i

Description

Concorde landing (old photograph)

Source

author not stated · original page

Licence

CC0 Public domain

Original photo, unretouched. The background is a version of it stylised by the site.

Sunk costs

What is already spent won't come back, yet it weighs on the decision to continue as if it could be recovered.

In brief

• What : a past investment (money, time, effort) weighs on the decision to continue, even though it is lost no matter what happens next.
• Why : stopping means acknowledging a loss and admitting a mistake; we want to avoid “waste” and stay consistent with ourselves.
• Detect : the main argument for continuing is “we've already invested,” “we can't stop now.”
• Avoid : compare only future costs and benefits, set stopping criteria in advance, have someone who didn't commit the spending make the decision.

On this page

1. Definition
2. The mechanism
3. Where you encounter it
   1. In project management
   2. In management
4. Famous cases
   1. Concorde, or the “Concorde effect” 1962-2003
   2. The theater subscription 1985
   3. Elf's “sniffer planes” (France) 1976-1979 ; revealed in 1983
5. Detection signals
6. Avoidance reflexes
7. With AI: detect, fix, mitigate
   1. Detect
   2. Fix
   3. Mitigate
   4. Prompts to integrate into your practices
   5. How to integrate them
   6. Precautions
8. Limits
9. Sources and verification
10. See also
    1. Combines with, or reinforces
    2. Related biases
    3. Counterweights and related reflexes

Definition#

Fictitious illustration

Fictitious illustration

A sunk cost (sunk cost) is spending already committed that no future decision can recover. In sound decision logic, it should not count: only the future costs and benefits of each option matter. The bias, often called the sunk cost fallacy, is the tendency to factor it in anyway—that is, to keep pursuing an action because we've already invested heavily in it.

The reference experimental demonstration is by Hal Arkes and Catherine Blumer (1985). The economic reasoning is much older; Richard Thaler linked it to choice psychology as early as 1980.

The bias is not limited to money: time (“I spent three months on it”), effort (“we rewrote everything”), and reputation (“I defended this choice before the committee”) weigh just as much.

The mechanism#

Several drivers combine:

• Aversion to waste : giving up means “losing for good” what we put in. As long as we continue, the loss stays virtual; it becomes real when we stop (see Loss aversion).
• Consistency and self-image : changing course means admitting the previous decision was bad, which carries reputational and self-esteem costs.
• Mental accounting : we keep a “project account” and want to close it in the black. That account closes only if we go all the way.
• Hope of making it back : we think a little more effort will be enough to make it all worthwhile, which brings the bias close to Optimism bias.

Where you encounter it#

In project management#

• Project drift : “we've already spent 800,000 euros, we're not going to stop at 90%.” The displayed completion percentage is often itself optimistic.
• Go / no-go arbitration : the committee compares the cost already incurred with the initial budget instead of comparing remaining cost with remaining value.
• Technical debt : a shaky component is kept “because we spent two years writing it,” even though rewriting or replacing it would cost less than continuing to work around it.
• Choosing a tool or vendor : the license is paid for three years, training is done, so we keep the tool even when an alternative would save several hours per week.
• Branches and features : people refuse to remove development never used because it was expensive.

In management#

• Hiring and assignment : someone struggling is kept in a role because we've invested in their training or onboarding period.
• Reorganization : a team, site, or business line is maintained on the grounds of resources already committed.
• Training plans and internal initiatives : a program launched by an executive keeps getting funded so they don't lose face.
• Personal career : staying in a path or role “after so many years of investment.”

Famous cases#

Detection signals#

• “We've already invested too much to stop.”
• “If we stop now, everything we've done will have been for nothing.”
• Budget consumed is cited as an argument for continuing, rather than the work remaining.
• The decision is made by the person who launched the initiative, or by a committee that already approved it.
• The more the project struggles, the more requests for additional resources are presented as “the final push.”
• People compare the current state with the initial budget, never with what would be decided if starting from scratch today.

Avoidance reflexes#

1. Look only at the future. For each option, list only future costs and benefits. What has already been spent belongs in the history, not in the calculation.
2. Ask: “If we were starting from scratch today, would we launch this project?” If the answer is no, continuing should be justified on other grounds.
3. Set stop criteria in advance. Cost threshold, date, value indicator: decide before committing under which conditions you stop (see Escalation of commitment).
4. Have an outsider make the call. A decision-maker who did not commit the spending or defend the original choice is less sensitive to the bias.
5. Break commitment into stages. Fund in phases, with a real decision point (and the possibility of stopping) at each stage.
6. Make stopping honorable. Highlight what has been learned, talk about “reorientation” and knowledge capture, so stopping is not experienced as an admission.
7. Run a pre-mortem. Before renewing funding, imagine the project failed in one year and look for why.

With AI: detect, fix, mitigate#

Prompts to integrate into your practices#

1. “From scratch” analysis. To use before a continue-or-stop committee meeting.

Here is the situation of a project: [goal, scope, progress status, remaining work, expected value, alternatives].
Completely ignore what has already been spent (money, time, effort). Assume I am discovering this topic today.
1. Estimate, with a range, the remaining cost and duration to reach the goal, and indicate the most fragile assumptions.
2. Compare three options based only on future costs and benefits: continue, reduce scope, stop.
3. State under which conditions each is preferable.
4. End with one sentence: would we launch this project today, at this cost?

2. Find past-based reasoning in meeting minutes. On anonymized committee or review minutes.

Here are project review minutes (names removed). Identify the arguments in favor of continuing and classify them:
- those based on what has already been spent or done (past);
- those based on what remains to be done, gained, or lost (future).
Quote the exact sentences. Conclude in one sentence: does the decision seem mainly justified by the past? Do not invent any information missing from the text.

Minutes: [text]

How to integrate them#

• Before the committee: the project manager runs prompt 1 on their own file and notes the result, without showing it.
• During: one person is responsible for asking the question “what if we started from scratch?”
• After: prompt 2 is used as a process review on meeting minutes, not to blame people.

Precautions#

• AI only knows what you give it: a flattering file yields a flattering analysis. Also provide unfavorable figures.
• Its remaining-cost estimates are orders of magnitude, not commitments: validate them with those doing the work.
• It may cite studies or references that do not exist: verify every source before using it.
• Do not send budgets, contracts, or personal information to an unauthorized tool.

Limits#

• The effect is robust in laboratory settings, but weaker among people trained in economics or decision-making, and among children or some animals, which has been discussed (Arkes and Ayton, 1999).
• Not everything boils down to a bias: continuing may be rational if remaining cost is genuinely lower than remaining value, or if stopping creates its own costs (contractual penalties, loss of trust, site closure). These future costs legitimately matter.
• Reputation is a real issue: a decision-maker may have an incentive not to stop. The bias then mixes with a conflict of interest, and diagnosis becomes more delicate.
• The Concorde case is often cited in support, but the reasons for continuing were also political and industrial.

Sources and verification#

• Arkes & Blumer (1985), “The psychology of sunk cost,” Organizational Behavior and Human Decision Processes — ski trip and theater subscriptions.
• Arkes & Ayton (1999), “The sunk cost and Concorde effects: Are humans less rational than lower animals?,” Psychological Bulletin
• Dawkins & Carlisle (1976), “Parental investment, mate desertion and a fallacy,” Nature — origin of the expression “Concorde fallacy.”
• Thaler (1980), “Toward a positive theory of consumer choice,” Journal of Economic Behavior & Organization
• Concorde (Wikipedia, in English) — treaty of November 29, 1962, estimate of 70 million pounds, cost of 1.5 to 2.1 billion in 1976, first flight and commercial service. Figures should be treated as orders of magnitude.

See also#