A survey asks what someone would do. A tax record shows what they did.
Both produce a number, and only one of them is measuring the thing this argument is about. That distinction looks obvious written down. It lasts about four seconds in a broadcast interview, which is why the same few figures keep arriving in this debate as evidence about migration when they measure something else.
What an answer costs
Someone asked whether they would leave the country if a tax were introduced pays nothing for saying yes. No accountant, no school place, no conversation with a spouse about their job, no decision about where to be ill in twenty years. The answer takes a moment, and it may buy something: irritation registered, a signal sent to somebody who will print it.
Leaving costs a great deal and buys none of that attention.
There is a second problem underneath the first. When a survey asks about a tax that does not exist and leaves its design unspecified, the respondent has to supply those details before answering: the threshold, the rate, whether their business is caught, what the rules would do about the house. Two people answering the same question are often pricing two different taxes, and neither is necessarily the one on the table.
A stated intention is not the same as an action. For this population and this tax, our register does not measure the gap, and we are not going to put a number on it.
This article quotes no survey finding, because our register holds none that has been verified. Figures of that kind are easy to find on this subject and hard to stand behind.
What surveys are actually good for
Sentiment is real information. If a substantial part of a wealthy population says it resents a tax, that is evidence of stated opposition to it. That is worth knowing, but it is not evidence of how many people will leave.
They are answers to a political question. The fiscal question is how many people actually go and how much of the tax base goes with them, and no amount of asking will produce it.
Three ways a migration number goes wrong
It projects and gets read as a count. The 16,500 figure in the British version of this argument, a net outflow of millionaires in 2025, is a forecast published by a firm that sells residence and citizenship advice. Locations were inferred partly from public professional profiles rather than from tax records. The hub article sets out the method in full.
It measures one thing and gets relabelled as another. Cutting the Swiss wealth tax by one percentage point raised reported taxable wealth by at least 43 per cent over six years. That figure circulates as though 43 per cent of rich people moved. The authors put only about 24 per cent of the response down to taxpayer mobility, and the mobility they found was between Swiss cantons, which involves no change of country or currency. The rest is people reporting and valuing their assets differently.
It has nobody to compare with. Wealthy people move in ordinary years for ordinary reasons. A number that counts departures after a tax change, without a comparison group who faced no change, cannot separate the policy from the background.
Four questions worth asking
Applied to any migration figure, in the order that saves the most time.
- Is it a count of what happened, or a projection of what someone expects?
- Counted from what? Tax records and administrative filings, or inferences drawn from public profiles and client enquiries?
- Compared with whom? Without a control group, the number describes a period in time and says nothing about the policy.
- Who published it, and what do they sell?
The third does most of the work and is skipped most often. Wealthy people are mobile in every year, including years when nothing changes, so a raw count of departures after a reform is a description of a period rather than a finding about a policy.
The fourth question is not an accusation. Commercial migration reports are produced by firms whose clients are the people already considering the move being counted, which is a selection problem and not a dishonesty one. The firm's enquiries and applications are among several inputs, so selection into its client base is relevant to how the forecast should be read. That does not make the figure false, but a reader is entitled to know how it was built.
What the alternative looks like
Britain has a properly measured answer to this question. After the 2017 non-dom reform, which cut the affected group's net-of-tax rate sharply, 4.9 per cent of them ceased UK tax residence over five years, established from HMRC records against a comparison group of people just below the threshold.
It estimates a different quantity from the figures that get repeated, and it took years and administrative data to produce. That is what the difference between a survey and a measurement looks like in practice.
The point
None of this makes the migration question go away, and none of it says that nobody leaves. It says that the numbers most often used to answer the question were built to answer a different one.
Ask what was counted, and by whom, and against what. Most of the figures in this debate stop being frightening once you know how they were built.
That is a duller argument. It is also one that can be won.
Sources and evidence status
Brülhart, Gruber, Krapf & Schmidheiny — Behavioral Responses to Wealth Taxes, AEJ: Economic Policy, 2022 (KB-056). Henley & Partners / New World Wealth — Private Wealth Migration Report 2025, methodology, published 24 June 2025 (KB-057). Advani, Burgherr & Summers, using HMRC administrative data (KB-052).
Tax Policy Associates published a critique of the Henley figures in July 2025. It is reported here as a critique and is not adopted as our own finding.
No verified survey of wealthy individuals' stated intentions is held in our Knowledge Base, and none is quoted in this article. The gap between what people say about leaving and what they do is not quantified here, because no claim behind this piece measures it.
0 comments