Mental Accounting and the Easy Money
Two bodies of evidence on mental accounting are moving in opposite directions, and the gap between them is wide enough to change what the idea is worth. We lay both out at full strength and decline to settle it.
Answer to the question
Why does a refund feel free to spend?
Because money gets sorted into separate pots and the pots are not treated as interchangeable, an idea called mental accounting. The field evidence is strongest for American food benefits, where fifty to sixty cents of each benefit dollar went on eligible food against far less of each cash dollar. The laboratory demonstrations are a different story: the house money result reversed in a preregistered re-test of 504 people.
Key takeaways
- A 2025 registered replication of Thaler's seventeen mental accounting problems supported eleven, partly supported three and did not support three, on 1,007 participants.
- The house money demonstration reversed: 143 of 504 people took the bet, 28 per cent, against about 70 per cent in the original.
- Out of each dollar of American food benefit, fifty to sixty cents went on eligible food, against far less out of each dollar of cash, and this held for households whose benefits were smaller than their grocery bill.
- The payday spending spike in 60 million transactions from 75,000 people is attributed by the journal's own editorial summary mostly to rent and utility payments scheduled to land with income.
Ask why a tax refund gets spent differently from the salary that produced it, and the question usually turns into an argument about whether that is irrational. That is the wrong place to start. Mental accounting, the idea that people sort money into separate pots and then treat the pots as though they cannot be swapped for one another, has been the standard answer for forty years, and the useful question is which evidence anyone is pointing at when they give it.
Because the two bodies of evidence behind mental accounting are now moving in opposite directions. The survey and laboratory demonstrations, the ones that reached the textbooks, are weakening under preregistered re-testing, meaning tests whose hypotheses and analysis were locked in and reviewed before any data was collected. The evidence from records of what households actually bought is holding or getting stronger. We are going to lay both out and decline to settle it.
What does mental accounting actually claim?
Richard Thaler set the idea out in Marketing Science in 1985 and returned to it in 1999 in the Journal of Behavioral Decision Making, where the paper is a tour of seventeen problems rather than a model that predicts anything in advance. The claim is about fungibility, the property money is supposed to have of being interchangeable. A pound is a pound whatever its source. Mental accounting says nobody behaves as though that were true.
The formal statement sits in an appendix. Daniel Kahneman reprinted his 1984 paper with Amos Tversky at the back of Thinking, Fast and Slow, and its section on transactions, pages 441 to 444, gives the machinery: a purchase can be booked to a narrow account, a topical one or the whole of a person's wealth, and which account it lands in changes the answer.
Credit for the idea belongs to Thaler rather than to the more famous author carrying it, and Kahneman says so himself in chapter 32, whose endnote points at Thaler's review. What Kahneman adds in his own voice is one sentence about a hierarchy of accounts, spending money and general savings and money earmarked for a child's education, which he says are drawn on with very different willingness. One sentence, with no data under it.
Two neighbouring questions are worth keeping separate from this one. A price split into a headline and a surcharge is a question about how the total is arranged rather than where the money came from. And paying with a card instead of notes varies how the money is handed over, not which pot it leaves. Mental accounting is the pot.
Why are the laboratory results shrinking?
In 2025 Mengfei Li published a registered replication of Thaler's 1999 problems in Royal Society Open Science, with Gilad Feldman. The team built testable versions of the seventeen problems and ran them on 1,007 online participants in the United States, roughly 500 per problem. Eleven were supported, three partly, and three not at all.
The house money effect is one of the three. The problem asks you to imagine you have just won thirty dollars, then offers a coin flip for nine either way. Thaler and Eric Johnson's 1990 paper in Management Science reported about seven people in ten taking the bet, which is where the phrase came from. In the replication 143 of 504 people took it, and 72 per cent turned it down. The result did not shrink. It reversed.
The house money demonstration did not shrink in a preregistered re-test of 504 people. It reversed.
| Problem, as Thaler put it | Original | 2025 replication |
|---|---|---|
| A coin flip for nine dollars after winning thirty | about 70 per cent took the bet | 28 per cent |
| A chance to break even after losing thirty | 60 per cent gambled | 24 per cent |
| A coin flip for thirty after losing thirty | 40 per cent took the bet | 17 per cent |
The third row is the one that held, in the sense that it moved in the same direction as the original. The first two did not.
What survived is the part about account boundaries. Told they had been given a fifty-dollar dinner, 83 per cent of participants would go on to buy a twenty-five-dollar theatre ticket; told they had spent fifty dollars on the dinner, 40 per cent would. That gap is large and it stood. It runs on the same machinery as a sunk cost, the one where money already spent keeps someone finishing a series they stopped enjoying, and it is the half of mental accounting that is not in trouble.
Two other pillars are wobbling. Hal Arkes and five colleagues ran the foundational windfall experiments in 1994, comparing students warned about a payment days ahead with students who learned of it on arrival; the unwarned group gambled more in one study and spent more at a basketball game in another. A preregistered replication of those studies was approved at its design stage in 2026 and has not reported. There is no result yet, which is not the same as a failure, and nothing can be read into the silence.
The payday spike is the weakest. Michael Gelman and colleagues examined 60 million transactions from 75,000 people for Science in 2014, and the journal's own editorial summary says the apparent spree after a paycheque arrives is mostly explained by the convenience of linking rent and utility payments to the date income lands. The Choice Factory uses that study at pages 134 to 135 as evidence of a psychological splurge, and credits the third author in its prose where its reference list correctly gives Gelman first.
What do the transaction records show?
Here the picture is better, and it gets better as the data gets closer to a till.
### Coupons, labels and gift cards
Katherine Milkman and John Beshears compared what customers of an online grocer bought when redeeming a ten-dollar-off coupon with what the same customers bought without one. Spending rose by $1.59 on the coupon trip, and the extra went on groceries the household did not normally buy. Their 2009 paper says the pattern fits mental accounting and then explicitly discusses other psychological explanations rather than claiming the mechanism.
Johannes Abeler and Felix Marklein attached a label to part of a budget, first in a field experiment and then in the laboratory. Consumption moved the way the label suggested, and the people most likely to break fungibility were those with weaker mathematical ability. Nicholas Reinholtz and colleagues found in 2015 that shoppers holding a retailer's gift card lean toward products typical of that retailer, even a fictional one, which is a finding about what gets bought and not how much.
### Petrol grades and food benefits
Justine Hastings and Jesse Shapiro built the two strongest tests. In 2013, in the Quarterly Journal of Economics, they used household records of petrol purchases and found that when petrol prices rise, people downgrade to a cheaper octane grade by more than their loss of income can account for. They reject the idea that households treat petrol money as interchangeable with the rest of their money. The thing measured is which grade gets bought, not how much is spent overall.
In 2018 they turned to American food benefits. Out of each dollar of benefit, between fifty and sixty cents went on benefit-eligible food; out of a dollar of cash, far less. The force of the paper is a subgroup: households whose benefits were smaller than their food bill, for whom the benefit is economically identical to cash. They still spent it differently.
Does any evidence point the other way?
Yes, in the same policy area. Hilary Hoynes and Diane Whitmore Schanzenbach studied the staggered county-by-county introduction of American food stamps in the 1960s and 1970s using a long-running household survey panel, and reported that households responded to a dollar of food stamps much as they did to a dollar of cash. We have not reached that article, so we are printing none of its figures here.
Put the two side by side and they are not the same study with a different answer. One reads modern retail transaction records; the other reads a historical survey panel across a national rollout. Different decade, different measurement, different households.
The sharpest objection to mental accounting comes from inside the seed book, and it is not about replication at all. On pages 443 to 444 Kahneman argues that if it genuinely feels better to save five dollars on a small purchase than on a large one, and genuinely more annoying to pay twice for a ticket than to lose cash, then those secondary consequences may be legitimate, and preferences built on them "cannot readily be ruled out as inconsistent or erroneous". Non-fungibility might not be an error.
What cannot be settled yet?
Quite a lot, and saying so is the finding rather than a dodge.
The house money effect cannot be described as established. Its central demonstration reversed in a preregistered test of 504 people. The windfall result cannot be described as settled either, because the re-test is running. The payday spike cannot be used as psychological evidence at all. The Choice Factory prints a figure for how much more freely windfall money gets gambled; that figure is the book's report, we did not reach the paper's own, and we are not printing it.
Nor can anyone claim money is non-fungible in general. It has been shown non-fungible in specific settings: petrol octane under a price shock, eligible food under a benefit, a labelled budget, and gift-card category choice. That list is the claim. There is also no citable version of the common complaint that mental accounting is merely descriptive because account boundaries are never specified ahead of time, so we attribute it to nobody.
So, a rule for reading the next mental accounting claim you meet, including ours. Three questions, in order:
- Did this come from a survey problem or from a record of a purchase? The two have diverged, and the problems are the half under revision.
- Which pot was actually tested? Petrol, eligible food, a labelled budget and a gift card are the pots with field evidence behind them.
- Was the money treated as separate by accident, or earmarked on purpose? Kahneman's hierarchy of accounts is a self-control device rather than a mistake, and a fence you put up on purpose, around a holiday fund or a bill account, is not a failure of arithmetic at all.
Common questions
Why does a refund feel free to spend?
Because the mind files it in a different account from wages. A 2025 registered replication of Thaler's 1999 problems with 1,007 participants supported eleven of seventeen, including the result that spending fifty dollars on a dinner, rather than being given it, cuts the appetite for a further twenty-five-dollar purchase from 83 per cent to 40 per cent. The house money and break-even demonstrations reversed.
Do people really spend windfall money more freely?
The original 1994 experiments by Hal Arkes and five colleagues found that students who were not warned about a payment gambled and spent more than students who were. A preregistered replication of those studies was approved in 2026 and has not reported, so the honest answer is that the foundational result is currently being re-tested and no outcome exists.
Is the payday spending spree real?
Not as psychology. Michael Gelman and colleagues looked at 60 million transactions from 75,000 people, and the journal's editorial summary attributes most of the spike after a paycheque to rent and utility payments scheduled to land with income. The spending rises because the bills do.
What is the strongest evidence that money is not interchangeable?
American food benefits. Justine Hastings and Jesse Shapiro reported in 2018 that fifty to sixty cents of each benefit dollar went on eligible food against far less of each cash dollar, and that this held for households whose benefits were smaller than their grocery bill, for whom the benefit is economically the same as cash.
Sources
Thaler, R. (1985). 'Mental Accounting and Consumer Choice.' Marketing Science, 4(3), 199-214. doi:10.1287/mksc.4.3.199. [Crossref item record read; abstract not supplied by Crossref for this item]
Thaler, R. H. (1999). 'Mental accounting matters.' Journal of Behavioral Decision Making, 12(3), 183-206. doi:10.1002/(SICI)1099-0771(199909)12:3<183::AID-BDM318>3.0.CO;2-F. [Crossref item record read; no abstract deposited]
Thaler, R. H., & Johnson, E. J. (1990). 'Gambling with the House Money and Trying to Break Even: The Effects of Prior Outcomes on Risky Choice.' Management Science, 36(6), 643-660. doi:10.1287/mnsc.36.6.643. [Abstract read in full; the full text was not reached and no sample sizes are printed in the abstract]
Hastings, J. S., & Shapiro, J. M. (2013). 'Fungibility and Consumer Choice: Evidence from Commodity Price Shocks.' The Quarterly Journal of Economics, 128(4), 1449-1498. doi:10.1093/qje/qjt018. [Abstract read in full; household count in the panel NOT VERIFIED]
Hastings, J., & Shapiro, J. M. (2018). 'How Are SNAP Benefits Spent? Evidence from a Retail Panel.' American Economic Review, 108(12), 3493-3540. doi:10.1257/aer.20170866. [Abstract read in full; panel household count and date range NOT VERIFIED]
Milkman, K. L., & Beshears, J. (2009). 'Mental accounting and small windfalls: Evidence from an online grocer.' Journal of Economic Behavior & Organization, 71(2), 384-394. doi:10.1016/j.jebo.2009.04.007. [Abstract read in full, from the identical SSRN version doi:10.2139/ssrn.1018382; customer and transaction counts NOT VERIFIED]
Arkes, H. R., Joyner, C. A., Pezzo, M. V., Nash, J. G., Siegel-Jacobs, K., & Stone, E. (1994). 'The Psychology of Windfall Gains.' Organizational Behavior and Human Decision Processes, 59(3), 331-347. doi:10.1006/obhd.1994.1063. [Crossref item record read; no abstract deposited and the full text was not reached, so the per-study figures and sample sizes are NOT VERIFIED]
Reinholtz, N., Bartels, D. M., & Parker, J. R. (2015). 'On the Mental Accounting of Restricted-Use Funds: How Gift Cards Change What People Purchase.' Journal of Consumer Research, 42(4), 596-614. doi:10.1093/jcr/ucv045. [Crossref item record read; abstract read only via secondary summaries; per-study sample sizes NOT VERIFIED]
Abeler, J., & Marklein, F. (2017). 'Fungibility, Labels, and Consumption.' Journal of the European Economic Association, 15(1), 99-127. doi:10.1093/jeea/jvw007. [Abstract read in full, from the identical working-paper deposit doi:10.2139/ssrn.1139870; sample sizes of both experiments NOT VERIFIED]
Gelman, M., Kariv, S., Shapiro, M. D., Silverman, D., & Tadelis, S. (2014). Science, 345(6193), 212-215. doi:10.1126/science.1247727. [Abstract and editor's summary read in full; the editorial summary attributes most of the payday spike to regular payments linked to regular income]
Li, M., & Feldman, G. (2025). 'Revisiting mental accounting classic paradigms: Replication Registered Report of the problems reviewed in Thaler (1999).' Royal Society Open Science, 12(9). PMCID PMC12445221, PMID 40977913. Endorsed by Peer Community in Registered Reports, doi:10.24072/pci.rr.100375. Data at https://osf.io/v7fbj/. [Abstract read in full; the problem-by-problem results table read directly]
Teng, Zhang, Qi and colleagues. 'Revisiting the psychology of windfall gains: Replication and extensions Registered Report of Arkes et al. (1994)'s Studies 1, 2, and 3.' Stage 1 registered on OSF 26 May 2025 (osf.io/b25tr); PCI RR Stage 1 recommendation doi:10.24072/pci.rr.101145, 23 June 2026. [No Stage 2 result exists; searched PCI RR and OSF]
Hoynes, H. W., & Schanzenbach, D. W. (2009). 'Consumption Responses to In-Kind Transfers: Evidence from the Introduction of the Food Stamp Program.' American Economic Journal: Applied Economics, 1(4), 109-139. [Abstract and framing read via search results only; the published article was NOT read and no DOI was verified, so no figures from it are printed]
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. Appendix B, the reprinted Kahneman and Tversky 1984 paper, section on transactions and trades, pp. 441 to 444, and ch. 32, pp. 342 to 352. [Read in the vault's own 26-row audited reading digests]
Shotton, R. (2018). The Choice Factory. Harriman House, pp. 134 to 136. [Read in the vault, audited from 108 photographs; the prose credits Shapiro where the reference list gives Gelman as first author]
Thaler, R. H., & Sunstein, C. R. Nudge: The Final Edition, p. 62, for the house money label. [Read in the vault's reading digests]