Ask people what is wrong with taxing wealth above £10 million and fairness rarely comes up first. What comes up is arithmetic. So few people would pay it, the argument goes, that the whole exercise is theatre with a tax form attached.
That deserves a number rather than a rebuttal.
Why this objection is worth answering
Most arguments about wealth taxes are arguments about values, and they go round in circles because the people having them want different things out of a tax system.
This one can be settled. Either the money is there or it isn't. If a tax at this threshold raises a rounding error then the policy is a gesture, and it should be defended as a gesture or dropped. A campaign that cannot say what its own proposal would raise has no business asking anyone to take it seriously. So the figure matters more here than anywhere else in the argument. Which is exactly the reason to get it right instead of getting it big.
Headcount and tax base are different things
How many people pay, and how much comes in, are separate questions. The objection assumes the first answers the second.
It doesn't, and the reason has nothing to do with clever tax design. Twenty-two thousand people in the UK hold net wealth above £10 million, about 0.04 per cent of adults. The charge doesn't fall on their wealth. It falls on the slice above the threshold, and at the top of a wealth distribution that slice grows far faster than the headcount does. Someone worth £11 million pays on one million. Someone worth £3 billion pays on very nearly the lot. Add one more person at the very top and you add more to the base than a thousand more people at the threshold would.
The objection swaps a small population for a small tax base. That swap is the whole argument.
The mistake that runs the other way
There is a second confusion in this debate, and it inflates instead of deflating.
The Wealth Tax Commission's famous number is roughly £260 billion. That comes from a 5 per cent charge on net assets above £500,000, reaching more than 8 million people, payable in five annual instalments. It is a one-off. Quoted as an annual yield it is wrong by a factor most people would struggle to guess. Anyone reaching for £260 billion to show that wealth taxes raise enormous sums is making the mirror image of the mistake this piece is about.
What the modelling says
The most careful published estimate is by Arun Advani, Helen Hughson and Hannah Tarrant, in Fiscal Studies. They model an annual charge on individual net wealth using survey data linked to HMRC records, with the top of the distribution corrected using rich-list information. They publish two behavioural scenarios instead of one.
At 1.12 per cent on wealth above £10 million, the model returns £10.0 billion a year under low avoidance and £8.8 billion under high avoidance.
Both belong in any summary of it. £8.8 billion is the pessimistic case, where avoidance behaviour sits at the top of the range the literature supports, and the distance between the two is about a fifth.
For scale, that is 0.33 per cent of UK GDP in the low case and 0.29 per cent in the high one. Around a third of one per cent of national output, from a group you could seat in a football stadium.
Not transformative, then. Real money all the same.
A second estimate from a different direction
In July 2026, King's College London, with the Paris School of Economics and Berkeley, modelled a 2 per cent minimum effective tax on households worth over £100 million. They put the yield at about £10.4 billion a year from fewer than 1,000 households.
That proposal differs from ours on every axis: households instead of individuals, and a threshold ten times higher. The figures are not comparable.
What is informative is the order of magnitude. Two research teams, different data, different designs, both landing somewhere near ten billion a year at the top of the distribution. Convergence like that proves nothing on its own, though it does make the ballpark harder to wave away.
The number nobody quotes
The same table carries a figure that gets almost no attention and settles the efficiency objection more or less on its own.
Administering the tax at this threshold costs the government three thousandths of one per cent of what it raises. Roughly one pound of collection cost for every three thousand pounds collected.
That is what a high threshold buys. Twenty-two thousand people, most of whom already file complicated returns and keep an accountant on retainer, is a very different administrative proposition from the adult population. Anyone arguing that a wealth tax would cost more to run than it brings in should be asked which threshold they have in mind, because at this one the figures are not close.
Where this is weaker than it looks
Two things weaken this.
About half the revenue at this threshold comes from a few thousand rich-list individuals. That makes the estimate a projection about a small and unusually mobile group, not a statistical property of a population. If enough of them left or restructured their affairs the figure would move, by more than headcount intuition suggests: their share of the base far exceeds their share of the people. This is the genuine fragility in the number, and the objection almost never raises it.
The underlying data runs from 2016 to 2018. Asset prices have moved a great deal since, and the top of the distribution has moved most. The likely direction is that the model understates today's base, though that is an inference from asset prices rather than a finding of the model.
On the rate
The published figures use 1.12 per cent, which is not what we propose. The authors picked that rate to hit a £10 billion target. Arithmetic, not policy advice.
Our proposal is 2 per cent. Applying the model's own parameters and its own stated behavioural rule at that rate gives £16.7 billion to £12.8 billion a year. That is our calculation from their table, not a figure they published.
Two things follow. Revenue doesn't scale linearly with the rate, because the model's avoidance response is itself proportional to the rate; doubling the rate does not double the yield, whatever the back of the envelope says. And on the model's own structure the revenue-maximising rate sits around 2.9 per cent even under the pessimistic avoidance assumption, which puts our 2 per cent below the peak.
Mobility
The other half of the objection is that the base can move. People with £10 million can leave.
Fair worry.
Departure is a design variable though, not a fixed constraint. Exit charges on accrued gains, trailing liability for a period after leaving, source-based charges on domestic assets: all of these exist, all operate in other jurisdictions, and each changes the arithmetic of going. None is free. Each buys a reduction in one behaviour at the cost of a distortion somewhere else.
The narrower point is that the revenue figures above already carry a behavioural response. That is what the two scenarios are for. The open question is whether the assumed range is wide enough, and reasonable people disagree about that.
The size of the base is not a design variable. That is simply what the distribution looks like.
What it settles
A tax is judged on what it raises against what it costs to collect. On that test: eight to ten billion a year at the published rate, collected for a rounding error, from a group small enough that assessing them is tractable.
Reaching few people is a property of the threshold. Setting one was the entire point.
So this raises real money and does not fix the public finances. Anyone claiming either more or less than that is selling something, and the figures above are why we can say so.
Where to go next
The estimates here already contain an assumption about avoidance, and that assumption is where most of the remaining disagreement lives. The next piece takes it on directly: how much of a wealth tax survives contact with the people paying it.
Sources and evidence status
Figures from Advani, Hughson and Tarrant, 'Revenue and distributional modelling for a UK wealth tax', Fiscal Studies 42(3–4), 2021, Table 1; Wealth Tax Commission final report, December 2020; and King's College London, Paris School of Economics and UC Berkeley, July 2026. The 2 per cent range and the revenue-maximising rate are our own calculations from the Fiscal Studies table and its stated behavioural assumptions, not published figures.
0 comments