What leaves no receipt

Posted on: 30 July 2026

The 5p cut to fuel duty was introduced in March 2022 as a temporary response to an energy shock, and it is now in its fifth summer. It has been extended at every fiscal event since, most recently on 20 May of this year, when the government confirmed that the full cut would hold until 31 December 2026 rather than beginning its phased withdrawal in September as the Budget had scheduled. Duty on petrol and diesel therefore stands at 52.95p per litre, its lowest level in more than sixteen years. Before the cut it was 57.95p, a figure that had itself been frozen since 2011. Four and a half years of temporary relief sitting on top of fourteen years of frozen rates, and the word temporary has survived intact throughout.

The Treasury's own estimate of what this saves the average motorist is roughly £120 over two years. On a 55 litre tank the 5p comes to £2.75, or £3.30 once VAT is added on top. The Office for Budget Responsibility scrutiny note observes that extending the cut is expected to reduce CPI by four hundredths of a percentage point in the final quarter of the year. That is the government's own arithmetic, not a critic's, and it describes an intervention whose measurable effect on the price level is very close to nothing.

Set against that, industry estimates put the additional cost imposed on British drivers by the Middle East disruption since February at around £3 billion, following a rise of roughly 26p per litre at the pump. On the same 55 litre tank that is a little over £14. The relief that gets announced covers under a quarter of the increase that does not.

None of this makes the policy indefensible. For a haulage operator running a fleet, 5p per litre is a genuine line in the accounts, and the additional 3.7p cut in the rebated rate on gas oil introduced in June has a real balance sheet consequence for agriculture and construction. The question worth asking is not whether the money is real. It is what the money is buying, and for whom, because the answer is not the one the announcement implies.

In 1981 Amos Tversky and Daniel Kahneman published an experiment in Science that remains the clearest thing ever written about this mechanism. One group of respondents was asked to imagine buying a jacket for $125 and a calculator for $15, and was told the calculator was available for $10 at another branch twenty minutes away. Sixty eight per cent said they would make the trip. A second group was given the same problem with the prices reversed, a $15 jacket and a $125 calculator on offer at $120. Identical journey, identical five dollars. Twenty nine per cent said yes.

The samples were ninety three and eighty eight respondents answering a questionnaire rather than actually driving anywhere, which is the honest limitation of the study and the reason it should not be swung about as a cudgel. The direction of the finding has held up well enough across replications to reason from, and what it shows is that willingness to spend twenty minutes has almost nothing to do with what twenty minutes is worth. It tracks the size of the saving relative to the price of the object. Five dollars against fifteen feels like a third. Five dollars against a hundred and twenty five feels like nothing. They are the same five dollars.

Two readings are available here and the comfortable one is wrong. The comfortable reading says people cannot do sums, which is the register in which behavioural economics has been popularised for four decades, delivered with the faint self satisfaction of someone explaining other people's mistakes to them. It has the disadvantage of failing to explain why the same error persists among competent, prosperous individuals who are perfectly capable of working out a percentage.

The uncomfortable reading was set out by Elias Khalil in Managerial and Decision Economics in 2024. Calculating the opportunity cost of your own time on every occasion is cognitively expensive, so the mind adopts a shortcut and uses the price of the item as a rough proxy for how much attention the decision deserves. Cheap thing, cheap deliberation. Expensive thing, sustained thought. As a general rule this works impressively well, and it is why nobody spends half an hour choosing toothpaste. The shortcut is not the problem. The problem is that the shortcut loses its validity at exactly the point where it is being applied with the greatest diligence.

Consider the subscription costing £9 a month that you no longer use. Cancelling it requires finding the login, working through three levels of menu, discovering that cancellation is only accepted by form, then waiting for confirmation. Forty minutes, realistically, spread across two attempts because the first one stalled. The price of the object says nine pounds, small thing, not worth attention. The actual calculation says £108 a year, which is what you are paying for not having spent forty minutes on something the unit price persuaded you was negligible. The firms that design those menus are not doing it out of malice. They have read the same studies and they know the heuristic is working for them.

The most instructive British example is not commercial at all. Fuel Finder, the price comparison scheme that obliges forecourts to publish their prices so that drivers can shop around, is the search for a better price installed as public policy. It is a sensible measure and it corrects a real information asymmetry. It also takes a national population and directs its attention towards differences of one or two pence per litre, on the reasonable assumption that motorists will act on what they find. Acting on it means driving somewhere. A supermarket forecourt three miles further on, selling at 4p less, returns about £2.20 on a fill up and costs a mile and a half each way of fuel plus whatever the queue is doing on a Friday afternoon. The scheme measures the pence. It has no instrument for the rest, because the rest generates no transaction and therefore no record.

That is the structural point and it generalises well beyond fuel. We treat as free anything that does not produce a receipt. This is not meanness or innumeracy, it is the architecture of mental accounting itself, a system built to track exchanges which has no idea what to do with a resource that depletes while you stand still. From which follows a conclusion that tends to go unstated because it flatters nobody: saving has a cost of production. Comparing, travelling, waiting for the promotion, assembling the furniture yourself, all of it consumes something that never appears in the ledger. When the cost of production exceeds the value produced the operation runs at a loss and continues to feel like a gain, because the only legible entry is the positive one.

Then there is the layer above, where somebody is genuinely making money. The sensation of having got something is the highest margin commodity ever brought to market, for the straightforward reason that it costs nothing to manufacture. The value does not sit in the discounted product. It sits in the discount. Richard Thaler, who built much of a career on this, called it transaction utility, the pleasure derived from the deal itself as distinct from the pleasure of owning the thing. It accounts for loyalty cards, for points that expire, for the struck through price beside the new one, for the countdown timer on the checkout page. Not one of these devices makes the product cheaper. All of them make the moment of purchase more intense.

Which brings us back to a temporary tax cut in its fifth year. Look at the structure rather than the intention. An intervention whose own impact assessment puts its effect on consumer prices at four hundredths of a point, extended by announcement at the moment of maximum public attention to pump prices, with the unwinding scheduled for 3p on 1 January and a further 2p on 1 March, followed by RPI uprating from April 2027, which is to say a return to annual increases for the first time since 2011 arriving conveniently after the extension has been forgotten. The principal product here is not the saving. It is the same thing the supermarket sells with its loyalty card, the sense that somebody has attended to your price. Government sells it through an identical mechanism, with the distinction that you pay the cost of production twice over, once as a driver receiving the relief and once as a taxpayer funding it out of the £24 billion the duty raises.

There is a short term test of whether any of this holds, which is the best property an argument can have. The weekly pump price data will show whether the phased increases from January pass through to the forecourt in full and stay there, or whether they vanish inside crude movements within the same fortnight. If duty changes track the pump cleanly in both directions then the instrument does what it claims and this reading can be discarded. If they do not, the question stops being whether the measure worked and becomes what it was working on.

I live a few miles from a national border, which is a useful vantage point on all of this. The differential in fuel prices between the two sides moves traffic with great reliability, predictable queues at predictable hours, thinning when the gap narrows and swelling when it widens. The people in those queues are not getting their fuel arithmetic wrong. They tend to have it to the last centime. The travelling time simply does not appear in the cost column, because time issues no invoice and nobody has ever watched it leave their current account.