Does paying by card actually make you spend more?
The honest version of a very popular claim: what the famous experiment found, why later work is much more mixed, and the one asymmetry between cash and card that holds regardless of the effect size.
This is one of the most repeated claims in personal finance: cash hurts, cards do not, so you spend more with a card. It appears in books, in articles, in app marketing, and usually with a specific number attached and a confident tone. It may well be true. It is nothing like as settled as the confidence suggests, and the way it is normally presented would not survive contact with the studies it rests on.
The study everyone is citing
The source of most of the confident numbers is a 2001 paper by Drazen Prelec and Duncan Simester with the memorable title "Always Leave Home Without It". They ran sealed-bid auctions in which participants — students — bid for tickets to a sold-out sporting event, with some told they would pay by credit card and others by cash. Bids in the card condition were substantially higher, and the headline figure that circulates from this work is often quoted as roughly a doubling of willingness to pay.
It is a genuinely interesting result. It also has limits that are almost never quoted alongside it, and they matter for whether it says anything about your grocery shopping.
- It was an auction for event tickets, not a purchase in a shop. Bidding for a scarce ticket is a very unusual purchasing situation, with no reference price and a competitive element that ordinary shopping lacks.
- The sample was small and consisted of students, a group whose relationship with credit cards is not typical of a household running a monthly budget.
- Willingness to pay in an experiment is not the same as spending over a month. Even a real effect on how much you would pay for one item does not straightforwardly become an effect on a household total, where budgets, income and habit all constrain the sum.
- It measured a stated maximum, not a completed purchase followed by the arrival of the bill.
None of this makes the finding wrong. It makes it narrow, which is what a single well-run experiment usually is; the problem is the retelling.
The mechanism people offer
The theoretical backing usually comes from mental accounting, associated with Richard Thaler, and from the idea of payment coupling. The argument goes roughly like this: people do not treat money as fungible, and instead sort it into notional accounts with different rules. Alongside that, the pain of paying is a real component of a purchase decision, and it depends on how tightly the payment is coupled to the consumption. Cash is maximally coupled — the money physically leaves at the moment you receive the thing. A card decouples them: the consumption is now, the payment is a statement three weeks away and merged with thirty other charges, so the pain is delayed, diluted and hard to attribute to any one purchase.
This is a good story: coherent, matching ordinary experience, and generating predictions. It is also worth remembering what kind of thing it is — a framework that explains a result, not independent evidence for it. Mental accounting has broad support as a description of how people handle money, and that support does not transfer automatically to any specific quantitative claim about cards and supermarkets.
It is also worth noting that several classic findings in this general area — the behavioural economics of payment and spending — have replicated unevenly. Some effects have held up, some have shrunk considerably when tested with larger samples and pre-registered designs, and some have not survived at all. Anyone citing a thirty-year-old laboratory result as settled fact is not tracking that.
What happened when people looked at real purchases
Later work using field data and actual purchase records is considerably more mixed than the laboratory picture. Some studies find effects in the expected direction but small. Some find effects concentrated in particular categories — impulse and hedonic items rather than staples — which is more specific and more plausible than a general inflation of all spending. Some find nothing detectable once the obvious confounds are accounted for.
And the confounds here are unusually severe, in three distinct ways.
People use cards for different purchases
The choice of payment method is not random with respect to what is being bought. Large purchases go on cards. Online purchases go on cards, necessarily. Small, local, immediate purchases are more likely to be cash. So the average card transaction is bigger than the average cash transaction in almost any dataset, and that fact tells you nothing at all about whether the card caused anything.
People who use cards are different people
Card use correlates with income, age, urban residence, banking status and a good deal else. Comparing heavy card users to heavy cash users is comparing two populations that differ in many ways besides how they pay, and untangling that is exactly the sort of problem observational data handles badly.
Cashless environments are not randomly assigned
Whole countries have moved towards cashless payment over the same decades in which incomes, prices, e-commerce and retail formats all changed. Comparing before and after such a shift means comparing a great many things at once.
There is a further problem specific to this literature: the studies span an enormous stretch of payment technology. A credit-card experiment from the 1990s involved producing a card, signing a slip, and receiving a monthly paper statement. Contactless payment involves a tap. Paying by phone involves a glance at a face and a tap, sometimes with an immediate notification showing the amount and often with an app showing a running balance. Whether the psychological mechanism proposed in 2001 applies unchanged to that is an open question, and it could easily run either way — the tap is even less effortful than a card, while the instant notification restores something of the immediacy that cash had.
The part that is actionable regardless
Here is what does not depend on how the research turns out, and it is the reason this page exists on a site about household ledgers.
Cash and cards are not equal in your records, whatever they do to your behaviour. A card transaction is recorded twice — by you if you bother, and by your bank regardless. It generates a notification at the moment of payment, which is a prompt arriving exactly when the memory is fresh. If you forget to enter it, it is recoverable later from a statement.
Cash has none of that. No second copy, no prompt, no recovery. And the purchases people still make in cash skew small, frequent and local, which means the part of your ledger that goes missing is concentrated in exactly the everyday discretionary spending you most wanted to see. The gap in a cash-heavy ledger is not a random third of the total; it is a biased sample, missing low.
So there is a real asymmetry, it is large, and it has nothing to do with payment pain. If you switch to cash to spend less and stop recording as a result, you have traded a possible modest behavioural effect for a definite and substantial loss of information about your own household. That is a bad trade unless you have a plan for recording the cash, which is a solvable problem but has to be solved deliberately.
The only experiment that applies to you
The honest conclusion is that the effect may well be real and probably modest, that its size in your household is not knowable from the literature, and that you are in a position to find out directly if it matters to you.
- Pick one category that is genuinely variable and where you would plausibly behave differently — food, eating out, or day-to-day miscellaneous spending. Not rent.
- Record it accurately for a month exactly as you normally pay. This is your baseline, and it has to be real recorded data rather than a recollection.
- For the next month, pay for that category in cash, and record every cash purchase at the moment you pay. This is the hard part and the whole experiment depends on it.
- Compare the two totals, and then be sceptical of yourself: check whether the months were comparable. A birthday, a visitor, a holiday or a five-weekend month will move the number more than any payment effect.
- Repeat it before believing it — two months each way, alternating. One pair of months is a story.
Two honest caveats about your own experiment. You will know you are running it, which is itself likely to change your behaviour more than the payment method does — that is not a flaw you can eliminate, only one you should remember when reading the result. And a single household over two months is a very small sample, so a difference of a few percent means nothing. If the difference is large and repeats, you have learned something about yourself, which is more than the literature can tell you.
For the recording side of that experiment, the practical requirement is being able to enter a cash purchase in a few seconds while standing in the shop, with no signal, since cash tends to get spent in markets, basements and places without reception. Fambook writes each entry to the phone first and pushes it afterwards, which is what makes that possible; the number of not-yet-uploaded entries is shown rather than hidden. Whether the experiment shows anything is not something this page will promise you, and nothing here suggests it will save you money — it will tell you what happened, which is all a ledger ever does.
Frequently asked questions
Do people really spend twice as much with a card?
That figure comes from a specific 2001 experiment by Prelec and Simester in which students made sealed bids for sports tickets, and it measured willingness to pay in an auction rather than spending in a shop. It is a real finding in a narrow setting. Generalising it to a household grocery bill goes well beyond what the study tested.
Is the cash-versus-card effect settled science?
No. Laboratory work supports a mechanism through mental accounting and payment coupling, while field and replication work is much more mixed — some effects are small, some are confined to impulse categories, and some do not survive at all. Several classic behavioural results in this area have replicated unevenly, so confident single numbers should be treated with suspicion.
Why is it so hard to study?
Because payment method is never randomly assigned. People use cards for different kinds of purchase than cash, card users differ from cash users in income and age and location, and cashless environments arrived alongside decades of other economic change. Any raw comparison of card and cash spending is measuring all of that at once.
Does tapping a phone count the same as a credit card?
Nobody knows, and it is a fair question. The studies span very different technologies: a 1990s credit-card purchase involved a signature and a monthly paper statement, while a phone tap is faster still but often comes with an instant notification showing the amount. That notification restores some of the immediacy cash had, so the effect could plausibly go either way.
Should I switch to cash to control my spending?
You can try it, but understand the trade. Cash leaves no statement, produces no notification and cannot be recovered later, and the purchases people make in cash are disproportionately the small frequent ones — so a cash-heavy month is systematically under-recorded. If you switch and stop recording, you have swapped a possibly modest behavioural effect for a definite loss of information.
How would I test it on myself?
Pick one genuinely variable category, record a normal month accurately as a baseline, then spend a month paying cash for that category and recording every purchase at the till. Compare, then check whether the two months were actually comparable — a birthday or a school holiday moves the number more than any payment effect would. Repeat before believing it.
Try it for one month
Fambook gives a household one shared ledger: anyone can add an entry in seconds, every entry says who spent it, and the month adds up in one place instead of two. Records with no signal and syncs afterwards. Recording, categories, budgets, statistics, CSV import and export, sync and sharing for two people are free — the subscription only buys you less typing.