Mental models

Sunk Cost Fallacy

Why does money you can never get back keep steering decisions it has no claim on?

The sunk cost fallacy is the tendency to continue an endeavour because of time, money, or effort already spent and unrecoverable, even when the remaining costs and benefits alone would say stop.

TL;DR

  • The sunk cost fallacy is an error about the reason for continuing, not about continuing itself. Persisting counts as the fallacy only when the justification is resources already spent; if the remaining benefits still beat the remaining costs, continuing is correct for a forward-looking reason instead.
  • Past costs are irrelevant to a forward-looking decision because no available choice can recover them. A rational actor compares only what lies ahead — the additional money, time, and risk from this moment onward — and treats what is already spent as fixed background, identical under every option on the table.
  • Sunk cost differs from opportunity cost: sunk cost is spending already made and unrecoverable, while opportunity cost is the value of the best alternative you give up by continuing. Only opportunity cost belongs in the decision, because the fallacy is letting a backward-looking quantity carry any weight at all.
  • Escalation of commitment is the organisational cousin of the fallacy: decision-makers pour further resources into a failing course of action, pushed partly by past spending and partly by the need to justify their earlier choice. The Anglo-French Concorde airliner gave the pattern its other name, the Concorde fallacy.
  • Continuing can be correct for reasons that have nothing to do with sunk cost, including contractual penalties for cancelling, reputation with future partners, and the option value of information the project will reveal. These are forward-looking payoffs, and naming one honestly is the test that separates judgment from fallacy.

When to use it

  • Check for the sunk cost fallacy whenever someone defends continuing with the phrase "we have put too much into this to quit now". That sentence cites a quantity no future choice can change, so it is a report about the past, never an argument about what to do next.
  • Check for it at every stage gate of a long commitment — follow-on funding rounds, renovation budgets, multi-year research programmes — because these are the moments when accumulated spending is most visible and most easily mistaken for a reason to proceed rather than a fact about history.
  • Check for it in personal decisions carrying large unrecoverable investments: a half-finished degree, a long relationship, a career you trained years for. The live question is whether the remaining years beat the best alternative, not how many years you already gave to get here.
  • Run the fresh-start test on any commitment you find yourself defending: if you were choosing today with nothing yet spent, would you begin? A clear no means sunk cost is the only thing holding you, so either stop or rebuild the case entirely on future value.

When it fails

  • The sunk cost accusation fails when continuing rests on genuine forward-looking value — cancellation penalties, reputation with future counterparties, or option value from information the project will produce. These are real payoffs that lie ahead, so labelling the choice a fallacy misreads the actual reason behind it.
  • It fails as an explanation when apparent escalation is really Bayesian updating, because what the spending revealed — test results, early traction, proven feasibility — can rationally raise the estimate of future value. Such a decision-maker responds to evidence rather than to regret, so committing more money is the correct call.
  • The rule fails as blanket advice when cheap abandonment is itself the bias, since a person who quits whenever a project turns hard destroys the credibility and follow-through that make future commitments possible. Ignoring sunk costs is a rule about reasons, not a licence to drop anything unpleasant.
  • It fails to describe organisations where the true driver is accountability rather than psychology, because a manager who cancels must explain the loss publicly while continuing defers that reckoning. That is an incentive problem, fixed by separating the continue-or-stop decision from the person who launched the project.

Worked example

A publisher has spent $400,000 developing a book series and needs $150,000 more to launch it. Fresh market research now projects $120,000 in lifetime revenue, and the manuscripts have no resale value, so none of the $400,000 comes back under any option. The only live comparison is $150,000 of remaining cost against $120,000 of expected return, a $30,000 loss, so shelving is correct. The $400,000 changes nothing, because it sits identically in every branch.

The fallacy arrives in the meeting when someone argues the series must launch because $400,000 is already committed. That reasoning runs backwards: the larger the sunk amount, the stronger it feels as a justification and the more it distorts. Escalation follows when the team approves a further $80,000 of marketing that the same research says will not move the $120,000 forecast, turning a $30,000 loss into a $110,000 one. Each step is defended by the spending of the step before it.

Concorde supplied the pattern's other name. Its development costs had outrun any plausible commercial return by the late 1960s, and airline orders collapsed in the early 1970s, yet Britain and France funded the supersonic airliner to completion. The case is not pure sunk cost: Britain sought to cancel in 1964 and was held by a 1962 treaty with no withdrawal clause, a forward-looking liability rather than regret. Richard Dawkins and T. R. Carlisle took the name in 1976 for the mistaken argument that an animal should defend a brood in proportion to past investment rather than remaining prospects.

Change one fact and the answer flips with no fallacy involved. If cancelling triggers a $200,000 contractual penalty to the printer, launching still loses $30,000 while cancelling now loses $200,000, so launching is $170,000 better even though it never turns a profit. If the launch would also reveal whether a new distribution channel works, that option value can justify proceeding on its own. Neither reason is the sunk $400,000; both are cash flows lying ahead.

Related concepts

  • Expected Value — The forward-looking calculation that replaces sunk cost reasoning: remaining costs weighed against probability-weighted future payoff.
  • Dominant Strategy — Sunk costs enter every option identically, so they can never make one strategy dominate another.
  • Tragedy of the Commons — An analogy rather than a shared mechanism: both keep actors on a losing course, but the commons runs on costs spread across users, not on spending already made.

FAQ

What is the sunk cost fallacy in simple terms?

It is refusing to walk away from something because of what you already spent on it. The money, time, or effort is gone whichever option you pick next, so it cannot make continuing the better choice. Only the costs and benefits still ahead of you should decide.

What is the difference between sunk cost and opportunity cost?

Sunk cost is spending already made that no future choice can recover, while opportunity cost is the value of the best alternative you give up by choosing one path over another. Opportunity cost belongs in every decision because it lies ahead; sunk cost never does, because it lies behind.

Is it always wrong to keep going after investing a lot?

No. Continuing is often the right call, but never because of what you spent. Legitimate reasons are all forward-looking: penalties for cancelling, reputation with future partners, or the value of information the project will reveal. The fallacy lies in the reason you give, not the choice you make.

What is escalation of commitment?

Escalation of commitment is the pattern of pouring more resources into a failing course of action instead of abandoning it. Sunk cost is one driver, but Barry Staw's research puts the main engine in self-justification and public accountability, so escalation appears even where little has been spent. The two overlap without being identical.

Why is it called the Concorde fallacy?

Because Britain and France kept funding the Concorde supersonic airliner long after its development costs had outrun any plausible commercial return. Richard Dawkins and T. R. Carlisle borrowed the case in 1976 to name the Concorde fallacy: the mistaken argument that an animal should defend offspring in proportion to past investment rather than future prospects.

How do you avoid the sunk cost fallacy?

Ask what you would choose today if nothing had been spent yet, then list only the costs and benefits from this moment forward. Require anyone arguing to continue to name a specific future payoff. In organisations, the strongest fix is separating the stop-or-continue decision from whoever launched the project.