Sunk Cost Fallacy: Why Money Already Spent Keeps You Spending

A cost that has already been paid and cannot be recovered is sunk, and in a decision about what to do next it should carry no weight, because it is the same whichever option you pick. Almost everyone agrees with that rule when it is stated plainly; the fallacy is that people do not follow it. The two versions of the problem above describe the same future: 1 million dollars buys a plane that a competitor already sells faster and cheaper. The only difference is that in one version 10 million dollars has already gone, and that difference moved most of Arkes and Blumer's participants from no to yes.

Several things push in the same direction. The first is loss aversion. Kahneman and Tversky's prospect theory, published in 1979, holds that losses weigh more than equal gains, and that people are willing to take risks to avoid closing a loss. Stopping the project books the 10 million as a loss today; continuing keeps the account open and the loss unrealized. Richard Thaler made this link in 1980.

The second is the wish not to appear wasteful. Arkes and Blumer argued that people continue because stopping would mean the earlier spending was for nothing, and that "not wasting" is a rule people follow even when following it wastes more. The third is self-justification. Barry Staw showed in 1976 that business school students playing a company's financial officer, after an earlier funding decision of theirs went badly, allocated more money to the same division than those who had not made the first decision, as if the new money could redeem the old choice.

Two findings put limits on the story. Arkes and Ayton reviewed the animal literature in 1999 and found little evidence that animals honor sunk costs, though Sweis and colleagues reported in 2018 that mice, rats and humans all showed sensitivity to time already invested in a foraging task. And whether it matters whose money was spent is unsettled; Arkes and Blumer found only equivocal support.

Hal Arkes and Catherine Blumer published "The psychology of sunk cost" in Organizational Behavior and Human Decision Processes in 1985. It reported ten experiments: nine questionnaires given to college students in Ohio and Oregon, and one field study with real money.

The first experiment is the one most often retold. Participants were asked to imagine that they had paid 100 dollars for a ski trip to Michigan and later 50 dollars for a trip to Wisconsin that they expected to enjoy more, then found that both trips fell on the same weekend and neither ticket could be refunded or resold. More than half, 33 of 61, chose the 100-dollar Michigan trip they expected to enjoy less. The money was gone either way; only the weekend was still to be chosen.

The radar-blank plane problem, reproduced above, was given in two forms to separate groups. Of those who read that 10 million dollars had already been invested and the project was 90 percent complete, 41 of 48, or 85 percent, said the last million should be spent. Of those who read the same competitive situation as a fresh proposal, 10 of 60, or 17 percent, said yes. Variations followed. Asked for the finished plane's chance of financial success instead of a decision, readers told of the prior investment put it at 41 out of 100, against 34 for the fresh proposal. Raising the fresh proposal's cost to 10 million dollars left the decision unchanged. A third-person version weakened the effect without removing it, and a ski-trip version with both tickets won rather than paid for did not differ significantly from the original; the authors called the support for personal involvement equivocal.

The field study used real money. Patrons buying season tickets for the Ohio University theater were randomly sold tickets at full price, at a 2-dollar discount or at a 7-dollar discount. Over the first half of the season, the people who had paid full price attended more performances than those with the discounts. The plays were the same; only the sunk cost differed.

Any situation in which something has already been spent and something more is being asked can show the pattern.

The sunk cost fallacy is unusual among biases in that most people already know the rule it breaks, so the counter is to make the rule easy to apply in the moment.

• Ask the fresh-start question. Would you choose this option today if you were starting from zero? The second version of the problem above is that question, and 83 percent of people who faced it said no.
• Separate the two accounts. Write down what has been spent and cannot be recovered, then put it aside and list only what each option will cost and return from here.
• Set the stopping rule before you start. A budget cap, a deadline or a milestone agreed in advance is much easier to honor than one invented after the money is gone.
• Hand the decision to someone with no history. Staw's participants who had not made the first decision were far less likely to pour money after it. An outside reviewer is a sunk-cost-free decision maker.
• Name the loss and let it be a loss. Stopping does not waste the earlier spending, which is already past; continuing is what turns a sunk cost into a larger one.

The fallacy is one decision about money already gone; its nearest neighbors differ in what drives the next one.

• It is not escalation of commitment, which is the organizational pattern Staw described: a sequence of decisions in which each new allocation is made to justify the last. Sunk cost reasoning is one of its causes; escalation also runs on reputation, politics and the fear of admitting a mistake in public.
• It is not loss aversion, the underlying finding that losses weigh more than equal gains. Loss aversion explains why closing out a sunk cost hurts; the fallacy is the decision that results.
• It is not the IKEA effect, in which effort spent on something makes you value the thing itself more highly. The sunk cost fallacy can run on a project you no longer value at all.
• It is not irrational escalation, the competitive version seen in the dollar auction, where each bid is made to avoid losing the last one to a rival.