INKWIRE(7) argument errors INKWIRE(7)
sunk-cost
The Sunk Cost Fallacy
Continuing because of what you already spent, not because of what is left to gain.
NAME
Definition
A sunk cost is money, time or effort that cannot be recovered no matter what you do next. The fallacy is letting it into the decision. The rational question is always forward-looking: from here, which option gives the better outcome?
The error is not "never finish anything". Finishing is often right, because abandoning has its own costs — lost value, contractual penalties, damaged trust. The error is counting a cost that is identical across all remaining options as if it differed between them.
It is stubborn because abandoning feels like a verdict on the past. Continuing protects the story that the earlier decision was sound, and that story is worth more to the decision-maker than the outcome is.
SIGNALS
-
S1
The justification refers to the past
"We have already put three years into this" is a statement about history, not about the choice in front of you. If removing it changes nothing about the future outcomes, it should change nothing about the decision.
-
S2
The two options are compared with different baselines
Continuing is measured from the current position; quitting is measured from zero, as though the spent resources come back if you stop. They do not.
-
S3
Escalation follows investment
Commitment increases as the sunk amount grows, which is the opposite of what a forward-looking calculation would do.
Three checks, in order
- Ask the clean question: "If I arrived here today with no history, what would I choose?" Compare the answer with what you are about to do.
- List only the costs and benefits that lie ahead. Anything already spent is removed from both columns.
- Separate the two decisions: was the original choice right, and is continuing right now? They have different answers surprisingly often.
Where the link breaks
| The sunk-cost sentence | The forward-looking version |
|---|---|
| We have come this far, we cannot stop now | Remaining cost X, remaining benefit Y, so continue if Y > X |
| I paid for the whole ticket, so I will stay to the end | The next two hours are worth more elsewhere or they are not |
| We already bought the machine, so we should use it | The machine is free now; use it only if it beats the alternative |
EXAMPLES
The project is eighteen months behind and the market has moved, but we have spent four million on it, so we have to see it through.
Why it fails
The four million is gone under every option, including cancelling. If the remaining spend buys less than it costs, stopping is the better decision, and the only thing the four million does is make stopping feel like an admission rather than a calculation.
Kindred entries
- [F] post-hoc B happened after A, therefore A made B happen.
- [F] motte-and-bailey Advance a bold claim, retreat to a modest one when challenged, then quietly re-advance.
- index — /t/fallacy/
[+] Notes and objections 3
- Is finishing what you started always the fallacy?
- No. Finishing is often correct — sometimes the last increment of work really does deliver most of the value. The fallacy is the reason, not the action. "Finish because the remaining value exceeds the remaining cost" is sound; "finish because of what we already spent" is not.
- How does this interact with reputation?
- Badly, because reputation is a real forward cost. If stopping now makes future partners trust you less, that belongs in the calculation. The trap is inflating it to protect a decision rather than estimating it.
- What is the best single question to ask?
- Ask what you would do with the remaining resources if they were handed to you as cash today. Most people would not buy back into the same project.
Scope — Reference material for reading arguments, not professional advice.