What is actually reachable

The destination first, since the rest of this page is how we get there

Global personal wealth is $470 trillion and the top one per cent hold between 37 and 40 per cent of it (UBS and the World Inequality Database, 2026). The proposal's ambition is to bring that to roughly a fifth within a century. Run properly, with both instruments and a controller that adjusts as conditions change, that ambition is affordable:

Two things about that table are worth pausing on, because they are what make it a design rather than a wish. Nobody's wealth ever falls. The charge sits below the return on what it falls on, so levied holdings keep compounding — what declines is their share, because the rest of the economy grows faster. And the rate is not a novel imposition invented for the occasion.

---

How it works

Two instruments, one legible base

Two instruments, because one cannot do both jobs. The transaction layer is kept — it turns out to be indispensable — but for what it is uniquely good at rather than for the revenue it was written to raise.

---

Dynamic levers, in all phases

A rate that responds to the state instead of being set once

The design above solves for 1.80% because that is what reaches the target if the return on that stock is 4.5% for a hundred years. It will not be. A rate calibrated to one number fails in both directions when that number moves, and neither failure announces itself:

• Returns fall to 3% and a 1.80% charge is suddenly 60% of the return. Holdings barely grow, the survivability constraint is silently breached, and the coalition against it becomes universal.
• Returns rise to 6% and 1.80% no longer keeps pace with compounding. The target is missed quietly, over decades, with nothing in the design noticing.

So the rate becomes a controller rather than a constant, and the levers differ by phase — because nothing can be charged on stock until the rails make stock legible.

The test that matters

Whether a controller hits its target on the path it was tuned for proves nothing. The question is what happens on paths nobody planned for — so both designs are run against five return futures, including a decade-long depression.

What the phases cost

The staging is not free, and the model shows the bill rather than assuming it away. For the first two decades nothing is levied on stock — the rails do not yet make it legible — and the top percentile compounds untouched throughout. Its share rises before it falls.

---

Biannual review, in all phases

The controller as a governance process, not a daemon

A rate that moves continuously is a rate nobody can plan against, and the planning horizon of the people being charged is exactly what determines whether they restructure around it. So the rate is set at a review, held constant until the next one, and the review looks at what actually happened rather than at the model's expectation: revenue against forecast, the observed return, coverage, and the remaining gap.

All three levers are reviewed together, because adjusting one in isolation is how a review ends up raising a rate when it should have widened a base.

The adjustment log

Every change is recorded with what was seen and why it moved. That is the difference between a policy that can be audited and one that merely runs — and over a century, the audit trail is most of what makes the thing legitimate.

---

Why not simply tax transactions

The road that found the design

The obvious version of this idea is to skip the second instrument entirely: put a tiny fee on every transaction, let the sheer volume of global finance do the work, and abolish income tax on the proceeds. It descends honestly from Edgar Feige's Automated Payment Transaction tax, and the numbers behind it are real — global OTC foreign-exchange turnover runs at $9.6 trillion a day and interest-rate derivatives at another $7.9 trillion (BIS 2025 Triennial, verified).

The pGSTC adds five economic tiers and a programmable cascading factor to Feige's flat base, so a migrant's remittance is untouched while algorithmic trading is not, and verified supply-chain intermediates escape the cascade that makes turnover taxes force vertical integration. A research-backed expansion paper takes it further still, projecting the top 1% falling from roughly 45% of global wealth to about 20% by 2125 — the same destination the design above reaches by another route.

It is a serious idea, and this page took it seriously by computing it. Three findings followed, in the order they arrived, and the third is what produced the design above.

First: the revenue is not there

What happens when the base is allowed to move

A transaction tax is not paid out of profit. It is paid on top of the cost of transacting that already exists — the spread, the interchange, the wire fee. What governs behaviour is therefore the rate relative to that existing cost. At the proposed T5 rates, that ratio is roughly 626× for FX and under 2× for card retail: wildly different interventions wearing the same number, with the heaviest rates falling on the most mobile base.

A clean result falls out. At unit elasticity revenue approaches V·c and never exceeds it: the ceiling on a transaction tax is the base multiplied by the spread already on it, not by the notional volume. Foreign exchange carries fifty-four times retail's volume and one eighth its revenue ceiling — $2.4 quadrillion of turnover can yield about $97 billion, because a market trading on 0.4bp of friction cannot be made to carry 250bp.

---

Second: the transfer is the wrong shape

Why no rate closes the gap

The expansion paper connects its mechanism to its outcome without computing the transfer between them, and that is the load-bearing step. A levy is a flow: dollars per year. Wealth shares are a stock — $470tn already allocated (UBS, 2026) — compounding while you try to move it. Reaching the projected shares takes … transferred every year for a century.

Only the static row funds it, and the static row is the one already shown to be overstated by …. The projection reads as plausible because it is implicitly spending a number that does not survive the base being allowed to move.

---

Third: the history already said what works

And the paper cites it without drawing the conclusion

Every case where policy moved wealth shares at scale moved stock directly, and none was a transaction tax.

The pattern is not incidental. A charge on transactions taxes the movement of wealth, and large fortunes are characterised precisely by not moving. Buy-and-hold pays almost nothing; the working household transacting its whole income pays on all of it. The instrument is not merely too small for the target — pointed at concentration, it is aimed slightly the wrong way.

Which is what sent the design at the top of this page to a second instrument, and left the transaction layer doing the one thing nothing else can: making wealth visible enough to charge.

---

Why this one is collectable

The objection that killed every previous attempt

Wealth taxes do not usually fail on fairness or affordability. They fail because they are avoidable, and the record is not ambiguous about it.

---

The state is not a spectator

Correcting a one-sided model

Everything above gave taxpayers a response and gave the state none. Rates were fixed, rules were fixed, and only the base moved. That is not how any real tax works, and the asymmetry biases the whole result against the proposal: an unanswered route erodes forever, when in practice a revenue authority notices, legislates, and closes it.

Regulators demonstrably do this, and the evidence was retrieved rather than recalled. France and Italy's transaction taxes use an issuance principle — the charge follows the security's issuer, not the trading venue — so moving the trade to London does not escape a tax on a French share. Pillar Two's undertaxed profits rule exists specifically to tax what escaped somewhere else, and national transposition was still advancing through August 2026. CRS and DAC6 make arrangements reportable.

But the record splits, and the same sources say so. Relocation is closable and is being closed. Substitution is not: on deployed transaction taxes, migration to non-taxed derivatives remains largely unchanged. An issuance principle stops you trading the French share abroad; it does not stop you holding a swap on it instead, and nothing deployed does.

If you are creative enough to avoid, contribute that creativity

There is a lever that follows directly from the objection, and it already exists in law. DAC6 obliges the intermediaries who design a reportable arrangement to disclose it — the adviser, not only the taxpayer. The expertise that builds avoidance is conscripted into compliance rather than merely policed after the fact. What it changes is not the regulator's willingness but their information: without it the state learns of a route when revenue falls, which is years late and after the structure is entrenched.

Applied to the transaction levy it does almost nothing, for two opposite reasons that both hold. At calibrated rates there is nothing worth avoiding — which is the design's strength, not a weakness. At punitive rates the loss is cessation, and no reporting obligation revives a trade that no longer pays.

The reason is structural. Wealth does not stop existing when it is taxed. There is no cessation channel at all, so every point of erosion is avoidance — relocation, relabelling, contested valuation — which is exactly the category disclosure was built for. And the historical failures confirm it: France's ISF and Sweden's wealth tax were not abandoned because wealth ceased to exist. They were abandoned because it moved, while the state learned of each route only once revenue had already fallen.

---

The hundred years

Staging, and the discipline that makes it safe

The century-long horizon is the part of the original proposal that survives contact with the arithmetic best, because it never claimed the revenue arrives at once. A no-shortfall rule — no jurisdiction leaves legacy taxation until replacement revenue is demonstrated — is exactly the discipline that would have caught the static estimate before it became policy.

Note what changes in the sequencing. The transaction layer now comes first not because it funds anything, but because nothing else can be enforced until it exists. The stock charge is switched on only once the rails make holdings legible — which is also the point at which the political argument becomes winnable, because the alternative to a visible charge is an invisible one that only the unsophisticated pay.