The sunk cost fallacy, and why the past keeps a vote it should not have

An everyday example

Someone buys a ticket to a concert, and on the day of the show they feel unwell and would rather stay home. They go anyway, telling themselves it would be a waste not to, even though the ticket is already paid for and non-refundable either way. Going does not get the money back. It only adds a miserable evening to an evening that was already going to cost the same amount.

The mistake is not caring about money that was spent. It is letting spending that cannot be undone by any future choice decide a future choice anyway.

The classic studies

Hal Arkes and Catherine Blumer's 1985 paper gave the pattern its name with a set of short scenarios. In one, people imagine paying 100 dollars for a non-refundable weekend ski trip to Michigan, then finding a better trip to Wisconsin for 50 dollars and buying that ticket too, before realising the two trips are on the same weekend. Asked which trip they would actually take, most people chose the Michigan trip, the one they said they expected to enjoy less, because it had cost more. The rational answer does not depend on either price, since both are already spent; it depends only on which trip is more enjoyable, which by the scenario's own description is Wisconsin.

Arkes and Blumer also ran a field study with the Ohio University theatre, mailing season-ticket offers at different prices to genuine buyers. Buyers who paid the full price attended more of the plays early in the season than buyers who received the same season ticket at a discount, even though every buyer already held a ticket to every play regardless of what they had paid for it. Price paid, not enjoyment or availability, was moving attendance for a ticket that could not be un-bought.

Barry Staw's 1976 study, published before the phrase "sunk cost" was attached to this literature, is usually read together with Arkes and Blumer's work because it isolates a related mechanism: personal responsibility. Participants played the role of a financial officer allocating research funding between two divisions of a company, then, after being told how that investment had turned out, allocated a second round of funding. Participants who had personally made the original choice, and were then told it had gone poorly, put more of the second round of money into the very division whose failure they were responsible for than participants who had not made the original choice themselves. Staw called the pattern escalating commitment to a chosen course of action: not just weighing sunk costs, but doubling down on a decision because it was yours.

Does it replicate?

Replication grade: Strong: reproduced across many studies and multi-lab replication, smaller with real stakes

The vignette version of the effect, choices like the Michigan and Wisconsin ski trips, has been reproduced in many follow-up studies since 1985 and was one of the effects retested in the large, multi-site Many Labs project, which reran a set of classic findings in dozens of laboratories at once specifically to see whether they held up outside their original lab. The basic direction of the sunk-cost effect reproduced there, which is evidence that the pattern is not an artefact of one lab, one country or one generation of students.

A real caveat is worth stating plainly: studies that put real money or real consequences on the line, rather than a hypothetical vignette read on paper, tend to find a smaller effect than the paper-and-pencil studies do, and the size varies with how directly the past spending is described as wasted if you stop. That does not overturn the finding. It means the vignette studies, including the famous ski-trip question, likely overstate how large the effect is when something genuinely costly is actually on the line, even though the direction of the effect, the pull of past spending on a forward decision, keeps showing up.

A scale that should ignore one of its own pans

The diagram on this page is a schematic, not data from a study. It draws a balance scale with the money already spent piled on one pan and the future cost and benefit on the other, and tips the scale toward the spent money, the pan that a correct decision should not weigh at all. The fallacy is not a miscalculation on the correct pan; it is letting the wrong pan onto the scale in the first place.

Try it: three scenarios

Three short situations, each with money or effort already spent. In each, decide what a purely forward-looking choice would do, ignoring the part that is already gone.

How to catch it

Sunk costs are hardest to ignore exactly when they are largest, which is also when the pull to keep going is strongest.

• Ask the question as if you were deciding fresh today, with none of the past spending in your history: would you start this now, knowing only what lies ahead?
• Separate the two pans out loud: write down what is already spent on one line and what remains to be spent or gained on another, and make the decision from the second line only.
• Notice the phrase "it would be a waste to stop now." The waste already happened; stopping does not create it, and continuing does not undo it.
• If personal responsibility for the original choice is part of what is pulling you, name that separately from the actual costs and benefits ahead, the way Staw's study isolated it.
• Ask what you would advise a friend in the identical position, with no history of their own attached to the choice.
• Set a stopping rule in advance, before any money is spent, for what would make you walk away; a rule written before the cost exists is not distorted by the cost once it does.