Romer's Progressive Tax on Digital Advertising

Summary

In a May 2019 New York Times op-ed, "A Tax That Could Fix Big Tech," the Nobel-laureate economist Paul Romer proposed a progressive tax on the revenue that platforms earn from targeted digital advertising — the business model on which Google and Facebook are built.[1] He developed the mechanism further in a 2021 essay, "Taxing Digital Advertising," which includes an interactive tax-schedule calculator.[2] The proposal is aimed squarely at the platform and data rents of the ad-funded internet, and it is designed to do two things at once: raise revenue that cannot be shifted offshore, and create a standing incentive for the dominant platforms to shrink, split up, or abandon the surveillance-advertising model altogether. Daron Acemoglu and Simon Johnson took up a close variant in a 2024 MIT policy brief, "The Urgent Need to Tax Digital Advertising," recasting it as a "sin tax" on a socially harmful business model.[3]

The Proposal

Romer's design rests on a single incidence insight the wiki treats as central: revenue has a location; profit does not. A corporate income tax is defined as revenue minus cost, and because a multinational's costs are incurred in many jurisdictions, clever accounting can move reported profit to low-tax havens — which is why, Romer argues, corporate-tax receipts from tech firms have fallen. Advertising revenue, by contrast, is attached to a viewer with a known location and cannot be relocated by accounting:

"Because revenue has a location, a tax on revenue will not be subject to this problem."[2]

The tax therefore falls on gross ad revenue attributable to users in the taxing jurisdiction (Romer's worked example: if a firm collects \$100m for ads shown to U.S. users and 7% of those users live in New York, New York taxes the \$7m). Two design choices distinguish it from a plain revenue levy:

• A steeply progressive rate schedule. The marginal rate rises with a firm's ad revenue, so the largest platforms pay the highest average rate. Romer's rationale is explicitly anti-concentration — "If big is bad, tax big" — and the schedule is calibrated so a dominant firm could cut its tax bill by splitting in two, turning the tax into a standing pressure toward deconcentration.[2]
• A deliberate escape hatch. Romer wants the tax to be avoidable — but only by changing the business model. A platform that abandons targeted advertising for a subscription model escapes the tax: "This is a tax that we don't want tech firms to pay. We want them to switch to a more normal subscription model."[2]

Romer frames the underlying harm as much political as economic: targeted digital advertising gives a few firms unprecedented surveillance-based influence over public discourse. In his 2021 essay he reports that Facebook controlled 59% of all U.S. digital political-ad spending and Google another 18%, arguing the ad model "poses a growing threat to democracy."[2]

The Acemoglu–Johnson Variant (2024)

Acemoglu and Johnson's MIT brief reaches the same instrument from the harm side. They argue social media built on digital advertising damages mental health, amplifies extremism, and diverts innovation toward manipulation, and that the standard democratic tool for a harmful but legal product is an excise tax:

"the best way forward is to impose a flat tax on digital advertising revenue, set at a high rate above some low threshold of revenues."[3]

Their design differs from Romer's in one respect worth flagging: they prefer a flat rate above a threshold rather than Romer's continuously progressive schedule, while sharing his core choice to tax revenue, not profit, precisely because "it is too easy for these multinational companies to hide profits in low tax offshore jurisdictions."[3] Both are, in effect, Pigouvian taxes on the advertising business model as much as rent-capture instruments.

Relation to the Georgist Case

The ad tax sits on the contested frontier of the rent gradient, and it is a useful test of what a frontier rent instrument is really trying to do.

• It shares the DST's incidence logic but targets a narrower base. Like a digital services tax, Romer's tax is levied on revenue in a location — which is what makes it hard to avoid — and it therefore faces the same question the DST case raises: a revenue tax is not automatically a rent tax, and its incidence depends on how the platform re-optimises. Romer's answer is that he does not especially mind if the tax is passed through, because the point is to make the ad model unattractive relative to subscriptions; the revenue is secondary to the behavioural incentive. That reframes the instrument as closer to Pigouvian correction and antitrust than to clean Georgist rent capture.
• Is the base rent or quasi-rent? Ad revenue is not pure rent: it funds the zero-price services (search, social, maps) that deliver real consumer value, and part of platform advertising returns reflect genuine engineering and matching quality — the quasi-rent the wiki refuses to assume away (see data-rents and the Crouzet & Eberly intangibles counter). Taxing that base heavily runs into the innovation-incentive objection the wiki steelmans at taxing quasi-rents kills innovation: a tax calibrated to force firms out of the ad model is, by construction, not a light touch on the return to building the service. Romer's counter is that the targeted-advertising slice specifically is a low-value, harm-generating activity society should discourage — but that is a contestable empirical judgement, not the settled land case.
• Where it fits the menu. Among the capture-or-dissolve instruments the wiki catalogues for the digital frontier, Romer's tax is a dissolve-leaning device: its success metric is a smaller, less concentrated, less surveillance-dependent industry, not a stream of captured rent. It is a cousin of the interoperability and antitrust remedies in the Furman Review, reached by a fiscal route.

Nuances and Limits

• Untested. No jurisdiction has enacted Romer's progressive ad-revenue tax; the behavioural predictions (firms split up, or switch to subscriptions) are theoretical. The actual DST record (revenue taxes largely passed through — see DST incidence) is a caution about assuming the intended incidence will materialise.
• Not, in the author's own framing, primarily a rent tax. Romer motivates the proposal as a defence of democratic discourse against concentrated informational power, and Acemoglu & Johnson as a public-health "sin tax." Read as rent capture it is a secondary use of the instrument, and the wiki should present it as a frontier proposal that overlaps the rent question rather than a canonical rent-capture design.
• Contested harm premise. The case leans on empirical claims about advertising's effect on democracy and mental health (Acemoglu & Johnson cite the U.S. Surgeon General's 2023 advisory and recent behavioural research). Those literatures are active and disputed; the ad tax is only as strong as the harm premise it rests on.
• Source provenance. The 2019 New York Times op-ed is behind a paywall; its title and 6 May 2019 date are corroborated by NYU Stern's faculty page. The substantive design detail and quotations here are taken from Romer's own freely-hosted 2021 essay and FAQ and from the Acemoglu–Johnson MIT brief, all fetched this session.

See Also

• Korinek–Stiglitz: AI and income distribution — the AI-rents theory (non-reproducible factors absorb the gains; taxing them is non-distortionary) · DMA interoperability — the dissolve pole as legislated
• Platform and Data Rents — the rents the ad tax targets
• Digital Services Taxes and Their Incidence — the sibling revenue-tax instrument, and the incidence cautionary tale
• Furman Review — Unlocking Digital Competition — the dissolve-the-rent-through-competition alternative
• Pigouvian Taxation — the harm-correction logic Romer and Acemoglu & Johnson share
• Objection: Taxing quasi-rents kills innovation
• Radical Markets — the data-as-labor alternative for the same firms
• Geoism

Sources

1. Paul Romer (2019), "A Tax That Could Fix Big Tech," The New York Times, 6 May 2019 (Opinion). Title and date corroborated via NYU Stern faculty page (the NYT original is paywalled) — used for the origin and headline claim of the proposal (D-claim; op-ed).
2. Paul Romer (2021), "Taxing Digital Advertising," adtax.paulromer.net, with the companion "FAQs on Targeted Digital Ad Tax." adtax.paulromer.net · FAQs · code (GitHub) — used for the revenue-has-a-location incidence argument, the progressive schedule and "if big is bad, tax big" / split-up incentive, the subscription escape hatch, the 59%/18% political-ad-spending figures, and the verbatim quotes (author's self-published position; D-claims; fetched this session).
3. Daron Acemoglu & Simon Johnson (2024), "The Urgent Need to Tax Digital Advertising," MIT Shaping the Future of Work, Policy Brief, April 2024. MIT PDF — used for the flat-tax-above-a-threshold variant, the tax-revenue-not-profit rationale, and the "sin tax" harm framing (D/E-claims; policy brief by two prominent economists; fetched this session).