A pharmaceutical company has spent $200 million developing a drug over eight years. Clinical trials now suggest the compound is unlikely to receive approval. The rational choice—abandon the project—feels almost impossible. Instead, executives approve another $50 million push, hoping to salvage what's already been spent.

This pattern appears across boardrooms, poker tables, home renovations, and romantic relationships. Economists call it escalation of commitment: the tendency to increase investment in a failing course of action precisely because prior investment has occurred. The money is gone. The rational analyst treats it as irrelevant. The human decision-maker treats it as a debt demanding vindication.

What makes sunk cost behavior particularly costly is its invisibility to the person committing it. Escalating investors rarely see themselves as irrational. They see themselves as committed, patient, or believing in their vision. Understanding the psychological machinery behind this bias—and the structural interventions that interrupt it—reveals why some organizations exit failing projects gracefully while others ride them into the ground.

Barry Staw's foundational 1976 experiments established the empirical pattern clearly. Participants who made an initial investment decision that produced poor results allocated significantly more resources to the same division than participants who inherited the failing situation. The difference wasn't information—both groups saw identical data. The difference was ownership of the prior choice.

Field evidence corroborates the laboratory findings. Studies of NBA draft picks show that players selected higher receive more playing time throughout their careers, controlling for performance. Bank loan officers extend credit to troubled borrowers longer when they originated the loan themselves. Real estate investors hold declining properties past objectively rational exit points, particularly when they set the initial purchase price.

The escalation pattern intensifies under specific conditions: when decision-makers feel personally responsible, when the investment is public and observable, when project completion remains theoretically possible, and when losses have been gradual rather than sudden. Each factor amplifies the psychological pull backward toward vindication of the original choice.

Notably, escalation is not universal irrationality. In many contexts, persistence in the face of setbacks produces excellent outcomes. The behavioral question isn't whether to persist, but whether the decision to continue reflects updated evidence or protection of prior commitments. The two look identical from inside the decision.

Takeaway

The past investment isn't the problem—the emotional relationship with that investment is. Ask whether you'd start this project today, knowing what you now know, if you were beginning from scratch.

The core engine of escalation isn't financial calculation—it's identity protection. Admitting a failed investment means admitting a flawed judgment, and human cognition resists that admission with remarkable ingenuity. Continued investment functions as a promise that the original decision will eventually be vindicated.

Leon Festinger's cognitive dissonance framework explains the mechanism. When actions and self-image conflict, we adjust our perception of reality rather than our self-image. Investors don't think, 'I made a bad choice, so I should exit.' They think, 'The market doesn't yet understand what I saw.' The failing project becomes evidence of everyone else's shortsightedness rather than the investor's.

This effect compounds with public commitment. Once an executive has defended a project in earnings calls, walking away means acknowledging error to shareholders, journalists, and colleagues. The reputational cost of admission grows with each additional round of defense. Continued investment becomes cheaper than the psychological price of reversal—at least in the short term.

Group dynamics amplify individual self-justification. Teams collectively invested in a project's success reinforce each other's optimism, screen out contradictory information, and stigmatize members who raise doubts. The organizational culture of loyalty and grit, generally virtues, becomes the mechanism by which billion-dollar mistakes get extended year after year.

Takeaway

Ego is the most expensive line item in any failing project. The question isn't whether the venture can succeed, but whether continuing serves the outcome or serves the identity of the decision-maker.

Structural interventions consistently outperform willpower in interrupting escalation. The most effective approach is separating the decision to continue from the person who initiated the project. When a fresh evaluator reviews the venture without ownership of the prior choice, decisions align more closely with objective prospects.

Pre-commitment devices work by binding future behavior to present judgment. Setting explicit exit criteria before beginning a project—specific metrics that will trigger abandonment—removes the discretion that escalation exploits. Venture capital firms use staged funding for this reason: each round requires justification against pre-set milestones rather than accumulated sentiment.

Reframing techniques also reduce escalation. Asking 'What would I do if I inherited this situation today?' strips away the ownership effect that drives commitment. Similarly, converting sunk costs into salient opportunity costs—'This next $50 million could fund three new ventures'—shifts attention from vindication to comparison.

At the organizational level, cultures that reward honest project termination outperform those that punish it. When abandoning a failing initiative is treated as good judgment rather than personal failure, escalation loses its psychological fuel. Some firms institutionalize this through 'kill fees' or promotion credit for executives who successfully wind down underperforming units, making termination a career-positive action.

Takeaway

Rational exits require systems, not resolve. Design the off-ramp before you enter the highway, because by the time you need it, you won't be capable of building it.

The sunk cost trap persists not because people can't calculate expected value, but because human decision-making integrates identity, memory, and social positioning alongside pure economics. Understanding this integration is more useful than pretending it can be trained away through logic alone.

The strongest defenses against escalation are architectural rather than motivational. Pre-committed exit criteria, rotating evaluators, and cultures that celebrate graceful termination outperform any amount of individual discipline. The bias is stable; the systems around it are not.

The next time you find yourself defending a choice by pointing to what it has already cost, pause. The past is fixed. Only the next dollar, hour, or year remains yours to allocate wisely.