Digital Services Taxes as Actually Implemented: Design, Revenue, and the Stalled Pillar One Bargain

Summary

The wiki's DST incidence page covers the theory (Cui & Hashimzade's location-specific-rent rationale) and the sharpest single incidence estimate (Muddasani & Langenmayr's finding that Amazon passed roughly half the UK DST through to sellers and consumers) in depth. This page does not repeat that finding; it documents the complementary layer — what the taxes actually collected, who actually paid, and what happened to the international bargain that was supposed to retire them — across the UK, France, India, and (for contrast) Canada, using government revenue reports, parliamentary responses, and the OECD's own status statements.

The headline pattern: a digital services tax (DST) is a levy on gross revenue, not a rent tax, and every jurisdiction examined here shows the same two symptoms — modest, concentrated, and administratively uncertain revenue on one hand, and acute political fragility to great-power trade retaliation on the other. Both symptoms are exactly what a tax on a genuinely fixed, immobile rent (a site's location value) does not exhibit, which is why the rent gradient treats the DST record as a caution rather than a template.

The UK DST: Revenue That Ran Hot, Then Fell Short of Its Own Later Forecast

The UK's DST (2% on revenue from search engines, social media, and online marketplaces attributable to UK users; in force since April 2020) is the best-documented case, thanks to two UK government audits: the National Audit Office's 2022 investigation and HM Treasury's five-year review, published November 2025.[1][2]

Design. Groups pay only if their worldwide in-scope revenue exceeds £500 million and more than £25 million of that is derived from UK users; the first £25 million of UK revenue is itself exempt, and firms with a very low operating margin on in-scope activity pay a reduced rate.[1]

Revenue, year by year.

Source: NAO (2020-21 figure) and HM Treasury's Table 4.A (2021-22 through 2024-25).[1][2]

The tax outperformed its first forecast dramatically: HMRC's July 2019 pre- implementation estimate was £275 million for 2020-21, and actual receipts came in 30% higher. HMRC told the NAO the main reason was that it had built in a 20% discount for anticipated avoidance behaviour that never materialised, plus a small number of groups paying far more than expected.[1] Revenue is also sharply concentrated: of the £358 million collected in 2020-21 from 18 paying business groups, five groups paid 90% (£324 million) of the total — a handful of the largest platforms, not a broad tax base.[1] Implementation cost HMRC £6.3 million, about £1.5 million under budget.[1]

But the overshoot did not compound indefinitely. In March 2022, with the 2020-21 overshoot already known, HMRC and the OBR jointly forecast that cumulative DST revenue would exceed £3 billion by 2024-25.[1] Summing the five outturns HM Treasury itself later published — 358 + 380 + 576 + 678 + 808 — gives a cumulative total of about £2.80 billion, short of that £3 billion-plus revised forecast (this sum is the wiki's own arithmetic on the published year-by-year figures, not a number stated by either report). The tax that beat its first estimate by 30% undershot its second, more-informed one.

On incidence, the government's own five-year review does not deny pass-through — it declines to test it. HM Treasury's November 2025 review acknowledges that consultation respondents flagged "the potential pass through of the tax to other businesses and customers," and that "it would be likely for some businesses within the scope of the tax to pass on the costs to their customers in the form of higher prices."[2] Its response is definitional, not empirical: DST "is not a tax directly on consumers... but a tax on the revenue," so pricing decisions are "for each company to determine."[2] On market impact it states plainly that "HM Treasury does not hold data suggesting there is a significant economic impact" and that rising sector revenue "does not provide conclusive evidence about the direct impacts of DST" — i.e., the government has not itself measured incidence; the one rigorous estimate on that question is the external academic study covered on the DST incidence page.[2] The NAO's earlier report separately records that Amazon, Google, and Apple publicly stated they would pass the DST's cost to their customers.[1]

France's DST: A Forecast Missed Low, Then a Rate-Hike Fight That Failed

France's 3% DST (on digital-interface intermediation and targeted-advertising revenue; groups with global revenue over €750 million and French revenue over €25 million) took effect from January 2019, a few months before the UK's.[3][4]

Revenue undershot its own forecast in year one — the opposite of the UK's pattern. An ex-ante economic assessment prepared for the tech-industry trade group CCIA projected roughly €400 million in French DST revenue for 2019.[4] The French government's own response to a parliamentary question (published February 2023, sourced to the DGFiP) instead reports €277 million in 2019 (27 paying companies), rising to €375 million in 2020 (28 companies) and €474 million in 2021 (36 companies, of which 18 were French entities, 14 other European entities, and 4 non-European entities of the taxed groups).[3] Revenue growth is real but the first-year shortfall against the €400 million forecast is a data point the UK record does not share.

An industry-commissioned ex-ante incidence estimate exists, and should be read with that funding disclosed. The same CCIA-commissioned assessment — prepared by Julien Pellefigue of Taj/Deloitte, which states explicitly that "this report has been prepared independently by Taj/Deloitte and does not necessarily represent the views of CCIA or its members" — modeled a total economic burden of about €570 million against the roughly €400 million the tax was expected to raise, split roughly 55% onto consumers, 40% onto businesses using digital marketplaces, and only 5% onto the large internet companies nominally taxed.[4] This is a pre-implementation industry-funded prediction, not an ex-post empirical result — unlike Muddasani & Langenmayr's later, independent, ex-post estimate for the UK (covered on the DST incidence page) — but the two studies point the same direction: most of the burden lands away from the platforms. Amazon's own 2019 announcement that it would raise French marketplace seller fees by 3% to cover the tax is documented on that sibling page and not repeated here.

France then tried, twice, to raise the rate — and as of this session, failed both times. A November 2024 budget amendment proposed raising the rate to 5% from January 2025; a further National Assembly vote on 24 November 2025 would have raised it to 6% from January 2026; a Finance Committee proposal reportedly went as high as 15% with a higher revenue threshold. According to the Tax Foundation Europe's dedicated 2026 tracker, the National Assembly voted against the 2026 Finance Bill provision that would have raised the DST to 6%, so France's DST rate remains 3% as of this fetch.[5] Given how contested and fast-moving this budget fight has been — and that it is intertwined with the tariff-threat dynamic described below — this figure should be re-checked before being relied on for a future date.

India's Equalisation Levy: Built in Two Pieces, Then Unilaterally Dismantled

India ran the widest-scoped DST-family regime of the three, in two separate levies, and has now withdrawn both.

• The 6% levy (2016) applied to non-residents' online-advertising and digital ad-space services, with a permanent-establishment exemption and a de minimis threshold of ₹1 lakh in annual payments.[6]
• The 2% levy (2020) applied more broadly to e-commerce operators facilitating online sales of goods and services to Indian users, with a permanent-establishment exemption and a de minimis threshold of ₹2 crore in annual turnover — India's closest analogue to the UK/France online-marketplace DSTs.[6]

Both are now discontinued. The 2% e-commerce levy stopped applying to transactions from 1 August 2024; the 6% advertising levy was discontinued from 1 April 2025, after which "the provisions related to equalisation levy are not applicable."[6][7] The government's stated reason for the 2% levy's withdrawal was that the levy was "ambiguous" and created a "compliance burden" — Finance Minister Nirmala Sitharaman's public framing — compounded by the practical difficulty of tracking Indian IP addresses to establish which transactions were even in scope.[7] The withdrawal is also widely reported alongside India's broader alignment with the OECD's two-pillar process and the reduction of trade friction with the United States, whose Trade Representative had opened Section 301 investigations into the levy and threatened retaliatory measures; sources consulted here describe this alignment in general "two-pillar" terms without consistently specifying whether Pillar One or Pillar Two was the operative link, so the wiki does not adjudicate which pillar India's own reasoning cited.[7][8] Revenue effect: one tax-practice source estimates the 2% levy's discontinuation cost India "a short-term revenue loss of over ₹3,000 crore" (roughly US $360 million at mid-2020s exchange rates); the wiki could not independently verify a full time series of annual equalisation-levy collections against a primary CBDT dataset in this session, so only this withdrawal-year estimate is carried here.[7]

The Pillar One Bargain: Designed to Retire DSTs, Never Closed

Every DST examined above was explicitly framed by its own government as an interim measure pending a multilateral deal. That deal is the OECD/G20 Inclusive Framework's "Two-Pillar Solution," agreed in headline form by more than 140 jurisdictions in October 2021. Pillar One's "Amount A" is the component that reallocates a slice of the largest multinationals' profits to market jurisdictions — and its entire DST-facing purpose, in the Inclusive Framework's own words, is to provide "for the withdrawal and standstill of Digital Services Taxes (DSTs) and Relevant Similar Measures (RSM) with respect to all companies."[9] The bargain was explicit: sign the deal, and every signatory withdraws its DST and agrees not to reimpose one.

That bargain has not closed. The Inclusive Framework's Co-Chairs' most recent public status statement (13 January 2025) records that a text of the Multilateral Convention (MLC) to implement Amount A was released in October 2023 and revised through mid-2024; when the revised text was submitted for formal "adoption" in June 2024, only one member objected. But adoption of a text is not the same as signature, and the Co-Chairs state plainly that "agreement to adopt the final text does not create an obligation to sign it."[9] As of that January 2025 statement, the MLC had not opened for signature: negotiations remained stuck on a handful of outstanding technical issues within the separate "Amount B" framework (simplified transfer pricing for baseline marketing and distribution activities), which some members treat as an essential precondition for signing Amount A at all.[9] The interim DST standstill that was meant to bridge this gap — originally running through the end of 2023 — was extended once, to mid-2024, and then lapsed without a signed MLC to replace it.

The practical result, verifiable from the country cases above: DSTs remain in force in the UK, France, and (until 2024-25) India, none of them withdrawn as part of a closed Pillar One deal, because there is no closed Pillar One deal for any of them to be withdrawn under.

The Actual Unwind Has Been Unilateral and Coercive, Not Multilateral

Where DSTs have actually come off the books, the mechanism has not been the promised multilateral swap — it has been bilateral pressure from the United States, case by case.

Canada is the clearest instance. Canada's Digital Services Tax Act imposed a 3% levy on digital-services revenue attributable to Canadian users above CAD $20 million annually, applied retroactively to 2022, with the first payment due 30 June 2025.[10] On 27 June 2025, President Trump suspended all US-Canada trade negotiations over the tax, calling it "a direct and blatant attack on our Country."[10][11] Canada's Department of Finance announced on 29 June 2025 — one day before the first payment was due — that it would rescind the tax "in anticipation of a mutually beneficial comprehensive trade arrangement with the United States," and trade talks resumed toward a target date of 21 July 2025.[11] The entire cycle, from tariff threat to full unilateral repeal, took three days; there was no OECD deal, no multilateral withdrawal-and-standstill mechanism, and no compensating international agreement — only bilateral leverage.

That leverage escalated further in mid-2026. On 26 June 2026, President Trump threatened blanket retaliation against any country that imposes or expands a DST, stating: "any Country that imposes such a Tax will immediately be met with a 100% TARIFF on any and all Goods" sent to the United States.[12] Reporting names France, the UK, Italy, Spain, Turkey, and Canada as the DSTs in scope of the threat; notably, the separate EU-US trade agreement capping most tariffs on EU exports at 15% does not address digital services taxes, leaving them an open flashpoint even where a broader tariff truce is in place.[12]

Read together with France's failed rate-hike votes and India's cited compliance- and-friction rationale for withdrawal, the pattern across all three unwinding cases is the same: none of it is the coordinated multilateral retirement Pillar One was designed to deliver. It is unilateral retreat under direct bilateral threat from the government whose companies pay the tax — a form of political fragility with no analogue in the land-tax case, where no foreign government has ever threatened tariffs over a jurisdiction's property tax.

The Georgist Reading

The wiki's rent gradient treats land as the clean case — fixed supply, no incentive story to damage, incidence that cannot be shifted because the owner cannot move or shrink the site. A DST was, on its most sympathetic reading (Cui & Hashimzade's, developed on the DST incidence page), a deliberate attempt to reach the digital analogue of that fixed factor — a platform's location-specific rent from a country's users. The record assembled here shows why the attempt falls well short of the land case on both of the dimensions that matter:

• It is a revenue tax, not a rent tax, and behaves like one. None of the revenue figures above are net of the tax's own economic cost; the UK government's own review acknowledges pass-through is "likely" without ever measuring it, and the one government that did commission an incidence estimate (France, via industry-funded modeling) found only about 5% of the burden landing on the taxed platforms.[2][4] The land-tax contrast is exact: a site cannot relocate or shrink to avoid an LVT; a platform can and does re-optimise its fee schedule.
• The revenue is real but modest and concentrated. £808 million from the UK's DST in 2024-25 is a small fraction of the UK revenues of the five firms who pay 90% of it — this is a narrow, thin instrument even where it "works" as a revenue-raiser, not the broad rent base an ACE/DBCFT design would reach.
• Political fragility here is geopolitical, not merely domestic. Every rent- targeting instrument on the wiki's tech-rents comparison faces some form of political fragility (universal ACE repeal, for instance). But DSTs face a fragility land taxation structurally cannot: the taxed firms' home government retaliates with trade weapons against the taxing country, not just domestic lobbying against the tax. Canada capitulated in three days; France's hikes have twice failed; a blanket 100% tariff threat now hangs over any DST anywhere. A tax that a foreign superpower can extinguish with a tweet is not behaving like a tax on an immobile local asset.

None of this shows platform rents are uncapturable — it shows that this particular instrument, aimed at revenue rather than the rent itself, inherits both the incidence problem and a geopolitical fragility problem that a genuine location- specific rent base would not have to the same degree. That is the honest reading the rent gradient demands: the DST record is evidence about a badly -aimed instrument, not a verdict on whether platform rent exists or is capturable.

Honest Limits

• Independent, ex-post incidence evidence exists only for the UK (Amazon). The France figure carried here (55%/40%/5%) is a pre-implementation, industry- commissioned prediction, not a measured outcome; no comparably rigorous ex-post incidence study for France or India was found or fetched in this research pass. The reader should weight the UK finding far more heavily than the French estimate.
• India's full revenue history is not independently verified here. Only the ₹3,000-crore withdrawal-year revenue-loss estimate is sourced to a direct fetch; broader multi-year collection totals reported elsewhere were not confirmed against a primary Indian government dataset and are omitted rather than presented as fact.
• France's current DST rate is a fast-moving fact. As of this session's fetch (mid-2026), the rate remains 3% after a 6% increase was voted down on 24 November 2025, but the same budget process had already produced two prior rate-hike attempts within roughly a year. This figure should be re-verified before reuse.
• Whether India's withdrawal reflects Pillar One or Pillar Two alignment is unclear from the sources fetched, and the wiki does not resolve the discrepancy rather than guess.
• This page does not re-derive the UK incidence finding — see the DST incidence page for the Cui & Hashimzade rent-rationale and the Muddasani & Langenmayr pass-through estimate in full, including that study's own caveats (a 2025 working paper; final peer-reviewed figures should be re-checked).
• The Trump tariff threat (26 June 2026) is very recent news, not settled policy — it is a threat and a live political dynamic, not yet a concluded trade action, and its ultimate effect on any of the DSTs discussed here is unknown at the time of writing.

See Also

• Digital Services Taxes and Their Incidence — the rent-rationale theory and the UK's rigorous ex-post incidence estimate (Amazon/Muddasani & Langenmayr), not repeated on this page
• Taxing Tech Rents — Instrument Comparison — where the DST is graded against ACE/DBCFT, Romer's ad tax, data dividends, and antitrust/DMA dissolution
• Platform and Data Rents — the underlying diagnosis a DST tries to reach
• Resource Rents — the royalty template Cui & Hashimzade's rent-rationale invokes
• Land Value Tax — the fixed-factor contrast: why a rent tax on an immobile site cannot be shifted the way a DST is
• Romer's digital advertising tax — a rival revenue-based instrument aimed at the same firms
• Digital Advertising Taxes After Romer: Maryland's Real-World Test — the US ad-tax parallel to this page's DST record: legal fragility and under-projection revenue
• Geoism — the rent-domain program and the gradient rule this page's Georgist reading applies

Sources

1. National Audit Office (2022), Investigation into the Digital Services Tax, HC
2. NAO PDF — used for the 2020-21 revenue figure (£358m, 30% above the July 2019 forecast of £275m), the five-groups/90% concentration figure, the £6.3m implementation cost, the March 2022 cumulative-to-2024-25 forecast (>£3bn), and Amazon/Google/Apple's public statements that they would pass the DST's cost to customers (A/B-claims; fetched and read via PDF text extraction this session).
3. HM Treasury (November 2025), Digital Services Tax Review Report, presented to Parliament under the Finance Act 2020. GOV.UK PDF — used for the 2021-22 through 2024-25 revenue table, the government's own statements on pass-through risk and the absence of measured market-impact data, and confirmation of the £6.3m implementation cost (A/B-claims; fetched and read via PDF text extraction this session).
4. Assemblée nationale (question n°2974, réponse publiée 14 February 2023, sourced to the DGFiP), "Rendement de la taxe sur les services numériques." Assemblée nationale — used for France's DST revenue by year (2019: €277m/27 payers; 2020: €375m/28 payers; 2021: €474m/36 payers, with the French/European/extra-European payer breakdown for 2021) (A-claim; fetched this session).
5. Julien Pellefigue (Taj/Deloitte, commissioned by the Computer & Communications Industry Association) (22 March 2019), The French Digital Service Tax: An Economic Impact Assessment. Deloitte PDF — used for the €400m 2019 revenue forecast, the ~€570m modeled total economic burden, and the ~55%/40%/5% (consumers/business-users/platforms) ex-ante incidence split; industry-commissioned, flagged as such (D-claim, attributed estimate, not adopted as fact; fetched and read via PDF text extraction this session).
6. Tax Foundation Europe (2026), "Digital Services Taxes in Europe" tracker. Tax Foundation — used for the current country-by-country DST rate table and the confirmation that the National Assembly voted against France's proposed 6% rate increase on 24 November 2025, leaving the rate at 3% (A-claim; fetched this session).
7. Income Tax Department, Government of India, "Equalisation Levy" (official guidance page). incometaxindia.gov.in — used for the design of both levies (6% online advertising, 2016; 2% e-commerce, 2020), their thresholds and exemptions, and confirmation that equalisation-levy provisions ceased to apply from 1 April 2025 (A/F-claims; fetched this session).
8. TaxGuru, "Equalisation Levy: Journey From Inception to Abolition & New Tax Regime." TaxGuru — used for the exact withdrawal dates (2% levy: 1 August 2024; 6% levy: 1 April 2025), the government's stated "ambiguous"/compliance-burden rationale, the IP-address tracking compliance problem, and the ₹3,000-crore short-term revenue-loss estimate (A/B-claims; practitioner source; fetched this session).
9. India Briefing (2024), "India to End 2% Equalisation Levy on Foreign Digital Companies." India Briefing — used for the US Section 301/trade-friction context and the general OECD two-pillar alignment framing (fetched this session; the wiki does not treat this source's Pillar One/Two attribution as resolved — see Honest Limits).
10. OECD/G20 Inclusive Framework on BEPS, Co-Chairs (13 January 2025), "Pillar One Update from the Co-Chairs of the Inclusive Framework on BEPS." OECD PDF — used for the Amount A withdrawal-and-standstill design language (verbatim quote), the October 2021/October 2023/June 2024 negotiation timeline, the "adoption is not signature" distinction, and the outstanding Amount B blockers (A/C-claims; primary institutional statement; fetched and read via PDF text extraction this session).
11. Al Jazeera (30 June 2025), "Canada rescinds digital services tax after Trump suspends trade talks." Al Jazeera — used for the design of Canada's Digital Services Tax Act (3%, CAD $20m threshold, retroactive to 2022) and the 27-30 June 2025 timeline (A-claim; journalistic source; fetched this session).
12. Government of Canada, Department of Finance (29 June 2025), "Canada rescinds Digital Services Tax to advance broader trade negotiations with the United States" (news release). Canada.ca — used for the primary-source rescission announcement and its verbatim rationale (A-claim; primary government source; fetched this session).
13. CBS News (26 June 2026), "Trump vows immediate 100% tariff if countries levy digital services tax." CBS News — used for the verbatim tariff-threat quote, the countries named, and the note that the EU-US tariff-cap agreement does not address DSTs (A-claim; journalistic source reporting a direct quote; fetched this session; flagged in Honest Limits as very recent, unresolved news).