The Sunk Cost Trap

The Rational Baseline and Why It Fails

In classical economics, a sunk cost is any cost that has already been incurred and cannot be recovered regardless of future action. The rational principle that follows is straightforward: sunk costs are irrelevant to decisions about what to do next. The only inputs that matter are the expected future costs and future benefits of each available option. What you have already spent, lost, or invested should exert precisely zero influence over that calculation.

This framework is logically airtight. It is also routinely violated. In 1985, economists Hal Arkes and Catherine Blumer published a now-landmark study in which participants who had paid more for a ski trip were more likely to drive through a blizzard to use it, even when a cheaper trip to a better resort had become available. The subjects who had paid more reported greater commitment to attending, not because the ski conditions were superior, but because abandoning the trip felt like acknowledging a loss. The higher the sunk cost, the more irrational the persistence.

The Mechanism: Loss Aversion as the Engine

The sunk cost trap runs on loss aversion, the cognitive asymmetry identified by Daniel Kahneman and Amos Tversky in their 1979 development of Prospect Theory. Their research demonstrated that losses register with roughly twice the psychological weight of equivalent gains. When a person has invested time, money, or effort into a course of action, abandoning it does not feel like reallocating resources toward a better opportunity. It feels like suffering a loss equal to everything already put in.

The mind reframes continuation as the path that avoids the loss and reframes exit as the path that confirms it. This reframing is not a deliberate calculation. It is automatic. The person experiencing it typically believes they are reasoning about future prospects when they are in fact processing the emotional weight of what they stand to write off. The more specific and personal the prior investment, the more powerful the distortion. Money produces a weaker sunk cost effect than time. Time produces a weaker effect than identity, when the investment has become part of how a person understands themselves.

Barry Staw's 1976 research on escalating commitment introduced a second layer: people who made the initial decision to invest tend to commit more aggressively to failing courses of action than people who inherited those same failing situations from predecessors. The decision-maker protects not just the investment but the judgment that produced it. Admission that a project is lost implies admission that the decision to begin was wrong.

Cross-Domain: Markets, Projects, and Institutional Budgets

The aviation industry produced one of the clearest large-scale examples in recorded economic history. The British and French governments launched the Concorde supersonic passenger jet program in 1962. By the mid-1970s, the economic case for the aircraft had collapsed: operating costs were prohibitive, the projected commercial market had not materialized, and orders from airlines beyond the two national carriers had evaporated. Both governments knew this. Both continued funding development and production anyway, committing additional billions precisely because they had already committed billions. The Concorde flew commercially for 27 years while losing money throughout its operational life. The phenomenon of continuing a demonstrably unviable project because of what has already been spent became known in behavioral economics as the Concorde fallacy.

The same pattern operates routinely in corporate capital allocation. Research by Mathew Hayward and Donald Hambrick, published in 1997, documented that CEOs who championed major acquisitions subsequently approved further resource transfers to the acquired unit at rates that tracked the original acquisition premium paid, regardless of post-acquisition performance. The higher the initial price overpaid, the more capital was subsequently directed toward rescuing the investment. Boards that approved the initial decision were least likely to authorize its reversal. The mechanism is identical to Arkes and Blumer's ski trip subjects: the decision-makers most responsible for the original commitment are the least capable of evaluating whether that commitment should continue.

Cross-Domain: War, Policy, and the Political Cost of Exit

Military and political contexts generate some of the most consequential sunk cost effects on record. During the United States' prosecution of the Vietnam War, internal government assessments produced by the Pentagon Papers, compiled between 1967 and 1969 and published in 1971, demonstrated that senior policymakers across multiple administrations privately concluded that the war could not be won on acceptable terms. The public rationale for continued escalation during this period centered increasingly on the sacrifices already made: the lives lost, the resources committed, the credibility at stake. Each addition to the casualty count became an argument for continuing, on the logic that withdrawal would render prior losses meaningless. The dead became a reason to produce more dead.

This is the sunk cost trap operating at national scale. The future costs and benefits of continued military engagement were being evaluated through the distorting lens of what had already been spent. Exit was reframed not as a rational response to an unfavorable cost-benefit calculation but as a betrayal of prior sacrifice, a framing that made the emotional cost of stopping greater than the strategic cost of continuing. Political leaders across partisan lines fell into the same pattern, demonstrating that the trap does not require poor analytical capability. It requires only that prior investment carry emotional and reputational weight, which it almost always does when the stakes are high enough.

Why the Trap Persists

The sunk cost trap has survived centuries of human decision-making because it is partly adaptive. Commitment to a difficult course of action, sustained through temporary setbacks, is genuinely valuable in many contexts. The trait that causes a person to abandon ship at the first sign of rough water is also a liability. The cognitive bias toward continuation evolved in an environment where most goals rewarded persistence and where the penalty for premature abandonment was real. The problem arises when the same disposition is applied to situations where the original conditions have fundamentally changed and continuation genuinely has a negative expected value.

The bias also resists correction because awareness does not dissolve it. Arkes and Blumer found that subjects who were explicitly told about the sunk cost fallacy still exhibited the effect. Knowing that prior costs are irrelevant in theory does not make them feel irrelevant in practice. The emotional processing that elevates exit to the status of loss occurs below the level where rational knowledge can intervene, which is precisely what makes the trap useful to those who know how to exploit it.

Detection: Separating History from Forecast

The most reliable diagnostic question for sunk cost distortion is: if you were encountering this situation for the first time today, with no prior history in it, would you choose to enter at current terms? If the answer is no, and the only argument for staying is what has already been committed, the sunk cost trap is active. The framing does not require any mathematical sophistication. It simply requires separating the question of what has happened from the question of what is likely to happen.

A second diagnostic is to examine who is generating the strongest arguments for continuation. If those arguments come overwhelmingly from people whose prior judgment is implicated in the current situation, their analysis should be treated as compromised, not by dishonesty, but by the same emotional processing that afflicts every decision-maker with skin in the historical game. Seeking the assessment of individuals with no prior stake is not a procedural nicety. It is the closest available approximation to the unbiased future-oriented analysis that the sunk cost trap prevents insiders from providing.

Counter-Measures That Work

Prospective accounting is the most operationally useful counter-measure: before beginning any significant investment, establish in writing the specific conditions under which you would consider it a failure and the criteria that would trigger exit evaluation. These criteria, written when future bias is absent, give you access to your own uncontaminated judgment at the moment it is most needed. Organizations that formalize pre-mortem analysis, which asks teams to imagine a project has already failed and to work backward to identify how, consistently catch sunk cost reasoning before it becomes institutionalized.

Structural separation of evaluation from execution is equally important at the institutional level. The people tasked with assessing whether a project should continue should not be the same people responsible for its current trajectory. This is not a guarantee against the trap, but it removes the layer of identity protection that makes the bias most acute. When decision-makers know that continuation will be evaluated by parties who have no reputational stake in the original choice, the incentive to persist as self-justification is reduced, and the evaluation is more likely to track actual future prospects rather than the weight of what has already been lost.

---