Sunk-cost fallacy
The sunk-cost fallacy is continuing a course of action because of resources already invested — time, money, or effort — even when quitting would be the better choice. Past costs that can’t be recovered should never drive future decisions, yet they do.
Why it happens
Loss aversion makes “wasting” a prior investment feel like a fresh loss, and a desire for consistency makes quitting feel like admitting a mistake. The mind treats unrecoverable past costs as if they were still at stake in the present.
Examples
- Sitting through a bad film to the end because you paid for the ticket.
- Staying in a failing project or relationship “after all this time invested.”
- Governments funding doomed megaprojects rather than absorb the loss — the original Concorde case.
How to counter it
- Ask: “Knowing what I know now, would I start this today?” If no, stop.
- Treat unrecoverable costs as gone — they are irrelevant to the next decision.
- Set kill-criteria before you begin, so quitting is a plan, not a defeat.
The deeper point
It isn’t really about the money — it’s about identity. We don’t escalate to protect the investment; we escalate to protect the story that we weren’t wrong to make it. Quitting feels like admitting, not just losing.
Frequently asked
- What is the sunk-cost fallacy in simple terms?
- Throwing good money (or time) after bad because you’ve already spent so much — letting unrecoverable past costs dictate future choices instead of expected future value.
- Why is it hard to ignore sunk costs?
- Quitting feels like wasting the investment and admitting you were wrong. Loss aversion makes that emotional cost loom larger than the rational case for stopping.
- How do you avoid the sunk-cost fallacy?
- Judge decisions only by future costs and benefits, ask whether you’d start fresh today, and set clear stop conditions in advance.
Related
Editorial synthesis © ReadGlobe 2026, drawing on Kahneman’s Thinking, Fast and Slow, the Tversky–Kahneman research program, and the primary cognitive-science literature. · Last reviewed 2026-05-29.