates that vary by the size of a business. Safe harbor rules could also be used to protect small or loss-making companies. However, such provisions might quickly become nonbinding even for small digitalized businesses given the nature of their cost structure (Annex I). With low (or falling) marginal costs, average variable costs will also be low, allowing small businesses to scale up operations profitably and quickly.⁵⁰

User-based royalties are a powerful fiscal instrument with a long history of application in a sector that shares a number of characteristics with data-intensive digitalized business models. However, they should be carefully approached so as not to generate other economic distortions. If royalties are intended as a substitute rather than a complement to the corporate income tax, then a lack of multilateral consensus risks unwanted international double taxation if they are not creditable against corporate income tax payable in home jurisdictions. If applied unilaterally such tax measures may also risk retaliation, especially if these measures end up being de facto targeted on firms from a few countries (Hufbauer and Lu, 2018).

The Implications of Moving Towards a Royalty on User-Based Revenues

In the absence of multilateral consensus, Box 1 explores—in a highly stylized setting—what the noncooperative outcome could be if all countries were to impose user-based royalties on the revenues from digital services. It illustrates that, ceteris paribus, countries will be indifferent (in terms of tax revenues) between a user-apportioned profit tax and a royalty on user-based revenues, if the ratio of the royalty rate and the profit tax rate applicable to the multinational enterprise is set in proportion to the company’s global average profit margin. This result holds irrespective of the global distribution of profits or revenues—and, by implication, the global distribution of costs.

While this is a very simplified result, which abstracts from some important real-world considerations and complexities, we can get a sense that should several countries move to implement such (short-term) royalties on user-based sales, then under the specified condition, the system would gradually come close to a system under which global profits are apportioned by the same user-based sales. This result should be familiar as it is the same logic that is commonly applied for calibrating standard turnover taxes for simplified small business regimes. In these systems, tax rates on turnover are typically set to ensure that the liability for the average small business would be the same as that under the application of the corporate income tax on profits.⁵¹

To take the logic one step further, given that user location and destination coincide, the widespread application of a royalty on user-based sales, when combined with certain forms of investment expensing can even lead us eventually to a destination-based cash-flow tax. Furthermore, user-based (or destination-based) approaches can be applied to any company engaged in digital activity, rather than only a subset of highly-digitalized companies.

As this section has discussed, there are several reasons why policymakers should not preclude the use of a user-based royalty tax instrument as a complementary measure or as part of a globally cooperative approach to tax data-intensive services. Firstly, royalty payments are common between private parties for the right to use intellectual property and other intangible assets, such as copyrights, patents, and trademarks. This sets a natural precedent for governments to charge similar payments for the use of a national (intangible) asset—such as a user data—by digitalized businesses. Secondly, as noted earlier, royalty taxes are often a necessary substitute when the direct taxation of rents is difficult, for example, in the resource sectors. Indeed, a royalty instrument is less vulnerable to profit shifting, particularly in a sector heavily reliant on hard-to-value intangibles. Thirdly, given the clear practical and political challenges of reforming income taxes to account for user value, well-designed user-based royalty instruments—which can also allow for some sort of netback to account for data processing—may provide a feasible and more realistic option. If designed and applied comprehensively and coherently to all services that use data as an input (i.e., little or no ringfencing around specific digital services), they could provide a valuable alternative or complement to the corporate income tax.

Some recent proposals for turnover taxes on digital services already take on the flavor of a user-based royalty. These include the digital services tax proposed by the European Commission, and variants subsequently proposed or enacted unilaterally by individual member states (Box 2). This formulation of digital services tax should be distinguished from non-user-based equalization levies on payments for digital advertising services. This type of levy was introduced by India in 2016 taking the form of a withholding tax on payments to nonresidents for certain digital services (Box 3). It has subsequently been expanded in scope in 2020 to a user-based turnover tax targeted only at nonresident businesses. Both types of tax attempt to target digital services provided by businesses operating remotely within a jurisdiction. They have also been formulated to sidestep income tax treaty issues by basing taxes on gross income (i.e., sales/turnover) rather than net income (i.e., profits)—for instance, as an excise tax on the supply of digital services or a tariff to protect domestic service providers.⁵²

^{Box 2.}

User-Based Turnover Tax: The European Commission Proposal

The 2018 European Commission draft directive is the most prominent example of a user-based turnover tax. It proposes a 3 percent levy on gross income of (resident and nonresident) businesses that supply specified digital services to domestic consumers and businesses. The tax is applied to a portion of revenues deemed to be derived from domestic users.

The tax is payable by large digitalized businesses, defined according to global and local thresholds. The thresholds are designed to ensure that only those dominant large businesses that benefit substantially from the exploitation of big data and network effects will be captured, while insulating small technology startups. Furthermore, the opportunity to engage in aggressive tax planning typically lies with larger companies. However, the global gross income thresholds inevitably restrict the companies liable to a handful of large tech firms and has sparked claims by the United States, where the majority of these companies reside, that the tax discriminates against foreign businesses. Large highly-digitalized multinational enterprises operating in other countries, such as China, with may also fall into the scope of this royalty (as they expand into European markets), leading to similar claims.

Only certain activities are included within the scope of the European Commission’s proposal (online advertising, intermediation services. and the sale of data), while other digital activities, such as online retail sales, financial and payment services, and businesses collecting and using data for their own internal purposes, are excluded. The rationale for this exclusion is that although online retail, digital content provision, and online services might also allow some degree of user interaction, such interaction is ancillary to the main purpose of the delivery of goods, content and services. That is, “value creation” in such cases lies mainly with the production of the goods, content, and services sold online, while the user’s role in value creation is less central. However, since many digital platforms engage in both in and out of scope activities (e.g., platforms that provide online retail, digital content provision and user intermediation services), the European Commission’s tax proposal would require the separation of revenue from taxable and non-taxable services.

In terms of revenue attribution, the European Commission proposal takes a volume-based approach, which circumvents the valuation issue. Revenue is allocated to member countries in proportion to the number of times an advertisement has appeared on users’ devices (in the case of advertising) and the number of users that have concluded transactions on a particular platform in a given tax period (for intermediation services), where the location of the user is determined based on their internet protocol or IP address. Questions remain as to how the residency of the user would determine the apportionment of revenues in the case of cross-border transactions, and how to apportion revenue derived from the sale of multinational data sets. At present, no ‘netback deduction’ for value added from post-data collection processing has been factored into the tax design, but may have factored in the choice of the tax rate.

Some EU states (e.g., Austria, Czech Republic, France, Italy, Spain, and the United Kingdom) have moved ahead with unilateral measures—though each have calibrated the tax differently, for example, with different rates and local (national) thresholds (see Table 2.1 for a list of those countries that have proposed/legislated/implemented digital services taxes). In many cases, the approach used to apportion revenues also deviates from the original European Commission proposal. For example, the United Kingdom’s approach is to account for user value through an indirect destination-based proxy: the value of the advertisements and search results targeted at domestic users and, in the case of online marketplaces, the commissions generated from transactions intermediated between domestic users (i.e., from both sides of the transaction).

^{Box Table 2.1.}

Examples of Country-Specific User-based Turnover Taxes

Sources: National authorities.

^1/DST originally expected to apply retroactively from Jan. 1, 2019. Final text of legislation included provision that for the 2019 tax year, the share of services connected with France will be assessed during the period between the day following the publication of the Act and Dec 31, 2019. However, following talks with the U.S., France will not collect revenue until end-2020.

^2/Italy introduced a 3 percent web tax law, effective as of January 1, 2020, established by the 2020 Budget Law (Law no. 160 of 2019) and 2019 Budget Law (Law no. 145 of 2018).

^3/Revenue allocation in proportion to (i) for social media platforms, revenues from targeting adverts at UK users, (ii) for search engines, revenues from displaying advertising against the result of key search terms inputted by UK users; (iii) for online marketplaces, commissions generated by facilitating a transaction between UK users.

^{Box Table 2.1.}

Examples of Country-Specific User-based Turnover Taxes

Country/Polity	Key dates	Details
European Commission	Mar 21, 2018; Draft directive	3 percent; dual thresholds: > EUR 750 million (global); > EUR 50 million (local, EU-wide); Online advertising, digital intermediation services, sale of user data; portion of annual worldwide revenues attributable to domestic users.53
Austria	Apr 5, 2019 (draft bill “Digitalsteuergesetz 2020”); to be implemented Jan 1, 2020	5 percent; dual thresholds: > EUR 750 million (global); > EUR 25 million (local); Online advertising (provision and contribution) only.
Czech Republic	Nov 19, 2019 (pending Parliamentary approval)	7 percent; dual thresholds: > EUR 750 million (global); > CZK 100 million (local, approx. EUR 2 million); Online advertising, digital intermediation services, sale of user data.
France 1/	Jan 25, 2019 (Draft Law No. 121/000039); Jul 11, 2019 (National assembly approval); Jul 11, 2019 (Senate approval)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 25 million (local); Online advertising, digital intermediation services, sale of user data.
Italy 2/	Jan 1, 2020 (Law no.145/2018; Law no. 160/2019)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 5.5 million (local); Online advertising, digital intermediation services, sale of user data.
Spain	Jan 25, 2019 (Draft Law No. 121/000039); Feb 18, 2020 (Parliamentary approval)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 3 million (local); Online advertising, digital intermediation services, sale of user data; Payments based on IP address of device used.
United Kingdom 3/	Apr 6, 2020 (Finance Bill 2019/20)	2 percent; 0 percent (loss-making companies); single threshold: >GBP 500 million (global). Safe harbor provisions for loss-makers and businesses with low margins. Annual GBP 25 million tax-free threshold; digital intermediation services (online marketplaces, search engines, and social media platforms).

Sources: National authorities.

^1/DST originally expected to apply retroactively from Jan. 1, 2019. Final text of legislation included provision that for the 2019 tax year, the share of services connected with France will be assessed during the period between the day following the publication of the Act and Dec 31, 2019. However, following talks with the U.S., France will not collect revenue until end-2020.

^2/Italy introduced a 3 percent web tax law, effective as of January 1, 2020, established by the 2020 Budget Law (Law no. 160 of 2019) and 2019 Budget Law (Law no. 145 of 2018).

^3/Revenue allocation in proportion to (i) for social media platforms, revenues from targeting adverts at UK users, (ii) for search engines, revenues from displaying advertising against the result of key search terms inputted by UK users; (iii) for online marketplaces, commissions generated by facilitating a transaction between UK users.

View Table

^{Box Table 2.1.}

Examples of Country-Specific User-based Turnover Taxes

Country/Polity	Key dates	Details
European Commission	Mar 21, 2018; Draft directive	3 percent; dual thresholds: > EUR 750 million (global); > EUR 50 million (local, EU-wide); Online advertising, digital intermediation services, sale of user data; portion of annual worldwide revenues attributable to domestic users.53
Austria	Apr 5, 2019 (draft bill “Digitalsteuergesetz 2020”); to be implemented Jan 1, 2020	5 percent; dual thresholds: > EUR 750 million (global); > EUR 25 million (local); Online advertising (provision and contribution) only.
Czech Republic	Nov 19, 2019 (pending Parliamentary approval)	7 percent; dual thresholds: > EUR 750 million (global); > CZK 100 million (local, approx. EUR 2 million); Online advertising, digital intermediation services, sale of user data.
France 1/	Jan 25, 2019 (Draft Law No. 121/000039); Jul 11, 2019 (National assembly approval); Jul 11, 2019 (Senate approval)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 25 million (local); Online advertising, digital intermediation services, sale of user data.
Italy 2/	Jan 1, 2020 (Law no.145/2018; Law no. 160/2019)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 5.5 million (local); Online advertising, digital intermediation services, sale of user data.
Spain	Jan 25, 2019 (Draft Law No. 121/000039); Feb 18, 2020 (Parliamentary approval)	3 percent; dual thresholds: > EUR 750 million (global); > EUR 3 million (local); Online advertising, digital intermediation services, sale of user data; Payments based on IP address of device used.
United Kingdom 3/	Apr 6, 2020 (Finance Bill 2019/20)	2 percent; 0 percent (loss-making companies); single threshold: >GBP 500 million (global). Safe harbor provisions for loss-makers and businesses with low margins. Annual GBP 25 million tax-free threshold; digital intermediation services (online marketplaces, search engines, and social media platforms).

Sources: National authorities.

^1/DST originally expected to apply retroactively from Jan. 1, 2019. Final text of legislation included provision that for the 2019 tax year, the share of services connected with France will be assessed during the period between the day following the publication of the Act and Dec 31, 2019. However, following talks with the U.S., France will not collect revenue until end-2020.

^2/Italy introduced a 3 percent web tax law, effective as of January 1, 2020, established by the 2020 Budget Law (Law no. 160 of 2019) and 2019 Budget Law (Law no. 145 of 2018).

^3/Revenue allocation in proportion to (i) for social media platforms, revenues from targeting adverts at UK users, (ii) for search engines, revenues from displaying advertising against the result of key search terms inputted by UK users; (iii) for online marketplaces, commissions generated by facilitating a transaction between UK users.

While it may be too early to evaluate these measures, so far, the revenue from such instruments that have been costed and included in national budgets appear to be relatively low, suggesting that the yields—as well as the perceived cost to highly-digitalized businesses—may be overstated (Table 2.2). As such, these measures may have some political appeal for remedying, without significant commercial impact, the popular perception that the government receives little tax from firms that feature so prominently in the daily lives of consumers (IMF, 2019).

^{Box Table 2.2.}

Revenue Estimates from Proposed Digital Services Taxes ^1/

Sources: National authorities.

^1/Estimates for the revenue yields vary from country to country given differences in design and structural characteristics of each market. For example, the Indian equalization levy, introduced in 2016, amounted to INR 7 billion in FY 2017/18 (<0.01 of GDP). The European Commission has estimated annual revenue yield from its digital service tax for member states of EUR 5 billion (<0.01 percent of EU GDP).

^{Box Table 2.2.}

Revenue Estimates from Proposed Digital Services Taxes ^1/

	2019	2020	2021	2022	2023
	(Local Currency Units, millions)
Austria	.	25	28	31	34
France	.	500	…	…	…
Italy	150	600	600	…	…
Spain	1,200	…	…	…	…
United	5	275	370	400	440
	(Percent of GDP)
Austria	.	0.006	0.007	0.007	0.007
France	.	0.020	…	…	…
Italy	0.008	0.033	0.033	…	…
Spain	0.096	…	…	…	…
United Kingdom	0.000	0.012	0.016	0.017	0.018

Sources: National authorities.

^1/Estimates for the revenue yields vary from country to country given differences in design and structural characteristics of each market. For example, the Indian equalization levy, introduced in 2016, amounted to INR 7 billion in FY 2017/18 (<0.01 of GDP). The European Commission has estimated annual revenue yield from its digital service tax for member states of EUR 5 billion (<0.01 percent of EU GDP).

View Table

^{Box Table 2.2.}

Revenue Estimates from Proposed Digital Services Taxes ^1/

	2019	2020	2021	2022	2023
	(Local Currency Units, millions)
Austria	.	25	28	31	34
France	.	500	…	…	…
Italy	150	600	600	…	…
Spain	1,200	…	…	…	…
United	5	275	370	400	440
	(Percent of GDP)
Austria	.	0.006	0.007	0.007	0.007
France	.	0.020	…	…	…
Italy	0.008	0.033	0.033	…	…
Spain	0.096	…	…	…	…
United Kingdom	0.000	0.012	0.016	0.017	0.018

Sources: National authorities.

^1/Estimates for the revenue yields vary from country to country given differences in design and structural characteristics of each market. For example, the Indian equalization levy, introduced in 2016, amounted to INR 7 billion in FY 2017/18 (<0.01 of GDP). The European Commission has estimated annual revenue yield from its digital service tax for member states of EUR 5 billion (<0.01 percent of EU GDP).

^{Box 3.}

Withholding Taxes on Payments to Nonresidents: India’s 2016 Equalization Levy

In 2016, India introduced an equalization levy in the form of a withholding tax on payments by domestic businesses to non-resident entities for online advertising services, at a rate of 6 percent. In March 2020, the levy was expanded to include income received by nonresident businesses for the supply of all e-commerce supply or services at a rate of 2 percent.

The first incarnation of the levy bears a close resemblance to the sorts of final withholding taxes already commonly applied to cross-border technical services (for example, accounting, management, and subcontractor services). For this version of the levy, the burden of compliance was placed on the domestic recipient of services.

Since payments to nonresidents for the provision of digital services abroad would otherwise go untaxed in the destination country, the levy attempted to equalize their (income tax) treatment vis-a-vis resident service providers. Indeed, to the extent that the levy reduces any tax-induced comparative advantage that foreign suppliers may enjoy over domestic suppliers, equalizing the tax treatment could also ease production inefficiencies.

Thus, the objective of these taxes is not to capture a portion of user value—while payable in the country of the purchaser of such services, the tax base may not in fact reflect the location of the users being targeted by such services (for example, in the case of an Indian retailer advertising products or services outside of India), and equally would not capture payments by the nonresident provider to companies in other (e.g., neighboring) countries, which are targeting Indian users.

Moreover, while simpler to design and administer, such taxes can potentially be more easily avoided by setting up appropriate structures. For example, if the resident company sets up an offshore related entity to make the payments to the nonresident supplier, it could be possible to avoid these withholding taxes.

Several other countries have also introduced variants of this tax. Thresholds have also been introduced to reduce the burden of this tax on small businesses. In Chile, Hungary, and Turkey, only payments for digital advertising service have been targeted. Tax rates also vary substantially from 6 percent in India to 15 percent in Turkey (see Box Table 3.1).

^{Box Table 3.1.}

Details of Other Turnover Taxes on Digital Activities by Country.

Sources: National authorities.

^1/Germany has a similar tax with proceeds earmarked for the promotion of national cinema.

^2/In August 2014, the Hungarian parliament passed a bill on a new type of tax on advertising published in Hungary. The tax is to be paid by media content providers settled in Hungary and includes online advertising activities. The tax is income based: a yearly income from advertising activity over HUF 100 million was initially to be taxed at the rate of 5.3 percent. The EC determined (Nov 4, 2016) that the rate structure violated rules on state aid because the lower rates were preferential to some companies.

^3/Constitutes an expanded scope of the existing nonresident income tax on Uruguayan-sourced income. Similar rules apply to the extended scope of the VAT.

^{Box Table 3.1.}

Details of Other Turnover Taxes on Digital Activities by Country.

Country		Details
France 1/	Jan 1, 2018 (Decree No. 2017–1364; Sep 20, 2017)	2 percent; (10 percent for certain content) on revenues from French residents from sales and rentals of video storage media, videos on demand, and advertising and sponsorship revenues derived by paid-for or free online video sites. Tax free allowance of EUR 100,000 for provider of free online access to Audiovisual (AV) content, 4 percent allowance for advertising revenues, 66 percent allowance for EU platforms sharing content created by private users, and exemption for sites where AV content is not primary business.
Hungary 2/	Jul 1, 2017	7.5 percent; applies to advertising revenues that exceed HUF 100 million.
Turkey	Dec 19, 2018 (Presidential Decision No. 476); Feb 15, 2019 (Communiqué No. 17)	15 percent; Payments made to nonresident online advertising service providers or to those who act as an intermediary regarding the provision of advertising services through the internet. (Article 94 of the Individual Income Tax and Article 15 of Corporate Income Tax).
Chile	Proposal issued Aug 23, 2018	10 percent; Nonresident companies. Withheld by payment intermediaries. Digital brokering, advertising, entertainment, intermediation and storage services. Revenues received from Chilean consumers (B2C).
India	Jun 1, 2016 (Finance Act 2016, Chapter VIII) Apr 1, 2020 (Finance Act 2020)	6 percent; Aggregate payments to nonresident >INR 100,000 (approx. US$ 1500) in a financial year. Nonresident companies. Withheld by payor. Online advertising purchased by Indian buyers. Revenues received from Indian residents or Indian PEs of nonresidents. 2 percent; Gross receipts in respect of goods sold/services provided to residents, non-residents and persons using IP addresses in India <INR 20 million. Charged and paid by nonresident e-commerce platform operators.
Uruguay 3/	Jan 1, 2018 (Law 19.535; Resolution 6,409/2018)	12 percent; In the case of intermediation services, the tax base will be 50 percent of the transaction if only one of the parties is based in Uruguay. Nonresident. AV services and digital mediation or intermediation services. Revenues received from Uruguayan residents (Uruguay-based IP address or user billing address).

Sources: National authorities.

^1/Germany has a similar tax with proceeds earmarked for the promotion of national cinema.

^2/In August 2014, the Hungarian parliament passed a bill on a new type of tax on advertising published in Hungary. The tax is to be paid by media content providers settled in Hungary and includes online advertising activities. The tax is income based: a yearly income from advertising activity over HUF 100 million was initially to be taxed at the rate of 5.3 percent. The EC determined (Nov 4, 2016) that the rate structure violated rules on state aid because the lower rates were preferential to some companies.

^3/Constitutes an expanded scope of the existing nonresident income tax on Uruguayan-sourced income. Similar rules apply to the extended scope of the VAT.

View Table

^{Box Table 3.1.}

Details of Other Turnover Taxes on Digital Activities by Country.

Country		Details
France 1/	Jan 1, 2018 (Decree No. 2017–1364; Sep 20, 2017)	2 percent; (10 percent for certain content) on revenues from French residents from sales and rentals of video storage media, videos on demand, and advertising and sponsorship revenues derived by paid-for or free online video sites. Tax free allowance of EUR 100,000 for provider of free online access to Audiovisual (AV) content, 4 percent allowance for advertising revenues, 66 percent allowance for EU platforms sharing content created by private users, and exemption for sites where AV content is not primary business.
Hungary 2/	Jul 1, 2017	7.5 percent; applies to advertising revenues that exceed HUF 100 million.
Turkey	Dec 19, 2018 (Presidential Decision No. 476); Feb 15, 2019 (Communiqué No. 17)	15 percent; Payments made to nonresident online advertising service providers or to those who act as an intermediary regarding the provision of advertising services through the internet. (Article 94 of the Individual Income Tax and Article 15 of Corporate Income Tax).
Chile	Proposal issued Aug 23, 2018	10 percent; Nonresident companies. Withheld by payment intermediaries. Digital brokering, advertising, entertainment, intermediation and storage services. Revenues received from Chilean consumers (B2C).
India	Jun 1, 2016 (Finance Act 2016, Chapter VIII) Apr 1, 2020 (Finance Act 2020)	6 percent; Aggregate payments to nonresident >INR 100,000 (approx. US$ 1500) in a financial year. Nonresident companies. Withheld by payor. Online advertising purchased by Indian buyers. Revenues received from Indian residents or Indian PEs of nonresidents. 2 percent; Gross receipts in respect of goods sold/services provided to residents, non-residents and persons using IP addresses in India <INR 20 million. Charged and paid by nonresident e-commerce platform operators.
Uruguay 3/	Jan 1, 2018 (Law 19.535; Resolution 6,409/2018)	12 percent; In the case of intermediation services, the tax base will be 50 percent of the transaction if only one of the parties is based in Uruguay. Nonresident. AV services and digital mediation or intermediation services. Revenues received from Uruguayan residents (Uruguay-based IP address or user billing address).

Sources: National authorities.

^1/Germany has a similar tax with proceeds earmarked for the promotion of national cinema.

^2/In August 2014, the Hungarian parliament passed a bill on a new type of tax on advertising published in Hungary. The tax is to be paid by media content providers settled in Hungary and includes online advertising activities. The tax is income based: a yearly income from advertising activity over HUF 100 million was initially to be taxed at the rate of 5.3 percent. The EC determined (Nov 4, 2016) that the rate structure violated rules on state aid because the lower rates were preferential to some companies.

^3/Constitutes an expanded scope of the existing nonresident income tax on Uruguayan-sourced income. Similar rules apply to the extended scope of the VAT.

As of March 2020, the second incarnation now displays the hallmarks of the user-based turnover tax from the EU proposal, but only payable by nonresident businesses, creating a discriminatory tariff on foreign-supplied digital services. By expanding the remit of the levy, the government aims to use the tax system to capture a portion of the value created from digital services within the Indian economy. The e-commerce supplies covered include the sale of online goods and services (including through platforms) to any person who uses an Indian internet protocol address and to any nonresident who is purchasing advertising services targeted at Indian residents. Unlike the original levy on advertising, the resident business is no longer responsible for deducting and remitting the tax. Instead, the e-commerce operator is responsible for charging and paying.

VAT and the “Free” Digital Barter

The identification of data collection as a taxable event raises the question of whether the ‘digital barter’—the transaction discussed in Section II through which data is collected—like regular barters should also be subject to VAT. In other words, should so-called “free” services (such as those provided by search engines or social media platforms), which amount to barter transactions—because users in fact “pay” for these services with their user data—be subject to VAT?

In the case of a business-to-business barter transaction which involves the supply of goods or services and the receipt of goods or services in payment, the two components of the transaction are typically considered as separate supplies on which VAT must be calculated, assuming both parties are registered for VAT. In the case of a cross-border barter where both parties are VAT-registered, no revenue would be expected to arise, particularly after the VAT-registered recipient claimed a credit for the VAT paid. However, if the VAT-registered parties were each in turn making exempt or partially exempt supplies (such that the VAT paid could not be fully reclaimed), then there is potentially revenue for one or both countries, as each would continue to be “destinations” for one of the supplies without a full corresponding credit. However, in the case of a business-to-consumer barter transaction—more common in the case of search and social media—consumers would not typically be VAT-registered, and as such only the digitalized business would have to pay VAT on the “valuable” service.

Moreover, there are questions over whether the business-to-consumer transaction constitutes a taxable barter in the first place. There is arguably no sufficiently direct link between the service provided and the non-monetary consideration received, given the universal access of the service or platform to users and the fact that the data collected varies from one user to another, with it even possible for a user to provide false data. Further, users are not attempting to use their data on a continued basis to derive income more generally—if given the opportunity to pay with traditional monetary means, they might agree or prefer to do so.

In any case, it would seem that applying VAT to “free” digital services raises the same complications as for the corporate income tax, in particular the need to value the non-monetary consideration (i.e. the user data) on a per-transaction basis. Attempting to evaluate both legs of the barter for the purposes of VAT seems unwieldy given the problems identified earlier of determining a value for data as well as the digital service provided—for taxing as well as crediting purposes—with both likely to fluctuate with demand and supply.