VoxEU Column Global governance Taxation

Profit-splitting rules and the taxation of multinational digital platforms: Comparing strategies

Tax avoidance by multinational firms presents a substantial challenge to policymakers and to international organisations. This column explores two possible policy regimes that could be introduced to target global firms focused on digital services: separate accounting and formula apportionment. The results of the study suggest that the separate accounting approach could be optimal, inducing lower efficiency costs and larger fiscal revenues. Such a policy regime would also make country-by-country reporting compulsory and reliable, which would induce additional outside benefits.

The current tax international rules, designed in the 1920s, aimed to reconcile two main objectives: avoiding double taxation and allowing countries to tax multinationals operating in their territories. In a digital world, these rules are not effective in controlling the behaviour of multinational enterprises, which can easily take advantage of differences in corporate tax rates (through what the OECD calls ‘base erosion’ and ‘profit shifting’). Profit shifting can be conducted via transfer prices on intangible assets, such as algorithms or licences (Choi et al. 2020). E-commerce allows a firm to avoid being taxed in a country by operating without a so-called Permanent Establishment. To restore fairness and public budgets, countries and international organisations are undertaking various initiatives.

Some countries have decided to unilaterally implement a digital service tax. France passed a law in July 2019 imposing a tax of 3% on the revenues based on intermediation and targeted advertising of 27 large digital platforms. Austria passed a similar law in October 2019 imposing a tax of 5% on online advertising. Other countries, including Turkey, Belgium, Spain, and the UK, have also declared their intention to introduce taxes on digital services (ranging from 2% in the UK to 7.5% in Turkey) in 2020. The proposed taxes, mostly based on activity and income indicators, are far from being efficient and, as of mid-2020, none of the proposed taxes have been implemented. But these proposals put pressure on digital platforms and international organisations. They triggered reactions by some platforms anticipating retaliation measures. For example, in 2016 Facebook decided to switch from a declaration of all its non-US profit in low-tax Ireland to a declaration of separate profit in all countries with sales offices (Lassmann et al. 2020). They also serve as a threat on international bodies to speed up agreements on reforms. 

Indeed, the OECD, the G20, and the EU have launched working programmes to adapt in a coordinated way some of the current rules of the digital economy. But progress is very slow.1 In particular, in June 2020 the US withdrew from the negotiations on the so-called Pillar One. Pillar One is focused on replacing physical presence by the virtual notion of ‘nexus’ and defining profit splitting rules. 

In a recent paper, we explore theoretically two different profit splitting rules which could emerge from current policy discussions: separate accounting (SA) and formula apportionment (FA). Each splitting rule splits the profit of a multinational firm into profits in each country where it operates. The country then can freely tax its allocated share. Our objective is to shed light on their effects on the policies of a monopolistic digital platform, which connect users who might be located in different jurisdictions. Users can be consumers, peers, firms, or advertisers. The platform generates network externalities that can either be positive or negative. Examples of positive externalities abound, ranging from peer-to-peer platforms like Spotify or collaborative platforms connecting users of similar type to platforms connecting buyers and sellers (like eBay, Airbnb, or Amazon marketplace), search engines and digital social media platforms (like Google or Facebook) connecting advertisers to consumers. Advertisers benefit from the presence of more consumers on the platform as this improves the quality of matching consumers to targeted ads, inducing positive network externalities. Externalities can also be negative. For example, consumers who dislike ads are harmed by the presence of advertisers on the platform. 

The two regimes of profit splitting depend on the availability of detailed data on the platform's profit. Under separate accounting, the platform (truthfully) declares its profit in every country. This corresponds to the objective of country-by-country profit reporting as set by the new rules proposed by the OECD. Under formula apportionment, fiscal authorities resort to an apportionment key, based on available data such as the number of users of the platform or the number of clicks to allocate profit across jurisdictions.2 Corporate tax rates – which apply to all firms in the countries not only digital platforms ¬– are taken to be exogenous and, for a first cut, we abstract away from tax competition across jurisdictions. We study how differences in corporate tax rates result in changes in the number of users and prices set by the platform in the different jurisdictions.

Our first result shows that, even in the absence of transfer pricing through intangible assets, the platform is able to shift profit away from the high-tax jurisdiction to the low-tax jurisdiction. The mechanism by which profit is shifted differs under each regime. Under separate accounting, the platform exploits the network externalities created by the platform, raising or reducing the number of users in one country to affect the demand and the profit in the other. Under formula apportionment, the platform manipulates the apportionment key to reduce the tax base in one country and increase it in the other. This manipulation can happen even in the absence of network externalities across jurisdictions. In both cases, an increase in the gap in corporate tax rates between the high- and low-tax jurisdictions reduces the platform’s overall pre-tax profit, increasing the fiscal revenues of the low-tax jurisdiction.

In a second set of results, we delve deeper into the effect of differences in corporate tax rates on the quantities and prices chosen by the monopolistic platform. This effect can be decomposed into a direct effect (assuming that quantities in the other jurisdiction remain fixed) and an indirect effect (taking into account the sequence of adjustments due to externalities across jurisdictions). Under separate accounting, the direction of the direct effect depends on the sign of network externalities. If externalities are positive, an increase in the corporate tax rate raises the number of users in the high-tax jurisdiction and reduces the number of users in the low-tax jurisdiction. If externalities are negative, the direct effect is reversed. Under formula apportionment, the sign of the direct effect is independent of externalities, and always results in a decrease in the number of users in the high-tax jurisdiction and an increase in the number of users in the low-tax jurisdiction, increasing the fraction of profit taxed in the low-tax jurisdiction. As a result, a platform with positive network externalities distorts the quantities in opposite ways under each regime, resulting in a decrease in the price in the high-tax jurisdiction under separate accounting but an increase under formula apportionment. 

Finally, we use numerical simulations to compare profits and tax revenues under the two regimes. For the entire range of tax rate values, distortions in the number of users are stronger under formula apportionment than under separate accounting. However, the tax bill is lower, resulting in comparable post-tax profits in the two regimes. Our analysis therefore favours the choice of separate accounting, both because it induces lower efficiency costs and larger fiscal revenues. Furthermore, separate accounting makes country-by-country reporting compulsory and reliable, which induces outside benefits.  However, it should be said that countries might disagree in the choice of the profit-splitting regime, as, in our simulations, the fiscal revenues of the low-tax jurisdiction are lower under separate accounting than under formula apportionment. 


Bloch, F and G Demange (2020), “Profit-splitting rules and the taxation of multinational Internet platforms”, CEPR Discussion Papers 15376.

Choi, J P, J Ishikawa and H Okoshi (2020), “Transfer pricing of intangible assets with the arm’s length principle”,, 27 July. 

Goolsbee, A and E L Maydew (2000), “Coveting thy neighbor’s manufacturing: the dilemma of state income apportionment”, Journal of Public Economics 75(1): 125-143.

Lassmann, A, F Liberini, A Russo, A Cuevas and R Cuevas (2020), “Taxation and Global Spillovers in the Digital Advertising Market. Theory and Evidence from Facebook", CesIfo working paper 8149.

OECD (2020), Tax Challenges Arising from Digitalisation – Report on Pillar One Blueprint: Inclusive Framework on BEPS, OECD/G20 Base Erosion and Profit Shifting Project, Paris: OECD Publishing.

Tørsløv, T R, L S Wier and G Zucman (2020), “Externalities in International Tax Enforcement: Theory and Evidence”,, 21 April. 


1 For example, the notion of Permanent Establishment is being redefined to account for digital presence (OECD 2020).

2 Formula apportionment is used in the US to allocate profits across states. The formula, which puts positive weight on sales, wages and assets, has been shown to induce distortions in the optimal input choices, both theoretically (Gordon and Wilson 1986) and empirically (e.g. Goolsbee and Maydew 2000).

1,155 Reads