Political costings often look authoritative because they contain large, precise numbers.
“£42 billion saved.”
“£90 billion of tax cuts.”
“£50 billion of waste removed.”
The number may be correct. But precision on the page is not the same thing as evidence behind it.
Here are five questions to ask.
1. What exactly is being measured?
Is the figure:
- annual or cumulative?
- gross or net?
- cash spending or an economic effect?
- a saving to government or a saving to the wider economy?
- expressed in today’s prices or future prices?
A ten-year total can sound much larger than an annual figure describing the same policy.
2. What is the baseline?
A “saving” only exists relative to what would otherwise happen.
Ask what spending, tax revenue or behaviour the model assumes if the policy is not introduced.
If the baseline is hidden, the saving is difficult to interpret.
3. Can you see the assumptions?
A useful costing should allow an outside reader to understand the arithmetic.
For a deportation programme, for example, assumptions might include the number removed each year, average detention time, voluntary/enforced mix, staffing, construction, flights and support costs avoided.
Without those inputs, a headline cannot be reproduced.
4. Does behaviour change?
Tax and spending policies can change what people and businesses do.
Cutting a tax rate does not necessarily reduce revenue by the simple static amount. Raising a tax does not guarantee the maximum theoretical yield. A new enforcement programme can become more or less expensive as scale changes.
Serious analysis tests those effects rather than assuming the world stays still.
5. Has anyone independent stress-tested it?
Independent scrutiny does not make an estimate automatically correct, but it exposes assumptions.
During the 2024 election, the Institute for Fiscal Studies examined Reform’s Contract and concluded that its package of tax cuts and spending increases was not financed by savings of the scale claimed. The IFS estimated that proposed tax cuts would cost more and several proposed savings would raise less than Reform stated.
That example is now historical — Reform’s policy platform has evolved — but the lesson remains current:
A costing is strongest when somebody outside the organisation can reproduce it, challenge it and still understand how the number was reached.
The two-minute test
When you see a political number, ask:
What period? What baseline? What assumptions? What behavioural response? Can I reproduce it?
If the answer to the last question is no, treat the number as a claim, not an established fact.