The Data Dividend

The problem

There is an industry that buys and sells the record of your life, and you have never seen a bill or a receipt from it. Estimates of the global data-broker market in 2025 cluster between $290 billion and $343 billion in annual revenue — the spread between reputable analysts is itself a finding, because nobody is required to report. Consumer data is the largest segment. Vermont was the first state to make brokers even register their existence, in 2018; California followed with a registry and a deletion mandate. Before that, the most reliable public fact about the industry was that it preferred not to be described.

The instinctive fix is to make data personal property and pay people for it. That instinct is popular, intuitive, and — we think — wrong. It fails on arithmetic and it fails on principle.

It fails on arithmetic because a single person's data is nearly worthless. Brokers pay fractions of a cent per record. If we perfected individual data markets tomorrow and every American negotiated flawlessly, the realistic annual result would be a few dollars, arriving as a stream of micropayments that cost more to administer than they deliver. It fails on principle because a market in personal data is a market in which the poor are induced to sell what the rich retain, and because the price would be set at the moment of maximum asymmetry — you, alone, versus a firm that knows exactly what your record completes.

The reason the industry is worth hundreds of billions is not that your record is valuable. It is that everyone's records, assembled, are valuable. The value is created by aggregation, and aggregation is a collective act performed on a population. Value that arises from the whole should be paid to the whole.

The proposal

Levy a royalty on the commercial exploitation of bulk personal and public data, and route it into the American Permanent Fund (№ 001) — where it is paid out as an equal dividend to everyone, because everyone made it.

Not a property right. Not a micropayment. Rent on a commons, collected once, at the point where the aggregation happens, from the firms that perform it.

How it would work

1. Register the aggregators. Any firm that buys, sells, brokers, or licenses personal data on more than a threshold number of individuals, or that trains a commercial model on bulk scraped corpora, registers federally — extending the Vermont and California registries to national scope. Registration is the precondition for everything else, and it is most of the reform: a great deal of this industry's margin is a shadow premium.

1. A royalty at the point of aggregation. A low single-digit percentage levy on revenue derived from bulk personal-data products and from commercial models trained on public corpora. It is assessed on the aggregator, not the individual, which means no consent theater, no clickwrap, and no negotiation any citizen can lose.

1. Deposit, don't spend. Receipts go to the Fund as principal. The dividend is equal per citizen because the input was equal per citizen — nobody's data is more of a commons than anybody else's.

1. Privacy rights stay separate and stay absolute. This proposal must never become a licence to trade in people. The royalty applies on top of deletion rights, purpose limits, and a hard prohibition on trafficking in certain categories — biometric, precise location, health, and the data of minors, which should not be commercially aggregated at any price. A tax is not permission. We would rather collect nothing from a category than legitimize it.

The numbers

Honest arithmetic, with the uncertainty shown. Take the midpoint of the market estimates — roughly $300 billion globally — and assume, conservatively, that about a third of that revenue is attributable to data about Americans. A 3 percent royalty on $100 billion yields about $3 billion a year.

Deposited as principal and spent at a 5 percent real return, that is $150 million in year one — under fifty cents per citizen. Compounding a $3 billion annual contribution for thirty years at 5 percent real builds a corpus near $200 billion, throwing off about $10 billion a year, or $30 per citizen per year, permanently.

We publish that number knowing it is small, because the alternative is to publish a flattering one. Anyone promising that a data dividend will pay your rent is selling something. What the Data Dividend does is establish a principle — that value extracted from a population belongs partly to that population — and attach a compounding asset to it. The principle is worth more than the first year's check, and if the AI economy grows the way its own builders forecast, the royalty base grows with it. This is the proposal in the catalog most likely to be worth ten times our estimate by 2050, which is also why we refuse to estimate it at ten times today.

The honest objections

"This is a tax on information, and it will be passed straight through to consumers." Some of it will be. The incidence of a royalty on an intermediated market is genuinely uncertain, and we will not pretend the whole burden lands on shareholders. Two things limit it: much of this industry's revenue is advertising margin rather than a priced consumer good, and the levy is small relative to the sector's growth rate. If the pass-through turns out to be near-total, the honest response is that the Fund still ends up owning an asset the public previously gave away for free.

"You are legitimizing surveillance by taxing it." This is the strongest objection and it deserves more than a rebuttal — it is the reason for the category prohibitions above, which we would keep even if they cost the Fund every dollar. The state taxes many things it also restricts. But we accept the burden of proof: if the royalty ever becomes an argument against banning a practice that should be banned, this proposal has failed and should be repealed. Write the sunset in.

"Defining 'bulk personal data' is a lawyer's paradise." True, and definitional fights will consume years. The registry threshold does most of the work — firms above it know exactly who they are — and existing state statutes have already survived the first round of line-drawing. Difficulty of definition is an argument for careful drafting, not for leaving the commons unpriced.

"Individuals should own their data. Your version pays them a pittance and calls it justice." We think individual ownership pays them a smaller pittance while stripping the protection of collective bargaining and calling that freedom. But we concede the emotional core: people want agency, not a dividend, and a royalty gives them no more control over the record of their lives than they have today. Control comes from № 016's companion rights — deletion, purpose limits, prohibition — not from the check. The check is only the rent.

"Models trained on public text aren't using 'your' data at all." Legally contested and not ours to settle. The claim here is narrower and does not depend on copyright: a commercial model trained on the collective written output of a civilization is capturing value from a commons, whoever holds the copyrights in it. That is a rent question, not an ownership question — the same logic that lets a nation charge for grazing on public land without asserting that it owns the grass.

Sources

• Data-broker market size estimates for 2025, $290B–$343B range: Grand View Research (grandviewresearch.com); market.us; Mordor Intelligence
• Vermont Act 171 (2018), first state data-broker registry; California Delete Act (SB 362, 2023) and the California Privacy Protection Agency broker registry (cppa.ca.gov)
• Federal Trade Commission, Data Brokers: A Call for Transparency and Accountability (ftc.gov)
• Proposal № 001, The American Dividend (the receiving institution)