The Psychology of Sunk Costs: Why Smart People Keep Making Losing Bets

In 1971, the Nixon administration received a report from its own economists recommending immediate termination of the supersonic transport program. The SST had consumed $1.1 billion in federal funds since 1963, produced no flyable aircraft, and faced mounting evidence that the economics of supersonic commercial aviation were fatally unfavorable. The economists’ analysis was straightforward: the money already spent was gone regardless of what the government decided next; the only rational question was whether continued investment would generate sufficient returns to justify additional expenditure. The answer was clearly no. The administration continued the program for another two years before Congress finally killed it.

Nixon’s advisors were not stupid people. Several held advanced economics degrees. They understood, in abstract terms, the logic of ignoring sunk costs. But understanding a principle in the abstract and applying it in the specific case where you have personal, institutional, and political capital invested in a different outcome are entirely different cognitive tasks. The gap between knowing the sunk cost fallacy exists and actually avoiding it is one of the most well-documented and least-solved problems in human decision-making, and it is not going to be solved by more economics education.

What the Fallacy Actually Is

The sunk cost fallacy, as formally defined in the economics literature, is the tendency to allow irrecoverable past expenditures to influence decisions about future actions. If you have spent $50 on a non-refundable concert ticket but are feeling ill on the night of the concert, the economically correct decision is based entirely on whether attending the concert is worth the discomfort — the $50 is gone either way and should not factor into the calculation. But most people feel a pull to attend precisely because they have paid for the ticket, as if sitting through the concert recovers some portion of the lost money.

The error is clear once stated. It is also extremely common in contexts far more consequential than concert tickets. The Concorde supersonic jet, developed jointly by the British and French governments from 1962, continued to receive funding through the 1970s long after it was clear that the program would never be commercially viable, partly because each government felt that abandoning the project would mean “wasting” the money already spent. The program cost approximately $1.5 billion in 1970s dollars and produced an aircraft that operated at a loss for its entire commercial service life. “Concorde fallacy” became a synonym for sunk cost reasoning in biology and psychology literature.

Corporate acquisitions offer some of the richest examples. When a company has publicly committed to acquiring a target and the bidding competition drives the price above any rational valuation, the acquiring company’s management almost never walks away. They have publicly committed capital and credibility to the deal. Walking away means acknowledging, in front of shareholders and the press, that the money spent on due diligence and the executive time invested in negotiations was wasted. Better, the psychology dictates, to complete the acquisition and attempt to make it work. The result is that the average large corporate acquisition destroys shareholder value — a finding so consistent across decades of research that it has become one of the most robust empirical regularities in corporate finance.

The Evolutionary Logic

The standard behavioral economics treatment of the sunk cost fallacy presents it as a cognitive error, an irrational deviation from the utility-maximizing standard that rational agents should achieve. This framing is wrong in an important way: it treats the bias as a bug when it is almost certainly a feature, adaptive in the environment where human cognition evolved even if maladaptive in the modern contexts where we encounter it.

Consider what life looked like for most of human evolutionary history. Resources were scarce. The effort required to find food, build shelter, or maintain social relationships was high. In this environment, persisting through difficulty was generally the correct strategy, because the alternative — abandoning partially completed projects whenever they became costly — would lead to a life of half-built shelters and uncompleted food caches. The psychological pressure to honor commitments and persist with investments served as a commitment device that prevented the kind of inconsistency over time that would undermine complex multi-period projects.

The sunk cost bias also served a social function. In small group societies where cooperation was essential for survival, individuals who abandoned commitments when conditions became unfavorable were unreliable partners. The tendency to persist even at personal cost signaled to potential collaborators that you could be trusted to honor obligations. A psychology that weighted past investment heavily was a psychology that produced reputationally trustworthy partners, which in a reputation-dependent social environment was a significant fitness advantage.

These adaptive functions do not disappear when we move into modern institutions and financial markets. They persist as cognitive defaults that activate in situations that superficially resemble their evolutionary context, even when the structural logic is entirely different. The manager who cannot abandon a failing project is activating the same psychological machinery that would have made her an excellent, persistent hunter-gatherer, but the machinery is running in an environment where the optimal strategy requires doing exactly what the machinery is designed to resist.

Why Experience Doesn’t Help

If the sunk cost bias were simply a matter of cognitive naivety, we would expect it to diminish with education and experience. The evidence suggests the opposite: experienced decision-makers in high-stakes environments show the bias as reliably as undergraduates in laboratory experiments, and in some studies show it more strongly.

The most revealing evidence comes from professional investors. Institutional investment managers are explicitly trained to ignore sunk costs and evaluate positions purely on the basis of forward-looking expected returns. Many of them can articulate the principle fluently. Yet empirical studies of trading behavior consistently find that professional investors are significantly less likely to sell losing positions than winning positions at equivalent holding periods — a pattern called the disposition effect. They are holding the losing positions too long precisely because selling them would mean crystallizing the loss, converting a paper loss into a realized one and triggering the psychological recognition that the original investment was a mistake.

The disposition effect is not eliminated by experience in markets. Studies comparing novice and experienced traders find that while experienced traders show somewhat less disposition effect on average, the difference is modest and the bias persists even among traders with many years of experience. What experience does teach is when to override the bias consciously through explicit decision rules — stop-losses, portfolio reviews with preset criteria, forced sell disciplines — but these are workarounds rather than cures. Absent an external forcing mechanism, even experienced professionals default to the same sunk-cost-influenced behavior as everyone else.

Military history provides some of the starkest examples of how catastrophically the bias can operate in consequential decisions made by experienced leaders. The German continuation of the Stalingrad offensive after encirclement in late 1942, over the objections of field commanders who understood the military situation, was partly driven by Hitler’s inability to accept that the enormous investment of German forces and prestige in the capture of Stalingrad had already been irretrievably lost. The correct decision — strategic withdrawal before encirclement was complete — was psychologically unavailable because it required acknowledging that the prior investment was gone.

When Persisting Is Actually Correct

The sunk cost fallacy’s ubiquity has given rise to a popular analytical overcorrection: the belief that any appeal to prior investment in a decision is automatically irrational. This is too simple. There are cases where information about past investment legitimately updates the forward-looking calculus, and distinguishing these from genuine sunk cost fallacies requires careful analysis.

The most important legitimate use of past investment information is what economists call the updating problem. If you have invested heavily in a project and the project is struggling, this fact provides evidence about the project’s underlying quality — evidence that Bayesian reasoning should incorporate into your forward-looking assessment. A project that has consumed twice its estimated resources and is half complete is not the same as a new project that requires the same remaining investment; the cost overrun is a signal about execution quality and underlying viability that should legitimately increase your estimate of future costs and decrease your probability of success.

The second legitimate case involves learning economies. In some industries, the knowledge and capabilities developed through early investment genuinely reduce the cost of future investment. The early losses in semiconductor manufacturing — where initial fabrication facilities cost far more than the revenue they generated — were correctly understood by their investors as the price of learning a manufacturing process that would become profitable only after years of iterative improvement. “Staying the course” in this context was not sunk cost reasoning. It was rational anticipation of a learning curve that could not be climbed without making the early losses.

The practical diagnostic for distinguishing genuine sunk cost fallacies from legitimate persistence is whether the reason for continuing involves the past investment itself or forward-looking information. “We should continue because we’ve invested so much” is always a fallacy. “We should continue because the investment we’ve made has generated learning that improves our forward-looking probability of success” may be correct, but requires the decision-maker to demonstrate the specific channel through which past investment improves future prospects.

The Institutional Fix

Individual psychological debiasing is largely ineffective against the sunk cost fallacy. The bias is too deeply rooted in motivational architecture to be eliminated through awareness alone. The more effective interventions operate at the institutional level, designing decision processes that make abandonment easier and less psychologically costly.

The most successful institutional fix is the “pre-mortem” approach developed by psychologist Gary Klein: before committing to a project, require the decision team to assume the project has failed and identify the most likely reasons why. This exercise activates the failure-analysis thinking that normally occurs only after commitment and makes decision-makers conscious of the specific conditions that should trigger project abandonment before any investment has been made. Having pre-specified the exit conditions, decision-makers find it psychologically easier to act on them — because exiting does not represent a failure of commitment but rather the rational execution of a pre-agreed plan.

The deeper structural fix is separating the decision about whether to continue from the people who made the original commitment. This is why central banks have independent boards rather than being run by the finance ministers who made the fiscal decisions that created the conditions the bank must manage. It is why courts are separated from executives. It is why the most sophisticated investment firms assign “devil’s advocate” roles to analysts whose explicit job is to argue for exit from existing positions. The psychology of sunk costs is not going to be eliminated from human minds. The appropriate response is to design institutions that do not rely on individual minds correctly applying a principle that evolution has spent hundreds of thousands of years teaching them to ignore.

Nixon’s economists were right about the SST. The money was gone. The question was only about future costs and benefits, and the analysis clearly favored exit. That they lost the argument to political considerations is not merely a story about bad policy — it is a story about how human psychology operates under institutional pressure, and about the gap between knowing the correct principle and being positioned to apply it. Closing that gap is an institutional design problem, not an educational one.