Cognitive biases
The sunk cost fallacy: why it is hard to quit what you have already invested in
Your gym membership is paid a year ahead. In February you realize you don’t want to go: it is far, it is inconvenient, and you like running in the park more. But you keep traveling across the city. “I’ve already paid for it.”
This is the sunk cost fallacy: we continue something because we have already put money, time, or effort into it, not because continuing pays off. Sunk costs are what has already been spent and will not come back, whatever you decide next.
Below, we look at the classic experiments, how strong this effect is in real decisions, and how to tell reasonable persistence from a trap. In the middle of the article you will choose between two tickets for the same evening.
Contents
In short: how the sunk cost fallacy works
What the sunk cost fallacy is
Hal Arkes and Catherine Blumer described a series of experiments in a 1985 paper. The best known is the problem of the two trips. Imagine: you bought a ticket for a ski weekend for 100 dollars. Later you bought a second one, for 50, to a different resort, and you are sure it will be more fun. Then it turns out that both trips are on the same weekend, and the tickets cannot be refunded or resold. Which trip will you take? Most participants chose the more expensive trip. Even though, by the terms of the problem, they would have enjoyed the second trip more.
There was also an experiment with real money. A university theater sold season subscriptions with random discounts: some people paid the full price, some paid less. Those who paid the full price attended more performances over the next six months. The performances were the same. The only difference was how much money had already gone.
From the standpoint of the logic of decisions, it is all simple here, and Arkes and Peter Ayton put it in a 1999 review: only future costs and benefits should influence the choice. The past can no longer be changed.
How strong the effect is in life. In problems on paper the effect shows up clearly. In real decisions the picture is more modest. A meta-analysis by Roth and colleagues confirmed that the effect exists, but its size depends strongly on the type of decision. In decisions like “should I use something I have already paid for,” it weakens over time, and it is less pronounced in older people. In a laboratory game with real money, the influence of past costs turned out to be surprisingly small. And the well-known NBA study, in which players picked earlier in the draft got more playing time even after accounting for how well they played, gave a weaker and shorter effect when the data were rechecked. The takeaway: the trap is real, but not all-powerful.
Try it yourself: two tickets
This is almost word for word the Arkes and Blumer problem, only with concerts. It takes a minute.
Step 1. You bought a ticket to a concert. For some players it costs €120, for others €20, and the computer picks the price. The ticket cannot be refunded or resold.
Step 2. Later you paid €40 for a ticket to a different event on the same evening, and you realize you will enjoy it more. Also non-refundable.
Step 3. You choose where you will go.
Step 4. You move the price of the first ticket and see what really changes in your evening. Then you answer where you would go if that ticket had been given to you as a gift.
Why is the problem built this way? Everything that matters for the evening is known in advance: the second event is better. Both tickets are already paid for, and the money will not come back whichever you choose. Only the amount that cannot be recovered differs. The random price is there for comparison: at €20 the first ticket is cheaper than the second, and there is almost nothing pulling you toward it. At €120 the pull is stronger. If your choice changes along with the price, sunk costs are influencing it.
What usually happens. In the original problem, most people chose the more expensive trip, although they knew in advance that they would like the other one more. The money for both trips had already been spent. But giving up the expensive ticket felt like throwing away a hundred dollars.
When it is not a mistake. If you promised to go to the concert with a friend, if it is the last concert of a favorite band, if people are waiting for you there, these are reasons about the coming evening, not about the money spent. The mistake begins where the only reason sounds like “but I paid for it.”
Why it is so hard to quit
Imagine that the money has fallen into a well. You cannot get it out, however hard you try. Now there is only one question: where to go from here. This is an image, but it describes the logic accurately: the well is equally deep whichever road you choose.
Arkes and Ayton offered an explanation. Adults have a useful rule: “don’t waste.” Don’t throw away food, wear a thing out, finish what you started. Usually it serves us well. But we apply it where it does not work: when the money is already spent, and continuing will not save it but only add new costs. A curious detail from the same review: small children in similar situations acted more sensibly than adults. The authors believe this may point to a learned rule rather than an innate error.
And there is a simple human feeling. To quit seems to admit that the investment was a failure. By continuing, we put off that unpleasant moment. Sometimes for years.
Where it gets in the way
Relationships. “We’ve been together for so many years” is a serious reason, because behind the years stand a shared history, children, closeness. But the phrase “so much has been invested” does not by itself answer the question of whether you are doing well together now and whether it will get better. It helps to separate two questions: “what have we invested” and “what do we have and what can come next.” For a decision about the future, the second matters more.
Work and business. “We have already spent half a year and forty thousand on this system, we can’t just drop it.” Sometimes a project is continued not because it has become more promising but because it feels like a pity to lose what has been done. The more has been invested, the harder it is to say “stop,” although every new month costs as much as the previous one.
Money and shopping. Repairing the old car one more time because the engine was replaced in spring. Holding stocks that have fallen “until they recover.” Finishing a plate at a restaurant when you are already full because “it’s paid for.” In each case one question is useful: what will I get from here on, and how much will that cost from here on.
Studies and time. The third year in a program that has stopped being interesting. A series you stopped liking long ago, but “there are only two seasons left.” A book you struggle through because you have read half. Time is a sunk cost too. But not everything here is a mistake: a degree may be useful in the future, and that is an honest reason to continue.
How to decide whether to continue
The sunk cost fallacy is one of the biases on the cognitive bias map, in the section on decisions. In the same section is the anchoring effect, which influences how much we are ready to pay from the very beginning.
Frequently asked questions
So is quitting what you started always the right thing?
No. Sometimes continuing is reasonable: you are almost at the finish, the thing has started to work, you gave a promise, or other people depend on you. The mistake is not persistence but letting what has already been spent become the main reason.
What is the Concorde effect?
It is another name for the same effect, after the supersonic Concorde airplane. It is often used to describe animal behavior. But according to the review by Arkes and Ayton, there are no clear examples of this error in animals. And small children in similar problems acted more sensibly than adults.
Do we get wiser or more stubborn with age?
According to the meta-analysis by Roth and colleagues, older people on average fall for the sunk cost effect less than younger people.
Is the sunk cost effect as strong in real life as in textbook problems?
Not necessarily. In problems on paper it shows up clearly. In real decisions the picture is more modest: the size depends strongly on the type of decision, in a laboratory game with real money the influence of past costs was surprisingly small, and when the NBA data were rechecked, the effect turned out weaker and shorter than in the original analysis. The trap is real, but not all-powerful.
