The reason you're still in that job. That relationship. That project.
Sunk Cost Fallacy
Some of your worst decisions feel like discipline.
You've already invested too much to quit.
That sentence feels rational and often it is not. The instinct behind it (hold on, stay the course, do not give up) gets praised far more often than it gets questioned.
By the end of this, you'll know how to recognise when past investment is making your decisions instead of future outcomes.
That's the meme version. It is sometimes correct.
The meme guards against quitting too early. The more expensive failure runs the other way: five more years in the wrong career or three more funding rounds for the wrong investment.
What the research shows: once people have invested, the amount already spent can sway whether they continue, even when it should not change the comparison of future options. Money you can't get back starts to act as a reason, though no future choice can change it.
How it works
In a field study, Arkes and Blumer (1985) randomly sold season tickets for a university theatre at full price or at one of two discounts. People who paid full price went to more plays than people who got a discount, though the plays were the same. The difference was statistically significant in the first half of the season but not in the second.
The same pull shows up in everyday choices:
driving through a storm to justify a ticket
finishing a bad book because you started it
staying in a failing project because of time already spent
What a sunk cost is
A sunk cost is any time, money or energy that cannot be recovered. It is gone whatever you do next. So it should not affect the forward-looking decision.
Not everything you put in is spent. Skills, code, relationships and reputation built along the way are still yours and they still change what happens next. The rule: ignore what is gone and count everything that still changes the future.
Ask instead:
Given where I am now, what leads to the best outcome from here?
Why past costs still count
The leading account combines two ideas. Thaler (1980) argued that people keep mental accounts: a payment opens an account and walking away closes it at a loss. Loss aversion then makes that booked loss feel larger than the future gain from stopping, while continuing keeps the hope of recovery alive. Arkes and Blumer (1985) added a simpler driver: people do not want to look wasteful to others or to themselves.
The effect is real but smaller and less durable than the meme suggests. A meta-analysis of 98 effect sizes (Roth, Robbert and Straus 2015) found clear evidence for it. Its size varied with the type of decision and in decisions about using what you bought it faded over time.
A field experiment in Zambia is a warning against reading every link between price and use as sunk cost. Ashraf, Berry and Shapiro (2010) found that households who paid more for a water purifier did not use it more. The higher price selected households that already planned to use it. When people who paid more use more, part of the gap may be selection: they were already the keenest.
Walking away feels like:
admitting failure
wasting effort
breaking a story about ourselves
So people continue because stopping is psychologically expensive, even when continuing is the worse option.
Also shows up in
- The sunk cost fallacy rarely looks irrational. It sounds like:
- 'I've put too much into this to quit now.'
- 'I can't leave after everything I've invested.'
- Every one of these argues from the past.
Countermeasure
Run the Fresh-Start Audit before any material new commitment to something already under way, especially one that is hard to reverse. A rule of thumb: at least a month of work or a tenth of the original budget counts as material. The audit has five steps.
1. Separate sunk from salvageable. List what is gone and what you can still use: skills, code, contacts, reputation.
2. Run the fresh-start test. Ask one question:
If I hadn't already invested anything, would I choose this today?
Not whether it is too late, what people will think or how much you have put in.
3. Run the forward-value test. Compare the options from here only: remaining cost, expected benefit, risk, the cost of exiting and the best alternative use of the same time and money.
4. Define the continuation condition. Write down what has to be true, by when and what evidence will show it.
5. Set a review date and the action you will take on it: continue, change course or stop.
A no from the fresh-start test is not an automatic stop. It shifts the burden of proof: going on now needs a forward reason that would convince someone who had spent nothing. The forward-value test is where that reason has to show up.
Large commitments are harder to audit because walking away feels more expensive. So write the continuation condition before you start, as a kill line: one sentence that names the condition and the date that end the project. 'We stop if we have fewer than 20 paying users by 30 June.'
A kill line can still move for a real reason. Material new information (a new market or a changed cost) can change the call. What it rules out is quietly moving the goalposts once the date arrives.
Try it now: pick one thing you're continuing mainly because of prior investment and write down what you'd decide if you were starting fresh today, with nothing already spent.
Where it hurts most
- Driving: Ho, Png and Reza (2018) used Singapore's car-permit policy, which changed how much buyers paid up front. Their model estimates that owners who paid more drove more: about 86 extra kilometres a month per car between 2009 and 2013 (5.6% more driving). The pattern held in Hong Kong data and the extra use faded over time, as in the theatre study. The estimate is model-based. It is consistent with sunk-cost reasoning but it cannot show what drivers were thinking.
- Public projects: (observational) Debates about cancelling large projects that have run over budget often cite the money already spent as a reason to finish. Overrun records show how often costs exceed budgets. They do not show how much of each decision to continue is sunk-cost reasoning and how much is real remaining value.
- Careers and training: (observational) People stay in degree programmes and professional tracks because of the years already put in, then count those same years again as a reason to stay one more.
The uncomfortable version
This gets harder over time. The longer you stay, the more stopping feels impossible.
Sunk costs trap you through consistency. The habit that makes you finish what you start also keeps you finishing things that should stop.
Persistence and misallocation look identical from the inside. One is discipline, the other is inertia. The same mechanism rewards both. The skill is knowing when they have diverged.
There is a second cost. Your biggest commitments (the career, the house, the company you started) are the ones least likely to have faced an audit.
What are you still in mainly because of what you've already put in?
CogniqOS publishes one cognitive trap per week: the mechanism, the myth, the countermeasure. Send this to someone who keeps calling persistence 'discipline' when it might be inertia.
If it landed, restack it.
When it doesn't apply
Past investment sometimes carries real information. The salvage you built along the way belongs in the forward-looking sum. A fresh-start test that throws it out with the spent money gives the wrong answer.
The 'don't waste' rule also earns its place in everyday life. For small repeat choices (finishing a meal or using a gym pass) deliberating each time costs more than the occasional error. The rule fails when the stakes are large and future costs keep rising. Run the fresh-start test there and let the cheap rule handle the rest.
Danger zone
Sunk Cost Fallacy + Loss Aversion + Escalation of Commitment - each new investment raises the perceived cost of walking away, Loss Aversion makes stopping feel worse than the loss already incurred and Escalation of Commitment converts a recoverable mistake into a structural one through repeated doubling.
Also a component of Convergence Trap 1: The Escalation Trap → · Also a component of Convergence Trap 8: The Stagnant Strategy →
Also mapped in Cross-Domain Transfer: AI Systems & Machine Learning → · Also mapped in Cross-Domain Transfer: Military Strategy & Geopolitics → · Also mapped in Cross-Domain Transfer: Product Design & Technology → · Also mapped in Cross-Domain Transfer: Finance & Investing → · Also mapped in Cross-Domain Transfer: Organisational Leadership →
Stress test
1. You paid a non-refundable $400 deposit for a course that no longer fits your plans. Knowing the $400 is gone and cannot be recovered, would you still pay the rest and attend?
2. Your team is 14 months into a vendor integration that needs 5 more. If the first 14 months had cost nothing, would the last 5 still be worth paying for?
3. You hold a stock you bought at $80 that now trades at $50. With fresh cash today, would you buy it at $50? If not, what keeps it in your portfolio?
4. List your reasons for continuing the thing you have in mind. How many point back to what is spent and how many point forward to what you will get?
Testable prediction
If teams write a kill line (condition and date) before a project starts, then the share of projects still funded six months after missing it will be lower than in teams that agree the same criteria only at the first review. If people who paid full price and people who got the same item at a deep discount are tracked for a year, then the gap in how often they use it will be smaller in the second six months than in the first.
Run this audit with your AI
Paste your decision in, and let your assistant audit it through this bias.
You are my CogniqOS sunk-cost surgeon. Bias lens: Sunk Cost Fallacy. Work only with the decision I give you. Do not invent metrics, numbers or forecasts. If you need a fact I have not given, ask for it. MY DECISION: [paste here] One decision you're weighing, in a plain sentence - e.g. "Keep funding the project we've spent 8 months on or kill it?" What are my options and what does doing nothing cost? (Required.) (Optional: what I've already sunk in - time, money, reputation.) If another bias fits better, say so before auditing. If my input is vague, ask me to sharpen it before auditing. If it contains more than one decision, make me pick one first. Audit: 1) Diagnose first: is past investment actually driving this decision? Quote the words in my input that show it before you recommend anything. 2) Sunk vs salvageable: list what is gone and what I can still use (skills, code, contacts, reputation). 3) Fresh-start test: starting today with nothing spent but keeping the salvage, would I choose this path? If not, make me name a forward reason that would convince someone who had spent nothing. 4) Forward-value test: compare continuing with my best alternative on remaining cost, expected benefit, risk and exit cost. Mark every input I have not given you as missing. 5) Continuation condition: what has to be true, by when and what evidence will show it? Set a review date and the action to take on it. If a missing fact is material, do not force a call. Return 'insufficient information', name the missing fact and say how it would change the decision. Return exactly as: Diagnosis: is sunk cost driving this decision? Quote the evidence from my input. Call: continue / change course / stop / insufficient information. Forward-only reasoning: compare the remaining options from today. Sunk vs salvageable: list each separately. Missing information: only facts that could change the call. Continuation condition: what must be true, by what date and what happens if it is not. Keep this in your stack. Next trap in the chain: Loss Aversion.
Related concepts
- Commitment Bias - The pressure to stay consistent with past statements
- Loss Aversion - Losses loom larger than equivalent gains
- Plan Continuation Bias - the pull of the plan itself. From the outside it looks like Sunk Cost because both keep you on a failing course. The difference is where the pull comes from.
- Sunk Cost pulls through what you have already spent. It needs a real investment and larger investments can make the pull stronger. The error is in how you value past costs - they count in a decision they cannot change. The fix is to compare future options without treating unrecoverable costs as a reason to continue.
- Plan Continuation pulls through the plan itself. It can take hold when almost nothing has been spent. The error is in perception - new warning signs get read as noise or as obstacles to push through so the plan never looks broken enough to drop. The fix has to be in place before you start: abort criteria written in advance, fixed review points and one person whose job is to argue for stopping.
- One study is consistent with the split. Wiegmann, Goh and O'Hare (2002) had pilots in a flight simulator meet worsening weather either early or late in a flight. Pilots who met it early, with less invested, flew farther into it. Sunk cost alone does not explain that result. It is one simulator study and not proof. In real decisions the two biases usually run together.
- A quick test - if you would keep going even with nothing spent, look for Plan Continuation. If the money and time already spent are the main reason, it is Sunk Cost.
- Cognitive Dissonance - Rewriting the past to justify the present
Also connected in the mapHyperbolic Discounting·Difficulty-Value Conflation·Path Dependence Bias·Choice-Supportive Bias·The Illusion of Control·The IKEA Effect·Face-Saving Behavior·The Disposition Effect·Opportunity Cost·The Arms Race Dynamic·Credibility Trap
Origin
Sources
- Arkes, H.R. & Blumer, C. (1985). The psychology of sunk cost. Organizational Behavior and Human Decision Processes 35(1)
- Thaler, R.H. (1980). Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization 1(1)
- Roth, S., Robbert, T. & Straus, L. (2015). On the sunk-cost effect in economic decision-making: a meta-analytic review. Business Research 8(1)
- Ho, T.H., Png, I.P.L. & Reza, S. (2018). Sunk Cost Fallacy in Driving the World's Costliest Cars. Management Science 64(4)
- Ashraf, N., Berry, J. & Shapiro, J.M. (2010). Can Higher Prices Stimulate Product Use? Evidence from a Field Experiment in Zambia. American Economic Review 100(5)
- Wiegmann, D.A., Goh, J. & O'Hare, D. (2002). The Role of Situation Assessment and Flight Experience in Pilots' Decisions to Continue Visual Flight Rules Flight into Adverse Weather. Human Factors 44(2)
Read next
Entry #110 of 631 in The Lexicon · see the full Lexicon