## wpiea2020076-print-pdf — I. INTRODUCTION and II. Understanding user-generated value ## Source details **Canonical URL:** [wpiea2020076-print-pdf — I. INTRODUCTION and II. Understanding user-generated value](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020076-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020076-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020076-print-pdf.pdf.json) --- ### Key findings and context (Introduction) - A new breed of “superstar” firm—businesses at the vanguard of the “digital revolution”—has emerged over the last decade, capitalizing on first mover advantages and network externalities to boost profitability and market dominance. - Policymakers are increasingly sensitive to perceived inadequacies of the current international corporate income tax system in generating sufficient tax revenues from large highly-digitalized multinational enterprises (see IMF, 2014, 2019). - The rapid growth of digital service providers has made them an attractive target for special taxes, which can also help mobilize revenues in the wake of a crisis. - For most sectors it is useful to consider a “digital asymptote”, with highly-digitalized businesses being closer to this asymptote than other businesses. ### Three central issues identified (Introduction) - Issue 1: Decoupling of economic and physical presence - Sophisticated ICT systems and the internet have facilitated remote cross-jurisdictional sales, straining the traditional permanent establishment concept that relies on fixed physical presence. - Example statistic: In the U.S., the share of e-commerce in retail sales has tripled to almost 10 percent compared to a decade ago. - Issue 2: Greater role of intangible assets - Modern multinationals derive a larger share of their value from intellectual property that is easy to shift across borders and hard to value for transfer pricing, frustrating the arm’s length principle and exacerbating profit-shifting opportunities. - Issue 3: User-generated value and its recognition - The online customer or “user” is now considered by many to be a critical driver of the value of digital services through the harvesting of user data at unprecedented scale. - Under the existing international tax framework, user participation is not recognized as a source of taxable value, raising questions about source, destination, taxable presence, and profit attribution. ### Observations on policy reactions (Introduction) - Policy proposals seek to limit tax avoidance and tax competition by pre-determining distributions of taxable profits; many proposals are predicated on the idea that users contribute to value creation and therefore justify source-based taxing rights where users are located. - Prominent proposals include pure formulary apportionment and hybrid residual profit allocation alternatives, often incorporating user value as an allocation key. - Some jurisdictions have implemented user-based turnover taxes targeting specific digitalized industries and activities in the interim. --- ### The User — interactions and the “digital barter” (Section II) - Definition: “user” = buyer or consumer of a good or service, physically or virtually. - Primary claimed channels of user-generated value: - Provision—or allowing for the collection—of personal data and creation of digital content that can be monetized by the business. - The role users play in building networks (network externalities) critical to many digital business models. - Examples of user interactions and characteristics of exchange: - When purchasing online, users reveal preferences recorded for future monetization; in two-sided markets both buyer and seller are users. - Users may use ‘free’ search engines or apps and share information in exchange for search and match services. - Users post content on social media (educational material, photos, videos, influencer promotion), trading data for the ability to share. - Users of internet-connected devices provide data via sensors (e.g., thermostats, vacuum cleaners, toothbrushes, cars). - “Digital barter”: users often receive “free” services but engage in multiple “micro-barter” transactions exchanging data for services; many users are unaware of the extent of passive data generation and monetization. - Debate on production of raw input data: - Some argue businesses “produce” raw input data via monitoring and thus users’ “mere acquiescence” should not be deemed co-creation of value. - Counterargument: distinction between active and passive provision is hard to draw and potentially irrelevant if both types of data are monetizable. ### Uses and markets for user data (Section II) - Uses of collected data: - Refine existing or develop new products and services; fuel machine learning and artificial intelligence. - Analyze behavioral trends for targeted advertising to own customers or third parties. - Customize and improve recommendation engines. - Sell raw or processed data to other businesses via data mining and brokerage firms. - Industry characteristics: - Consumer data is a strategic asset; firms invest heavily in collection and analysis to maintain competitive advantage. - Data brokerage industry harvests and packages consumer datasets for sale; some data mining can leave users uncompensated and unaware. - Core taxation question: - Are users adequately compensated in the digital barter? If not: (i) Should users be remunerated directly and taxed under personal income tax with corporate deduction? (ii) Or should governments tax a portion of digital firms’ profits on users’ behalf (establishing source-based taxing rights where users reside)? ### The Value of User Data — existence, attributes, and valuation challenges (Section II) - Core question: Does raw user data have inherent value prior to firm processing? - Some firms claim data is only valuable after processing (algorithms/intangibles); others argue without user data there would be no processed value. - Data attributes and determinants of value (from firm perspective): - Vintage, quality, and sensitivity. - User economic profile (affluence, propensity to spend, consumption habits); average value varies by country distribution and economic size. - Life events (e.g., marriage, childbearing) can increase data value via advertising opportunities. - Intensity of engagement: active users generate larger volumes and potentially more accurate profiles. - Aggregate size: value likely increases more than proportionately with volume collected due to network externalities and existing firm data stock. - Valuation difficulties and proxies: - Heterogeneous nature of data; little market trading of raw user data; value depends on firm’s capacity to extract useful information. - Proxies from M&A: Microsoft–LinkedIn (~US$ 260 per user), Facebook–Instagram (~US$ 20 per user), Facebook–WhatsApp (~US$ 42 per user). - Small-scale experiments (e.g., Datacoup, Meeco) provide limited user-value signals. - Advertising auctions may provide person-specific proxy values but blended data and cross-country mixes complicate attribution. - Analogy with extractive industries: - Data likened to “oil” as an asset extracted and processed; differences: data is nonrival and access (not ownership) is the appropriate concept. - Similarities useful for valuation methods, institutional frameworks, and taxing location-specific rents. ### Network externalities and policy implications (Section II) - Network externalities: - Platform value increases with number of users—direct and indirect effects in two-sided markets. - Digital amplification: digitalization exponentially increases scale and speed of network effects; user content attracts subscribers and further content. - Information externalities: user reviews influence other users’ choices and provide commercial data for firms. - Policy implications: - Network externalities exist in both digital and non-digital markets; not necessarily a justification for ring-fencing digital services for tax. - Some view network externalities as a source of location-specific rents. - Measuring user-generated network value is challenging; methods require measuring demand curves and producer surplus changes (Cui, 2018). - For taxation design the paper focuses on user data value rather than network externalities. ### Market concentration, regulation, and taxation trade-offs (Section II) - Market concentration and monopoly tendencies: - Network effects can lead to monopolies without interoperability and intervention; consolidation observed in internet search and social media. - Example: Facebook has over 2 billion users as of 2019, which is about 70 percent of the world’s population that has access to internet. - Internet access: about 40 percent of the world population has an internet connection today; in 1995, it was less than 1 percent. The number of internet users increased tenfold from 1999 to 2013; first billion reached in 2005, second in 2010, third in 2014. - Regulatory precedents: - European Commission fines cited: EUR 2.42 billion (2017) and EUR 4.34 billion (2018) against Google for different abuses. - Amazon’s distribution role raises “utility” and anti-competitive concerns (pricing below cost to eliminate competitors). - Policy choice considerations: - When markets tend to natural monopoly, regulation (interoperability mandates, price controls, sector-specific rules on data collection and privacy) is typically first-best rather than taxation. - Taxation may treat the symptom (redistributing rents) rather than the cause (market structure and anti-competitive behavior). - Appropriate policy responses likely vary by business model; a business-by-business or sector-by-sector approach may be necessary. - For universal or messaging/social media services, comprehensive regulation (akin to the U.S. 1996 Telecommunications Act) covering data collection, privacy, and ethical content may be needed. - Concluding policy questions: - If user-generated data creates location-specific rents, should taxing rights be allocated to the user’s country? - If so, how should value/profits be apportioned and taxed given valuation, attribution, and cross-border blending challenges? --- ### Nexus, profit allocation, and taxation instruments ### Users as dual agents and the compensation problem - Users act as both producers of valuable information (and generators of network externalities) and consumers of digital services. - Governments often view users as insufficiently compensated for inputs supplied to firms that provide services at near-zero marginal cost (Annex II). - The uncompensated value accrues to firms exploiting the information, boosting their rents, often beyond legal reach of governments. - Key statistic: E-commerce accounted for almost 10 percent of total U.S. retail sales in 2018, with over 80 percent carried out through online platforms. - Finding: For the world’s most populous emerging market economies, the average increase in internet usage in the 10 years since 2007 has been over 30 percent and represents approximately 1.4 billion new users. ### Designating taxing rights — production, ownership, and nexus - Property rights over personal data are largely undefined; ownership is implicitly vested in companies collecting data absent legal regimes. - Once collected, non-rival user data can be reused and sold; without rights or control, users cannot extract full returns. - Scenarios for ownership/control and compensation: - Company ownership analogy: user resembles an employee whose inputs (data) are owned by the firm; compensation could reflect present value of data used repeatedly. - User ownership: users retain ownership/control and businesses remunerate users per-use or over time; such remuneration would be deductible at corporate level and taxed under personal income tax schedules. - Government custody proxy: government holds custody of citizens’ data as a collective national asset and companies compensate the jurisdiction (e.g., via corporate tax or royalty). - Finding: Data portability and system interoperability are prerequisites for individual control and enforcement of personal data rights. - Finding: The microtransaction nature of data exchanges (“microbarters”) complicates individual valuation and compensation due to variable values and lack of transparent markets. ### Reappraising permanent establishment (PE) for user-based nexus - Traditional PE (fixed place of business or dependent agent authority to conclude contracts) typically excludes remote data collection and thus fails to grant source taxing rights for user-generated value. - Activities once considered auxiliary (local data collection, warehousing) now form core elements of digital business models and may justify expanded nexus. - Policy developments and examples: - OECD Action 7 and Article 12 of the Multilateral Instrument extend PE application to functions no longer merely auxiliary. - European Commission (Proposed) digital presence criteria (taxable year): >EUR 7 million in annual revenues; >100,000 users; >3000 business contracts for digital services. - India (April 2019): proposed significant economic presence test based on domestic transactions in respect of goods/services/data exceeding a threshold or systematic soliciting of domestic business activity/users; rule deferred to April 2021. - Israel (April 2016): introduced significant economic presence test listing five indicators including contracts with domestic customers, domestic user counts, localization, level of internet activity, and correlation between resident use and payments to foreign entities. - Nigeria (January 2020): Finance Bill introduced principle of significant economic presence for non-resident companies in digital services and e-commerce. - Slovakia (March 2018): expanded fixed place of business definition for foreign online platforms offering transport or accommodation. - Legal precedents and novel nexus concepts: - U.S. Supreme Court (June 2018, South Dakota v. Wayfair, Inc.): physical presence not required for state to compel out-of-state sellers to collect state sales tax; upheld nexus rules for yearly deliveries of more than US$ 100,000 or engaging in 200 or more separate transactions. - State-level “cookie nexus” and “in-state software” laws (e.g., Massachusetts; Ohio) treat cookies/apps/software as creating nexus under thresholds (example cookie-based collection requirement: annual sales of at least US$ 500,000 or 100 individual transactions). - Server presence rulings (India Authority for Advance Rulings, February 2012) and Model Convention commentary treat whole-server ownership/lease as PE in some cases. ### Approaches to establishing nexus — user-based, investment/asset-based, and marketing intangibles - User-based approaches: - Options include quantitative and qualitative benchmarks: number of registered users, active users, revenue in market, expenditure, local domain, dedicated local platform, local payment options. - Proposal: new PE nexus with user threshold, time threshold, and de minimis revenue threshold. - Concept: “sustained user relationship” as proxy for continuous data collection and exploitation; could serve as nexus basis or intangible asset for profit attribution. - Investment- or asset-based approaches: - Nexus arises if firm has invested capital in a market to access a customer base; requires qualitative and quantitative thresholds to avoid proliferation of small PEs. - Challenge: carving out general vs. market-specific investments (e.g., search functions vs. advertising sales). - Marketing intangibles: - Broadly defined assets (brands, goodwill, trademarks, user data) could be treated as location-specific for tax purposes. - Risk: may shift profits from R&D to populous end markets and lead to ringfencing of certain business types. - Valuation and distinction from other IP remain difficult and subjective. ### Allocating profits — from separate accounting to formulary systems - Arm’s length pricing (incumbent): - Attempts to regulate internal transactions among multinational affiliates; faces practical problems with unique intangibles, transfer mispricing, and conceptual issues in attributing worldwide income. - Unitary taxation/formulary apportionment (challenger): - Top-down division of worldwide income across countries using formulae based on measurable factors (e.g., capital, payroll, sales). - Formulary need not apply to all profits; can be limited to “residual” profits or a fraction attributable to user contributions. - User data as a factor in formulary apportionment: - Recommendation: expand formula to account for demand- and supply-side factors relevant to digitalization, and decide weights relative to sales, assets, payroll. - Candidate factors: - Demand-side: sales; population (total or by age); mobile phone and internet penetration. - Supply-side: assets (tangible, intangible); employees or payroll; users (volume, value of data). - Measurement and proxy issues: - Data factor proxies: population, mobile/internet access, number of active users, number of clicks, number of completed purchases. - Challenges: population is dynamic (seasonal tourism, migration); unclear definition of “active user”. - Distributional implications: - Inclusion of user-based factors would increase profit shares for large, populous economies sensitive to population/user measures (e.g., United States, China). - Demographic trends imply gradual profit reallocation to regions with growing younger populations in parts of Africa and Asia. ### Residual profit allocation and hybrid approaches - Concept: distinguish “routine” returns (arm’s length) from “non-routine” or residual profits; allocate residual profits by formula. - Routine return estimation methods: markup on costs, fixed minimum return on tangible assets, or similar to existing rules (e.g., Global Intangible Low-Taxed Income). - Application to digital context: - Allocate a portion of residual profits to value created by users; that portion is then apportioned among jurisdictions where users are located using agreed allocation factors. - Hybrid approach may be politically palatable and administratively feasible compared with full formulary consolidation. - Note: OECD’s unified approach under Pillar One follows residual profit allocation logic. --- ### Revenue instruments beyond corporate income tax: royalties, DSTs, VAT, and withholding ### Royalties on user-generated value — lessons and design (Section III) - Rationale (analogy with extractives): - Royalties useful where direct taxation of rents is difficult, assets hard-to-value, administrative capacity weak, and profit taxation vulnerable to cost-based profit shifting. - Royalties are simpler to administer than profit-based taxes because they rely on revenues. - Royalties sidestep permanent establishment/treaty challenges and secure revenue as soon as production (or data use) begins. - Design considerations: - Ad-valorem royalty on value of flow of data collected from a country in a year is ideal. - Periodic tax on stock of data held is an alternative if firms retain ownership. - Volume-based alternative: specific royalty per unit (e.g., per terabyte) if valuation unavailable. - Two-step revenue-approximation: apportion company final revenues by country users and allow a standardized netback deduction for post-data collection processing. - Rate-setting and incidence: - Netback deduction and royalty rate jointly determine effective tax burden; rates should be modest because royalties apply to revenues rather than profits. - Example numeric equivalence: “For example, a royalty of 4 percent is the equivalent to 20 percent corporate income tax rate assuming a 20 percent average profit margin across all businesses.” - Trade-offs and risks: - Gross-based royalties can distort production and are regressive relative to rent-based taxes. - Unilateral royalties risk double taxation absent home-country creditability and may invite retaliation. - Safeguards: modest rates, progressivity/thresholds by firm size, calibrated netbacks, and standardization to reduce cross-country inconsistency. ### Stylized equivalence result (Box summary) - If all countries impose user-based royalties on revenues from digital services, countries will be indifferent (in tax revenue terms) 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. - Implication: widespread adoption of user-based royalties calibrated to average profit margins would tend toward the revenue outcome under a user-based formulary apportionment of global profits. ### Digital services taxes, withholding and equalization levies — Box 1 and Box 3 country examples - European Commission (2018 draft directive) - Proposed 3 percent levy on gross income of businesses supplying specified digital services. - Dual thresholds: > EUR 750 million (global); > EUR 50 million (local, EU-wide). - Scope: Online advertising, digital intermediation services, sale of user data. - Revenue allocated to member countries in proportion to user interaction measures; user location determined by IP address. - Country examples of user-based turnover/DS taxes (rate; thresholds; scope) - 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 only. - Czech Republic (Nov 19, 2019; pending): 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 (Jan 25, 2019; Jul 11, 2019 approvals): 3 percent; dual thresholds: > EUR 750 million (global); > EUR 25 million (local); Online advertising, digital intermediation services, sale of user data. - Italy (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; 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 (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). - United Kingdom revenue allocation specifics: - (i) social media platforms: revenues from targeting adverts at UK users; - (ii) search engines: revenues from displaying advertising against key search terms inputted by UK users; - (iii) online marketplaces: commissions generated by facilitating a transaction between UK users. - Revenue estimates reported for proposed DSTs (Box Table 2.2 — Local Currency Units, millions; percent of GDP) - Austria: 2019: . ; 2020: 25; 2021: 28; 2022: 31; 2023: 34 - France: 2019: . ; 2020: 500; 2021: ...; 2022: ...; 2023: ... - Italy: 2019: 150; 2020: 600; 2021: 600; 2022: ...; 2023: ... - Spain: 2019: 1,200; 2020: ...; 2021: ...; 2022: ...; 2023: ... - United Kingdom: 2019: 5; 2020: 275; 2021: 370; 2022: 400; 2023: 440 - Box Table 2.2: Revenue Estimates (Percent of GDP) - Austria: 2019: . ; 2020: 0.006; 2021: 0.007; 2022: 0.007; 2023: 0.007 - France: 2019: . ; 2020: 0.020; 2021: ...; 2022: ...; 2023: ... - Italy: 2019: 0.008; 2020: 0.033; 2021: 0.033; 2022: ...; 2023: ... - Spain: 2019: 0.096; 2020: ...; 2021: ...; 2022: ...; 2023: ... - United Kingdom: 2019: 0.000; 2020: 0.012; 2021: 0.016; 2022: 0.017; 2023: 0.018 - Observed pattern: national budget yields appear relatively low; European Commission estimated annual revenue yield from its DST for member states of EUR 5 billion (<0.01 percent of EU GDP). - Withholding and equalization levies — India - India’s equalization levy (Jun 1, 2016; Finance Act 2016, Chapter VIII): 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. - India expansion (Apr 1, 2020; Finance Act 2020): 2 percent; Gross receipts in respect of goods sold/services provided to residents, non-residents and persons using IP addresses in India 1. - Specifics: - Increasing returns and network effects predispose technology sectors toward monopoly or tight oligopoly absent contestability. - Marginal product of data may decline after a certain level; marginal contribution of data differs by context (advertising vs. medical research). --- ### Policy synthesis — options, trade-offs, and recommended orientation - Two core technical challenges for taxing user-generated value: - Assigning taxing rights to data-producing (user/market) countries via revised nexus rules (e.g., new permanent establishment concept). - Designing revised profit allocation rules reflecting user contribution amid absence of market-based prices or benchmarks for user data valuation. - Policy options: - Formulary apportionment (with user as a factor) offers cleaner reduction of profit shifting but entails winners and losers and requires international agreement. - Intermediate approaches: marketing intangibles allocation and residual profit allocation aim to reallocate income to market jurisdictions while preserving some arm’s length elements. - User-based royalties/turnover taxes: feasible unilateral instrument, less vulnerable to profit shifting, administratively simpler, but risks distortion, regressivity, and double taxation unless coordinated. - VAT extensions and withholding/equalization levies: important complements for consumption and cross-border payment scenarios; VAT vendor-registration models recommended for nonresident suppliers. - Overarching recommendation: - Given user data’s broad role as an economic input across sectors, policy responses should avoid ringfencing and recognize user contribution across businesses; shifts in the international tax system to reflect their worth are likely inevitable and transformative. - Risks of uncoordinated measures: - Disincentives for investment; proliferation of double taxation; administrative complexity; misclassification risks; variation in effective tax rates affecting global allocation; political economy risks including retaliation. - International coordination: - Standardization of measurement, netback rules, thresholds, and allocation factors is crucial to limit cross-border inconsistencies and inefficiencies. *Source: IMF working paper (wpiea2020076-print-pdf) — Sections I, II, Box 1, Box 3, Annex I and Annex II (excerpt).* ### References ............................................................................................................. ### I. INTRODUCTION ### Key findings and context - A new breed of “superstar” firm—businesses at the vanguard of the “digital revolution”—has emerged over the last decade, capitalizing on first mover advantages and network externalities to boost profitability and market dominance. - Policymakers are increasingly sensitive to perceived inadequacies of the current international corporate income tax system in generating sufficient tax revenues from large highly-digitalized multinational enterprises (see IMF, 2014, 2019). - The rapid growth of digital service providers has made them an attractive target for special taxes, which can also help mobilize revenues in the wake of a crisis. - For most sectors it is useful to consider a “digital asymptote”, with highly-digitalized businesses being closer to this asymptote than other businesses. ### Three central issues identified - Issue 1: Decoupling of economic and physical presence - Sophisticated ICT systems and the internet have facilitated remote cross-jurisdictional sales, straining the traditional permanent establishment concept that relies on fixed physical presence. - Example statistic: In the U.S., the share of e-commerce in retail sales has tripled to almost 10 percent compared to a decade ago. - Issue 2: Greater role of intangible assets - Modern multinationals derive a larger share of their value from intellectual property that is easy to shift across borders and hard to value for transfer pricing, frustrating the arm’s length principle and exacerbating profit-shifting opportunities. - Issue 3: User-generated value and its recognition - The online customer or “user” is now considered by many to be a critical driver of the value of digital services through the harvesting of user data at unprecedented scale. - Under the existing international tax framework, user participation is not recognized as a source of taxable value, raising questions about source, destination, taxable presence, and profit attribution. ### Observations on policy reactions - A range of policy proposals aim to limit tax avoidance and tax competition by pre-determining distributions of taxable profits across countries; many proposals are predicated on the idea that users contribute to value creation and therefore justify source-based taxing rights where users are located. - Prominent proposals include pure formulary apportionment and hybrid residual profit allocation alternatives, often incorporating user value as an allocation key. - Some jurisdictions have implemented user-based turnover taxes targeting specific digitalized industries and activities in the interim. ### Challenges in recognizing and valuing user-generated value - Contentious acceptance: Businesses and governments must first agree that “user-generated value” exists; some claim user data has no value until processed. - Valuation difficulty: Objective bases for valuing user data and network externalities are tricky to establish due to lack of spot markets and heterogeneity of user characteristics. - Tax design implications: - How should user-generated value confer source-based taxing rights? - How should profits be allocated to reflect user value, e.g., via formula apportionment? - Are alternative fiscal instruments beyond corporate income tax appropriate to secure a jurisdiction’s “fair” share of user-based revenues? ### Related analytical material in the paper (selected) - Annexes I and II: Characteristics of highly-digitalized businesses and theoretical background for cost structures and strategic behavior. - Empirical and conceptual materials include Tables, Figures, and Boxes addressing country approaches, user-based turnover taxes, digital usage trends, and firm-level indicators (listed in the source). *Source: IMF working paper section titled "I. INTRODUCTION" (excerpt).* ### Section II explores the role of the user in digitalized business models. Section III then discusses ### Section II — Understanding user-generated value (from wpiea2020076-print-pdf) ### Overview - Focus: the exchange between users and digitalized businesses and whether users generate “value” for these businesses. - Two primary claimed channels of user-generated value: - Provision—or allowing for the collection—of personal data and creation of digital content that can be monetized by the business. - The role users play in building networks (network externalities) critical to many digital business models. - Structure: clarifies key terms and relationships, examines user interactions and data flows, evaluates valuation challenges, and considers policy implications for taxation and regulation. ### The User — interactions and the “digital barter” - Definition: “user” = buyer or consumer of a good or service, physically or virtually. - Examples of user interactions: - When purchasing online, users reveal preferences recorded for future monetization; in two-sided markets both buyer and seller are users. - Users may use ‘free’ search engines or apps and share information in exchange for search and match services. - Users post content on social media (educational material, photos, videos, influencer promotion), trading data for the ability to share. - Users of internet-connected devices (internet of things) provide data via sensors (e.g., thermostats, vacuum cleaners, toothbrushes, cars). - Characteristics of the exchange: - Increased exchanges → greater flow/stock of data about the user. - Data collected is often vast, real-time, and unstructured (web browsing activity, app usage). - “Digital barter”: users often receive “free” services but engage in multiple “micro-barter” transactions exchanging data for services. - Many users are unaware of the extent of passive data generation and monetization. - Passive data collection can occur with formal consent (terms of service) but without direct user involvement; tracking cookies and background location access are examples. - Debate on production of raw input data: - Some argue businesses “produce” raw input data via monitoring and thus users’ “mere acquiescence” should not be deemed co-creation of value. - Counterargument: distinction between active and passive provision is hard to draw and potentially irrelevant if both types of data are monetizable. ### The User Data — uses and markets - Uses of collected data: - Refine existing or develop new products and services; fuel data-intensive machine learning and artificial intelligence. - Analyze behavioral trends for targeted advertising to own customers or third parties. - Customize and improve search and matching functions (recommendation engines). - Sell raw or processed data to other businesses via data mining and brokerage firms. - Industry characteristics: - Consumer data is a strategic asset; firms invest heavily in collection and analysis to maintain competitive advantage. - Data brokerage industry harvests and packages consumer datasets for sale; some data mining can leave users uncompensated and unaware. - Central taxation question: - Are users adequately compensated in the digital barter? If not: - Should users be remunerated directly and taxed under personal income tax with corporate deduction? - Or should governments tax a portion of digital firms’ profits on users’ behalf (establishing source-based taxing rights where users reside)? ### The Value of User Data — existence, attributes, and valuation challenges - Core question: Does raw user data have inherent value prior to firm processing? - Some firms claim data is only valuable after processing (algorithms/intangibles); others argue without user data there would be no processed value. - Data is a nonrival but excludable “club” good; firms’ proprietorship over data allows market power and rent extraction. - Determinants of data value (from firm perspective): - Vintage, quality, and sensitivity (e.g., changes from security issues, litigation, legal regulations). - User economic profile (affluence, propensity to spend, consumption habits); average value varies by country distribution and economic size. - Life events (e.g., marriage, childbearing) can increase data value via advertising opportunities. - Intensity of engagement: active users generate larger volumes and potentially more accurate profiles. - Aggregate size: value likely increases more than proportionately with volume collected due to network externalities and existing firm data stock. - Valuation difficulties: - Heterogeneous nature of data; little market trading of raw user data; value depends on firm’s capacity to extract useful information. - From user perspective, value depends on privacy preferences and value placed on traded services. - Limited observable market benchmarks; some proxies from M&A: Microsoft–LinkedIn (~US$ 260 per user), Facebook–Instagram (~US$ 20 per user), Facebook–WhatsApp (~US$ 42 per user). - Small-scale user-centric and “pay for privacy” experiments (e.g., Datacoup, Meeco) provide initial user-value signals but are not widespread. - Revenue-based proxies face blending issues: data is processed, combined, and profits may be realized years after collection, making ex-post attribution to initial data inputs difficult. - Advertising may be an exception: person-specific ad placements have associated auction values and could serve as a proxy, though blended data and cross-country mixes remain complicating factors. - Analogy with extractive industries: - Data likened to “oil” in being an asset extracted and processed; differences: data is nonrival and access (not ownership) is more appropriate concept. - Similarities useful for valuation methods, institutional frameworks, and taxing location-specific rents. ### User value from network externalities - Network externalities: - Platform value increases with the number of users—direct and indirect network effects (two-sided market complementarities). - Examples: ride-sharing (more users increase pool size and service value), farmer’s markets, newspapers/TV (readers/advertisers). - Digital amplification: - Digitalization exponentially increases scale and speed of network effects; user content attracts subscribers and further content. - Information externalities: user reviews influence other users’ choices and provide commercial data for firms. - Policy implications: - Network externalities exist in both digital and non-digital markets; not necessarily a justification for ring-fencing digital services for tax. - Some view network externalities as a source of location-specific rents. - Measuring user-generated network value is challenging; methods require measuring demand curves and producer surplus changes (Cui, 2018). - For the purposes of designing taxation tied to user-generated value, the paper focuses on value from user data rather than network externalities. ### Taxation or Regulation? - Market concentration and monopoly tendencies: - Network effects can lead to monopolies in the absence of interoperability and intervention; rapid consolidation seen in internet search and social media firms. - Example: Facebook has over 2 billion users as of 2019, which is about 70 percent of the world’s population that has access to internet. - Internet access: about 40 percent of the world population has an internet connection today; in 1995, it was less than 1 percent. The number of internet users increased tenfold from 1999 to 2013; first billion reached in 2005, second in 2010, third in 2014. - Regulatory precedents and actions: - Antitrust and regulatory remedies can include breakups and price regulation; European Commission fines cited: EUR 2.42 billion (2017) and EUR 4.34 billion (2018) against Google for different abuses. - Amazon’s distribution role raises “utility” and anti-competitive concerns (pricing below cost to eliminate competitors), prompting discussion of price regulation. - Policy choice considerations: - When markets tend to natural monopoly, regulation (e.g., interoperability mandates, price controls, sector-specific rules on data collection and privacy) is typically first-best rather than taxation. - Taxation may treat the symptom (redistributing rents) rather than the cause (market structure and anti-competitive behavior). - Appropriate policy responses likely vary by business model; a business-by-business or sector-by-sector approach may be necessary. - For some universal or messaging/social media services, comprehensive regulation (akin to the U.S. 1996 Telecommunications Act) covering data collection, privacy, and ethical content may be needed. - Concluding policy questions raised: - If user-generated data creates location-specific rents, should taxing rights be allocated to the user’s country? - If so, how should value/profits be apportioned and taxed given valuation, attribution, and cross-border blending challenges? *Source: Section II (and transition into Section III) of the IMF working paper (wpiea2020076-print-pdf).* ### Section II established how the nature of digital service provision sees the user play a dual role in ### wpiea2020076-print-pdf - Section II established how the nature of digital service provision sees the user play a dual role in ### Users as dual agents and the compensation problem - Finding: Users act as both producers of valuable information (and generators of network externalities) and consumers of digital services. - Finding: Governments often view users as insufficiently compensated for inputs supplied to firms that provide services at near-zero marginal cost (Annex II). - Finding: The uncompensated value accrues to firms exploiting the information, boosting their rents, often beyond legal reach of governments. - Key statistic: E-commerce accounted for almost 10 percent of total U.S. retail sales in 2018, with over 80 percent carried out through online platforms. - Finding: For the world’s most populous emerging market economies, the average increase in internet usage in the 10 years since 2007 has been over 30 percent and represents approximately 1.4 billion new users. ### Designating taxing rights — production, ownership, and nexus - Finding: Establishing a right to tax income (or rents) from user-generated value requires disentangling relationships among user, digitalized business, and data. - Finding: Property rights over personal data are largely undefined; ownership is implicitly vested in companies collecting data absent legal regimes. - Finding: Once collected, non-rival user data can be reused and sold; without rights or control, users cannot extract full returns. - Scenarios for ownership/control and compensation: - Company ownership analogy: User resembles an employee whose inputs (data) are owned by the firm; compensation could reflect present value of data used repeatedly. - User ownership: Users retain ownership/control and businesses remunerate users per-use or over time; such remuneration would be deductible at corporate level and taxed under personal income tax schedules. - Government custody proxy: Government holds custody of citizens’ data as a collective national asset and companies compensate the jurisdiction (e.g., via corporate tax or royalty). - Finding: Data portability and system interoperability are prerequisites for individual control and enforcement of personal data rights. - Finding: The microtransaction nature of data exchanges (“microbarters”) complicates individual valuation and compensation due to variable values and lack of transparent markets. ### Reappraising permanent establishment (PE) for user-based nexus - Finding: Traditional PE (fixed place of business or dependent agent authority to conclude contracts) typically excludes remote data collection and thus fails to grant source taxing rights for user-generated value. - Finding: Activities once considered auxiliary (local data collection, warehousing) now form core elements of digital business models and may justify expanded nexus. - Policy developments and examples: - OECD Action 7 and Article 12 of the Multilateral Instrument extend PE application to functions no longer merely auxiliary. - European Commission (Proposed) digital presence criteria (taxable year): >EUR 7 million in annual revenues; >100,000 users; >3000 business contracts for digital services. - India (April 2019): proposed significant economic presence test based on domestic transactions in respect of goods/services/data exceeding a threshold or systematic soliciting of domestic business activity/users; rule deferred to April 2021. - Israel (April 2016): introduced significant economic presence test listing five indicators including contracts with domestic customers, domestic user counts, localization, level of internet activity, and correlation between resident use and payments to foreign entities. - Nigeria (January 2020): Finance Bill introduced principle of significant economic presence for non-resident companies in digital services and e-commerce. - Slovakia (March 2018): expanded fixed place of business definition for foreign online platforms offering transport or accommodation. - Legal precedents and novel nexus concepts: - U.S. Supreme Court (June 2018, South Dakota v. Wayfair, Inc.): physical presence not required for state to compel out-of-state sellers to collect state sales tax; upheld nexus rules for yearly deliveries of more than US$ 100,000 or engaging in 200 or more separate transactions. - State-level “cookie nexus” and “in-state software” laws (e.g., Massachusetts; Ohio) treat cookies/apps/software as creating nexus under thresholds (example cookie-based collection requirement: annual sales of at least US$ 500,000 or 100 individual transactions). - Server presence rulings (India Authority for Advance Rulings, February 2012) and Model Convention commentary treat whole-server ownership/lease as PE in some cases. ### Approaches to establishing nexus — user-based, investment/asset-based, and marketing intangibles - User-based approaches: - Options include quantitative and qualitative benchmarks: number of registered users, active users, revenue in market, expenditure, local domain, dedicated local platform, local payment options. - Proposal: new PE nexus with user threshold, time threshold, and de minimis revenue threshold. - Concept: “sustained user relationship” as proxy for continuous data collection and exploitation; could serve as nexus basis or intangible asset for profit attribution. - Investment- or asset-based approaches: - Nexus arises if firm has invested capital in a market to access a customer base; requires qualitative and quantitative thresholds to avoid proliferation of small PEs. - Challenge: carving out general vs. market-specific investments (e.g., search functions vs. advertising sales). - Marketing intangibles: - Broadly defined assets (brands, goodwill, trademarks, user data) could be treated as location-specific for tax purposes. - Risk: may shift profits from R&D to populous end markets and lead to ringfencing of certain business types. - Valuation and distinction from other IP remain difficult and subjective. ### Allocating profits — from separate accounting to formulary systems - Two broad approaches: - Arm’s length pricing (incumbent): attempts to regulate internal transactions among multinational affiliates; faces practical problems with unique intangibles, transfer mispricing, and conceptual issues in attributing worldwide income. - Unitary taxation/formulary apportionment (challenger): top-down division of worldwide income across countries using formulae based on measurable factors (e.g., capital, payroll, sales). - Critiques of separate accounting: - Hard to find comparables for unique intangibles; facilitation of profit diversion offshore; conceptual artificiality when income is generated by interactions across affiliates. - In digital context, pooling of data across countries increases complexity and transfer mispricing opportunities. - Strengths of formulary apportionment: - Ignores intercompany transactions and reduces profit-shifting; formulae can be sector-specific and adjusted for digital factors. - Formulary need not apply to all profits; can be limited to “residual” profits or a fraction attributable to user contributions. - Examples of domestic formulary use: U.S. states apply various three-factor or sales-only formulas for apportionment. ### User data as a factor in formulary apportionment - Recommendation: expand formula to account for demand- and supply-side factors relevant to digitalization, and decide weights relative to sales, assets, payroll. - Candidate factors: - Demand-side: sales; population (total or by age); mobile phone and internet penetration (proxy for usage). - Supply-side: assets (tangible, intangible); employees or payroll; users (volume, value of data). - Measurement and proxy issues: - Data factor proxies: population, mobile/internet access, number of active users, number of clicks, number of completed purchases. - Challenges: population is dynamic (seasonal tourism, migration); unclear definition of “active user” (no industry standard). - Example: Monthly active usage commonly computed by firms, but definitions vary (passive visitors vs. registered users). - Policy examples and proposals: - EU Common Consolidated Corporate Tax Base draft includes a fourth data factor reflecting collection/use of personal data of online platforms and services users. - India (April 2019) proposed assigning different weights to users depending on business user-intensity: 10 percent for low/medium user-intensity, 20 percent for high user-intensity (intensity remains undefined). - Distributional implications: - Inclusion of user-based factors would increase profit shares for large, populous economies sensitive to population/user measures (e.g., United States, China). - Other destination countries with large markets would also gain; small-source countries could lose. - Demographic trends (growing share of younger populations in parts of Africa and Asia) imply gradual profit reallocation to these regions under user-based factors. ### Residual profit allocation and hybrid approaches - Concept: distinguish “routine” returns (arm’s length) from “non-routine” or residual profits; allocate residual profits by formula. - Routine return estimation methods: markup on costs, fixed minimum return on tangible assets, or similar to existing rules (e.g., Global Intangible Low-Taxed Income). - Application to digital context: - Allocate a portion of residual profits to value created by users; that portion is then apportioned among jurisdictions where users are located using agreed allocation factors. - Hybrid approach may be politically palatable and administratively feasible compared with full formulary consolidation. - Note: OECD’s unified approach under Pillar One follows residual profit allocation logic. *Source: wpiea2020076-print-pdf (Section II and related passages).* ### Section II introduced the analogy between user data and natural resources. In this section, we ### wpiea2020076-print-pdf - Section II introduced the analogy between user data and natural resources. In this section, we ### Overview - Explores how revenue-raising instruments used in extractive industries could compensate countries for use of their citizens’ personal data. - Discusses need to change the VAT to include remote digital services provided by nonresident businesses and considers fiscal instruments that could substitute for or complement the corporate income tax to capture user value. - Shows that under certain conditions, widespread unilateral user-based tax measures would tend in the limit to the outcome under a user-based apportionment of global profits. ### Royalties on User-Generated Value — lessons from extractive industries - Historical analogy: extractive-sector fiscal regimes evolved from low initial bonuses/royalties to combinations of state participation, income taxes, ad-valorem royalties, and other instruments after decolonization and the “permanent sovereignty” principle. - Key objective in extractives: maximize compensation to host governments while maintaining investor incentives given respective risk profiles. - Regimes typically combine production and profit-based instruments to form a combined “resource charge,” balancing efficiency, neutrality, timing, and revenue risk. - Parallel drawn: personal data has largely undefined property rights and has gone uncompensated except indirectly via corporate tax; governments are now exerting greater sovereignty and seeking a fairer distribution of tax revenues. - Rationale for royalties: - Useful substitute when direct taxation of rents is difficult, especially where assets are hard-to-value, administrative capacity is weak, and profit taxation is vulnerable to cost-based profit shifting. - Simpler to administer than profit-based taxes because royalties rely on revenues and reduce need to monitor costs. - Sidestep legal challenges tied to permanent establishment rules and income tax treaty changes. - Royalties secure some revenue as soon as production (or data use) begins. - Design considerations: - Broad applicability desirable to avoid proliferation of multiple rates/bases across countries and activities. - Under simplifying conditions, consistent user-based royalties across countries are broadly equivalent in revenue terms to user-based formulary apportionment of profits. - Degree of progressivity may be desirable to capture large rents given variability in profitability across data-economy firms (see Annex I). - A minimum charge for all data collection activity could be desirable, but must be moderated to avoid deterring growth of data-intensive firms given high fixed and sunk costs. ### Designing a Royalty on User-Generated Value — bases, measurement, and rates - Royalty types and rationale: - Ad-valorem royalty on the value of the flow of data collected from a country in a given year is ideal (common in extractives). - Periodic tax on the stock of data held is an alternative if firms retain ownership; justified because companies repeatedly generate value from the stock (see Section III and footnote on stock reporting). - Both approaches require standardized valuation methods and/or establishment of a spot market for data. - Volume-based alternative: - If valuation methods are absent but volume (stock or flow) can be measured, charge a specific royalty per unit (e.g., pre-specified monetary charge per terabyte). - Requires periodic adjustment to reflect evolving data compression technologies. - Two-step revenue-approximation (when valuation/volume measurement lacking): 1. Estimate portion of a company’s final revenues from digital services generated by a particular country’s users (apportion user-generated revenues by location). 2. Allow for a (standardized) netback deduction from these country-specific revenues to account for post-data collection processing and analysis. - Netback could be defined company-by-company or standardized at broader (sub)industry levels; standardization plausible if multinationals collect similar-quality data and centralize processing. - Apportionment challenges: - Data is not traceable like physical commodities and is nonrival; the same data may be used simultaneously in production of multiple goods/services and over many years (see Section II and Section III for country apportionment characteristics such as population, purchasing power, and sales revenue). - Risk of inconsistent revenue-attribution methods across countries and activities, creating inefficiency and competing cross-border claims. - Rate-setting and incidence: - Netback deduction and royalty rate jointly determine effective tax burden; countries can adjust either to calibrate burden. - Some extractive regimes disregard costs in the base and lower royalty rates to compensate. - Rates should be modest because royalties are applied to revenues rather than profits. - Where royalties are charged on a gross basis they may not tax pure economic rents and can distort targeted activities. - Differential rates or thresholds may be appropriate for lower-value consumer markets to avoid deterring data extraction or market entry. - Risk of pass-through to consumers through higher prices exists; however, if marginal cost of providing the service is low, a royalty primarily taxes quasi-rents and may affect future investment more than current pricing. - In two-sided markets incidence is complex: a tax on advertising can incentivize platforms to raise prices or reduce subsidies to users while advertising prices may fall; tax incidence can fall on platform, advertisers/producers, or consumers depending on marginal cost (references to Cui and Hashimzade (2019), Bourreau and others (2016), Kind and others (2008 and 2010)). - Firm lifecycle considerations: - Different tax treatment, thresholds, or safe harbor rules could protect small or loss-making companies. - Such provisions may quickly become nonbinding given low/falling marginal costs that allow small businesses to scale profitably (see Annex I and Cui (2018)). - Example numeric equivalence: - “For example, a royalty of 4 percent is the equivalent to 20 percent corporate income tax rate assuming a 20 percent average profit margin across all businesses,” noting that effective income tax burden increases as profit margin declines and royalties may be too high for low-profit or loss-making businesses. ### Policy trade-offs, risks, and international considerations - Advantages: - Royalties secure early revenue, simpler administration, avoid treaty/permanent-establishment legal hurdles. - Can act as substitute where profit taxation is hard due to hard-to-value assets or weak administrative capacity. - Risks and downsides: - Gross-based royalties can distort production decisions and are regressive relative to rent-based taxes. - Potential excessive pass-through to consumers and production inefficiencies if digital services are business inputs. - Unilateral royalties as substitutes for corporate income tax could create international double taxation absent creditability in home jurisdictions. - Unilateral measures risk retaliation, particularly if de facto targeting occurs against firms from a few countries. - Implementation safeguards: - Set rates modestly, consider progressivity and thresholds by firm size, calibrate netback deductions, and consider standardization to reduce cross-country inconsistency. - Monitor and adjust base (value or volume) to technological changes (e.g., data compression). ### Implications of moving towards user-based royalties — noncooperative outcome and equivalence result - Stylized Box result summary: - If all countries impose user-based royalties on revenues from digital services, 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 indifference holds irrespective of the global distribution of profits or revenues and, by implication, the global distribution of costs. - Under these conditions, if several countries implement such royalties, the system would tend toward one under which global profits are apportioned by the same user-based sales. - This logic parallels standard turnover taxes calibrated for simplified small business regimes, where turnover tax rates are set to match average small-business liability under corporate income tax. *Source: wpiea2020076-print-pdf* ### Box 1. Digital User-based Royalties as a First Step on the Way to Formulary Apportionment ### Box 1. Digital User-based Royalties as a First Step on the Way to Formulary Apportionment ### Static two-country comparison: user-based royalty vs user-based formulary apportionment - User-based royalty tax revenue in country i: 휏_i^r R_i, where 휏_i^r is the royalty rate in country i applied to user-based revenues R_i earned by the nonresident highly-digitalized multinational enterprise for services provided to domestic users in country i. - User-based formulary apportionment tax revenue in country i: 휏_i^c α_i π_G, where - α_i = R_i / R_G is country i’s share of total global user-derived revenues, - ∑_{i∈{A,B}} α_i = 1, - R_G = ∑_{i∈{A,B}} R_i, - 휏_i^c is the corporate income tax rate for country i, - π_G = ∑_{i∈{A,B}} π_i is the multinational’s globally consolidated profits, - π_i = R_i − C_i is the profit generated in country i (but not necessarily booked there). - Revenue neutrality condition for each country i under the two systems: - 휏_i^c α_i π_G = 휏_i^r R_i. - Substituting α_i yields: 휏_i^r / 휏_i^c = π_G / R_G. - Key analytical implication: - For the two systems to deliver equal tax revenues for each country, the ratio of the two tax rates for each country i must equal the ratio of global profits to global revenues (the global average profit margin). - Therefore, assuming all countries introduce user-based royalties, a country can earn the same revenues as if all countries agreed on a user-based split of globally consolidated profits if and only if the ratio of the corporate income tax rate to the royalty rate is the same for all countries. ### Distributional and behavioral caveats - Turnover tax calibration issue: - “Just as in the small business setting, a turnover tax calibrated using average profit margins would be too high for low-profit or loss-making digitalized businesses (and vice versa), disincentivizing investment.” - Progression toward broader tax forms: - If user location and destination coincide, widespread application of a royalty on user-based sales, when combined with certain forms of investment expensing, can lead eventually to a destination-based cash-flow tax. - User-based (or destination-based) approaches can be applied to any company engaged in digital activity, not only a subset of highly-digitalized companies. ### Rationale for allowing user-based royalty instruments - Precedent and analogies: - Royalty payments are common between private parties for the right to use intellectual property and other intangible assets (copyrights, patents, trademarks); this sets a natural precedent for governments to charge similar payments for the use of a national intangible asset—such as user data—by digitalized businesses. - Practical tax policy advantages: - Royalty taxes are a necessary substitute when direct taxation of rents is difficult (e.g., resource sectors). - A royalty instrument is less vulnerable to profit shifting, particularly in sectors heavily reliant on hard-to-value intangibles. - Given practical and political challenges of reforming income taxes to account for user value, well-designed user-based royalty instruments—which can allow for some sort of netback to account for data processing—may be a feasible and 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. ### Recent policy proposals and unilateral measures - European Commission (2018 draft directive): - Proposes a 3 percent levy on gross income of (resident and nonresident) businesses that supply specified digital services to domestic consumers and businesses. - Dual thresholds: > EUR 750 million (global); > EUR 50 million (local, EU-wide). - Scope: Online advertising, digital intermediation services, sale of user data. - Revenue allocated to member countries in proportion to user interaction measures (e.g., number of times an advertisement has appeared on users’ devices; number of users concluding transactions), with user location determined by IP address. - No netback deduction for post-data collection processing is factored into current design. - Distinction from non-user-based equalization levies: - Non-user-based levies (e.g., India 2016 equalization levy) are withholding taxes on payments to nonresidents for certain digital services and differ from user-based turnover levies that aim to capture user value. ### Country-specific user-based turnover tax examples (design and thresholds) - 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. - 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 (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 (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 (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). - United Kingdom revenue allocation specifics: - (i) social media platforms: revenues from targeting adverts at UK users; - (ii) search engines: revenues from displaying advertising against key search terms inputted by UK users; - (iii) online marketplaces: commissions generated by facilitating a transaction between UK users. ### Revenue estimates reported for proposed digital services taxes - Box Table 2.2: Revenue Estimates from Proposed Digital Services Taxes (Local Currency Units, millions) - Austria: 2019: . ; 2020: 25; 2021: 28; 2022: 31; 2023: 34 - France: 2019: . ; 2020: 500; 2021: ...; 2022: ...; 2023: ... - Italy: 2019: 150; 2020: 600; 2021: 600; 2022: ...; 2023: ... - Spain: 2019: 1,200; 2020: ...; 2021: ...; 2022: ...; 2023: ... - United Kingdom: 2019: 5; 2020: 275; 2021: 370; 2022: 400; 2023: 440 - Box Table 2.2: Revenue Estimates (Percent of GDP) - Austria: 2019: . ; 2020: 0.006; 2021: 0.007; 2022: 0.007; 2023: 0.007 - France: 2019: . ; 2020: 0.020; 2021: ...; 2022: ...; 2023: ... - Italy: 2019: 0.008; 2020: 0.033; 2021: 0.033; 2022: ...; 2023: ... - Spain: 2019: 0.096; 2020: ...; 2021: ...; 2022: ...; 2023: ... - United Kingdom: 2019: 0.000; 2020: 0.012; 2021: 0.016; 2022: 0.017; 2023: 0.018 - Observed pattern: - So far, revenue estimates included in national budgets appear relatively low, suggesting yields—and perceived cost to highly-digitalized businesses—may be overstated. - The European Commission estimated annual revenue yield from its digital service tax for member states of EUR 5 billion (<0.01 percent of EU GDP). ### Withholding and equalization levies: India’s experience and variants - India’s equalization levy: - Introduced 2016 as 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, expanded to include income received by nonresident businesses for the supply of all e-commerce supply or services at a rate of 2 percent. - Design and objectives: - The original levy resembled final withholding taxes on cross-border technical services; compliance burden placed on the domestic recipient of services. - Objective: equalize income tax treatment of nonresident suppliers with resident suppliers where payments to nonresidents would otherwise go untaxed in the destination country. - Limitations: the tax base may not reflect the location of the users targeted by services (e.g., an Indian retailer advertising outside India); resident companies could avoid withholding by using offshore related entities. - Other country variants and rates (Box Table 3.1 details): - France (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 (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. - 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). *Source: wpiea2020076-print-pdf — Box 1. Digital User-based Royalties as a First Step on the Way to Formulary Apportionment* ### Box 3. Withholding Taxes on Payments to Nonresidents: India’s 2016 Equalization Levy ### Box 3. Withholding Taxes on Payments to Nonresidents: India’s 2016 Equalization Levy (concluded) ### Withholding and turnover taxes on digital activities — country specifics - 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 1 increases output by more than λ: - f(A, λK, λL) = λ f(A, K, L) < f(λA, λK, λL). - Cobb-Douglas example: f(A,K,L) = A^{α_A} K^{α_K} L^{α_L}, with α_i>0. Increasing returns to scale occur given that: - α_A + (α_K + α_L = 1) > 1. (text preserves original formatting) - Euler’s Theorem implication: when increasing returns hold, factors cannot be paid their marginal products from output; i.e., - f(A,K,L) < f_A A + f_K K + f_L L. - Growth literature responses: externalities or imperfect competition—both imply at least one factor paid less than its social marginal product and generally inefficient equilibrium (Jones, 2005). B. Cost functions and marginal cost behavior - Firm cost minimization problem: minimize p·z subject to f(z) ≥ y. First-order conditions for inputs i = 1,...,I−1: - p_i ≥ λ ∂f(z*)/∂z_i with equality if z_i* > 0 (1). - Lagrange multiplier λ equals marginal cost ∂c(p,y)/∂y. - Cobb-Douglas two-input example (rival z1 and non-rival z2; p = (p1,p2); f(z1,z2)=z1^{α1} z2^{α2}) yields cost function: - c(p,y) = θ p1^{α1/(α1+α2)} p2^{α2/(α1+α2)} y^{1/(α1+α2)} - where θ is a constant equal to: θ=(α1^{α2} α2^{α1})^{1/(α1+α2)} + (α1^{α2} α2^{α1})^{-α1/(α1+α2)}. (text preserves source expression) - Second derivative of cost with respect to y: - ∂^2 c(p,y)/∂y^2 = (1/(α1+α2))(1/(α1+α2) − 1) θ p1^{α1/(α1+α2)} p2^{α2/(α1+α2)} y^{1/(α1+α2) − 2} - Sign of ∂^2 c/∂y^2 depends on whether 1/(α1+α2) equals, is greater than, or is less than 1: - If α1 + α2 > 1 (increasing returns), then 1/(α1+α2) < 1 and marginal costs fall as output increases (second derivative negative). - Conclusion: non-rival technology that generates increasing returns implies falling marginal costs and average (variable) cost c(p,y)/y tends to decrease as y increases. C. Specifics for digitalized businesses - Digitalized businesses use non-rival inputs (intellectual property, user data, intangibles) and thus are subject to increasing returns to scale, analogous to Romer’s entrepreneurial innovators seeking profits from innovation. - Increasing returns can persist over a large output range, but in the short term, after very high levels of service production, constant or decreasing returns to scale may arise and average costs could become constant or rise (Varian, 2019). - Data collection specifics: - Marginal product of data could decline after a certain level of data collection; additional data may yield decreasing incremental value because existing data already gives a good user profile. - Marginal contribution of data differs by context: may decline quicker for advertising than for medical research, where large numbers of observations are particularly important. *Sources: IMF, World Economic Outlook; Thomson Reuters DataStream, and authors’ calculations.* --- _Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020076-print-pdf.pdf_