How The World Works | Sunk Costs

You paid for the ticket.

The film is terrible.

You have already spent forty minutes watching it.

There are ninety minutes left.

Do you stay?

A surprising number of decisions in business, government, investing, education and ordinary life have exactly this shape.

We continue because we have already paid.

Already waited.

Already worked.

Already defended the idea.

Already told everyone it would succeed.

The past starts voting on the future.

That is the problem of sunk costs.

A sunk cost is a cost that has already been incurred and cannot now be recovered. Economically, the cost is gone. Psychologically, it can feel painfully present.

Quick Read

Sunk costs are past expenditures of money, time or effort that cannot be recovered and therefore should normally not determine what you do next.

OpenStax makes the forward-looking rule explicit: current decisions should compare future costs and future benefits rather than let unrecoverable past spending control the choice.

Arkes and Blumer’s classic 1985 experiments called the behavioural distortion the sunk-cost effect: people become more likely to continue an endeavour after investing money, effort or time in it.

The central question is:

If I had not already spent the money, time or reputation, would I choose this option now?

The One-Sentence Answer

Sunk costs distort decisions when unrecoverable past investment is treated as a reason to continue, even though only the future consequences of continuing versus stopping should determine the current choice.

The Sunk-Cost Chain

investment is made → cost becomes unrecoverable → stopping makes investment feel wasted → continuation protects the story of the past decision → more resources are committed → exit becomes emotionally and politically harder → bad project can survive longer than its future value justifies

The cost does not become more relevant as it grows.

It becomes more psychologically powerful.

What Makes a Cost Sunk?

A cost is sunk when it has already happened and cannot be changed by the decision now being considered.

The movie ticket is sunk if it is non-refundable.

Last year’s research budget is sunk.

Three years already spent building software are sunk.

The hours already spent studying a course are sunk.

The future maintenance bill is not sunk.

The resale value you can still recover is not sunk.

A contractual cancellation fee you would pay only if you stop is not sunk until it becomes unavoidable.

The classification depends on what the current decision can still change.

Sunk Cost Is Not Opportunity Cost

The distinction is one of the most useful in decision-making.

Sunk cost looks backward.

Opportunity cost looks forward.

You paid $20 for a cinema ticket.

That $20 is gone.

The next ninety minutes are not gone.

If staying means giving up ninety minutes of something better, that foregone alternative is the opportunity cost of staying.

The rational decision compares the next ninety minutes of movie with the next ninety minutes of the best alternative.

See How The World Works | Opportunity Cost.

Sunk Cost Is Not Irreversibility

Irreversibility describes a state of the world that cannot easily be undone.

A sunk cost describes a past expenditure that cannot be recovered.

The two can interact.

Destroying a wetland may be irreversible.

The engineering money already spent on the project may be sunk.

The irreversibility of future environmental damage can be highly relevant to whether construction should continue.

The money already spent is not.

See How The World Works | Irreversibility.

Sunk Cost Is Not Loss Aversion

Loss aversion concerns the asymmetric psychological weight of losses relative to equivalent gains.

Sunk-cost reasoning concerns the relevance of past unrecoverable expenditure to a current decision.

Loss aversion can help explain why abandoning a project feels painful.

Stopping forces the organisation to recognise that earlier spending will not be recovered through the hoped-for outcome.

But loss aversion and sunk cost remain different owners.

See How The World Works | Loss Aversion.

The Theatre Experiment

Arkes and Blumer’s classic research included a field experiment with theatre season tickets.

Customers who paid more for their subscriptions attended more plays during the early part of the season than customers who had received discounts.

The performances were the same.

The future entertainment value was the same.

What differed was the amount already paid.

The sunk expenditure appeared to influence later attendance.

The simple interpretation is powerful:

The more we have paid, the stronger the pressure to make the past investment feel worthwhile.

Why “Wasting” the Investment Feels Wrong

People are taught not to waste.

That rule is usually useful.

Do not waste food.

Do not waste money.

Do not waste somebody’s effort.

But after a cost becomes unrecoverable, continuing does not unspend it.

The anti-waste instinct can therefore reverse direction.

We throw more resources into the project to prove earlier resources were not wasted.

Now the attempt to avoid waste creates more waste.

The Business Project That Refuses to Die

A company invests $5 million in a new platform.

After eighteen months, the technical architecture is unstable and a competitor has launched a superior solution.

The team estimates that finishing will cost another $3 million.

The board says:

We cannot walk away after spending five million dollars.

That sentence sounds responsible.

It can be exactly backwards.

The relevant question is not whether $5 million was spent.

The relevant question is whether the next $3 million produces more value than the best alternative use of that $3 million.

The past cost is evidence about the quality of the original decision process.

It is not automatically a reason to continue the project.

The Replacement Test

A strong governance question is:

If we did not own this project today, would we buy it at the price of all future costs required to complete it?

If the answer is no, the past investment may be controlling the decision.

Escalation of Commitment

Sunk costs often appear inside a broader phenomenon called escalation of commitment.

A decision-maker commits to a course of action.

Bad news arrives.

Instead of reducing commitment, the decision-maker increases it.

Why?

sunk costs;
reputation;
self-justification;
political incentives;
hope of recovery;
uncertainty about whether failure is temporary;
organisational identity;
contractual obligations;
loss aversion;
confirmation bias.

Escalation of commitment is therefore broader than sunk costs.

The past investment is one driver among several.

Sunk Costs and Confirmation Bias

Once an organisation invests heavily in a strategy, evidence is no longer neutral.

Good news protects the investment story.

Bad news threatens it.

Supportive evidence is accepted quickly.

Conflicting evidence is called temporary, incomplete or unfair.

Now the sunk cost changes the evidence pipeline.

See How The World Works | Confirmation Bias.

Sunk Costs and Planning Fallacy

A project is late.

The team says, “We are nearly there.”

Three months later, the same sentence returns.

Past investment creates pressure to continue.

The planning fallacy makes the remaining cost look smaller than it will really be.

Together they form a dangerous pair:

We have spent too much to stop, and there is not much left to spend.

See How The World Works | Planning Fallacy.

The 90%-Complete Trap

Projects can be “90% complete” for most of their lives.

Why?

The easy visible parts finish early.

Integration, edge cases, testing, legal approval, migration and operational readiness remain.

The percentage complete is anchored to a task list that underestimated difficult remaining work.

A sunk-cost organisation then sees “90% complete” and concludes stopping would waste almost-finished work.

The right measure is not percentage of planned tasks completed.

It is expected future value minus expected future cost from this point forward.

The Investment Example

You bought a share at $100.

It falls to $60.

You refuse to sell because “I need to get back to $100.”

The purchase price is psychologically powerful.

The market does not care.

The decision today is whether owning the share at $60 has a better expected future return, adjusted for risk and alternatives, than selling and reallocating the capital.

The $40 already lost is not recoverable by insisting the original thesis remain true.

The Disposition Effect Is Related, Not Identical

Investors sometimes hold losing positions too long and sell winners too quickly.

Loss aversion, reference points, taxes, beliefs and sunk-cost reasoning can all contribute.

Do not collapse every refusal to sell into one behavioural label.

The forward-looking question remains stable:

Would I buy this asset now if I did not already own it?

The House Renovation Example

A renovation begins with a $50,000 budget.

Halfway through, hidden structural problems appear.

Another $80,000 is required.

The owner says, “We cannot stop now. We already spent fifty thousand.”

Maybe continuing is correct.

Maybe it is not.

The $50,000 does not decide.

The relevant comparison is:

future cost to finish;
future value of the completed renovation;
cost and value of alternative options;
sale value in the current state;
financing cost;
risk of further hidden problems.

Past spending belongs in the post-mortem.

Future economics belong in the decision.

The Software Rewrite Example

A company has spent three years building an internal platform.

A commercial product now offers better capability at lower expected lifetime cost.

The internal team argues:

We cannot abandon three years of development.

Those three years are evidence about sunk investment and organisational learning.

They can also create reusable components, expertise or intellectual property.

Those recoverable future assets are relevant.

The hours already consumed are not.

Not Everything From the Past Is Irrelevant

This is where simplistic sunk-cost advice becomes dangerous.

Past investment can change the future state.

You spent three years training an engineer.

The training time is sunk.

The engineer’s current skill is not.

You spent money building a factory.

The construction spending is sunk.

The factory’s resale value and productive capacity remain relevant.

The correct rule is not “ignore the past.”

It is:

Ignore the parts of the past that the current choice cannot change, while fully accounting for every present capability, liability and option the past created.

Sunk Cost and Learning

A failed project can create valuable knowledge.

The money spent may be sunk.

The lessons are not necessarily sunk.

Code may be reusable.

Staff may have learned.

Customer information may be valuable.

Contracts may retain options.

A mature post-mortem separates unrecoverable expenditure from surviving assets.

This prevents the opposite mistake: abandoning useful capabilities merely because the original project failed.

When Continuing Is Actually Rational

Stopping is not automatically smart.

Suppose a bridge is 80% complete.

The first 80% cost far more than expected.

Finishing the remaining 20% is still cheaper than every alternative and the completed bridge has enormous future value.

Continue.

Not because 80% has already been spent.

Continue because the future benefits still exceed the future costs.

This distinction is the heart of sunk-cost discipline.

The “Almost Finished” Question

“Almost finished” matters only if it changes the future cost-benefit calculation.

If one more hour completes a valuable project, that one hour can be worth spending.

If one more year is needed to finish a product nobody wants, “almost finished” can be a dangerous phrase.

Completion percentage is not value percentage.

The Student Who Chose the Wrong Course

A student spends two years preparing for a pathway they no longer want.

They say:

I cannot change now. I would waste two years.

The two years are gone either way.

But they may have created knowledge, maturity, credentials or transferable skills.

Those surviving assets belong in the forward decision.

The relevant question is:

From today onward, which path creates the better expected future?

The Student Who Should Not Quit

Now reverse the example.

A student is three weeks from an examination and wants to abandon a subject because recent revision feels painful.

The future cost of completing the course is small.

The credential value is large.

Continuing can be rational.

Again, not because “you have already studied so much.”

Because the remaining effort has a strong forward return.

Sunk Costs in Education

Education is especially sensitive because effort becomes identity.

“I have always been a Science student.”

“We have already spent three years preparing for this school.”

“We bought the entire programme.”

Past commitments can make families continue a learning route that no longer matches the child’s current needs.

But educational decisions also contain switching costs, transition risks and prerequisite structures.

Those future consequences are real and should be counted.

Sunk-cost discipline is not “change whenever unhappy.”

It is “do not let unrecoverable past spending substitute for current diagnosis.”

The Tuition Example

A family pays for a long package.

After several weeks, the fit is clearly poor.

If fees are non-refundable, those fees are sunk.

The remaining learning time is not.

The decision should compare the educational value of staying with the educational value and switching cost of alternatives.

The past invoice should not own the child’s next six months.

Sunk Cost and Reputation

Leaders often continue bad projects because stopping would admit that the original decision was wrong.

Now reputation becomes part of the payoff.

If cancelling damages the leader personally, continuation can be individually rational even when it is organisationally destructive.

This is a governance problem.

The system must make correction cheaper than denial.

Separate the Person From the Project

A project should be allowed to fail without automatically making its sponsor a failure.

Otherwise the sponsor has an incentive to keep the project alive.

Good organisations reward accurate stopping as well as successful completion.

“We stopped because the evidence changed” should sometimes be evidence of competent management.

Kill Criteria

Before starting a project, define conditions under which it should stop.

If cost exceeds X.
If demand remains below Y after Z months.
If technical milestone A fails twice.
If regulatory approval is not obtained by date B.
If expected future return falls below the alternative investment threshold.

Precommitted kill criteria are useful because they are written before the sunk cost becomes psychologically powerful.

They convert future stopping from an emotional admission into a previously agreed governance rule.

Stage-Gate Funding

Instead of funding a five-year project all at once, divide it into stages.

After each stage, reassess.

Has the evidence improved?

Have assumptions changed?

Does the next tranche still beat alternatives?

Stage gates make continuation a fresh decision rather than the automatic default.

This is mechanism design against escalation.

See How The World Works | Mechanism Design.

Independent Review

The people who designed the project know the most about it.

They also have the strongest psychological and reputational investment in it.

Independent review introduces a decision-maker who did not pay the original psychological cost.

The reviewer can ask the replacement question more cleanly.

Would we start this today?

The New-CEO Effect

New leaders often cancel projects old leaders protected.

Sometimes that is politics.

Sometimes it is rational reset.

The new leader did not sponsor the original decision and therefore carries less self-justification pressure.

Fresh eyes can reduce sunk-cost attachment.

They can also destroy useful continuity if cancellation becomes symbolic rather than analytical.

The same forward-looking test applies.

Sunk Costs in Public Infrastructure

Large public projects are difficult because abandoning them can be politically visible.

A half-built station.

An unfinished highway.

A cancelled procurement.

Citizens can point to the physical evidence of past spending.

That visibility makes stopping politically costly.

But continuing an uneconomic project only hides the original mistake inside a larger bill.

Public governance needs explicit reappraisal gates precisely because political sunk-cost pressure can be enormous.

When Cancellation Costs Are Real

Stopping can have future costs.

contract penalties;
staff redundancy costs;
reputational damage;
supplier claims;
environmental restoration;
transition costs;
lost network compatibility;
customer disruption.

These are not sunk if the current decision determines whether they occur.

Count them.

Sunk-cost discipline does not mean pretending exit is free.

It means separating true exit costs from dead historical spending.

Reputation Can Also Have Future Value

“We promised to finish” is not automatically a sunk-cost argument.

Breaking a commitment can damage future trust.

That future reputational consequence is relevant.

The analytical task is to price it honestly rather than hide it inside “we have come too far to stop.”

See How Trust Works.

Sunk Costs and Bargaining

Specific sunk investment can weaken an outside option.

A supplier builds custom tooling for one buyer.

The money already spent is sunk.

The fact that the equipment has little value elsewhere changes the supplier’s current bargaining position.

This is where sunk investment connects to the Hold-Up Problem.

The old spending is not directly relevant because it is past.

The weak outside option created by the specialised asset is highly relevant because it affects the future.

See How The World Works | The Hold-Up Problem and How The World Works | Bargaining.

Sunk Costs and Default Continuation

Many projects continue because nobody makes a new decision.

Budget renews.

Staff remain assigned.

Meetings continue.

The project’s existence becomes the default.

Sunk costs make the default emotionally easier to defend.

A useful governance design reverses the burden:

Continuation requires a fresh case.

See How The World Works | Defaults.

The Portfolio Rule

When evaluating ten projects, do not ask which ones have consumed the most money.

Ask where the next dollar produces the most expected value.

A new project may deserve funding more than an old project with millions already invested.

Capital allocation is a competition among future returns, not a ceremony honouring past expenditure.

Marginal Analysis Is the Correct Forward Lens

What happens if we spend one more dollar?

One more month?

One more hour?

What future benefit does that incremental commitment produce?

This is why marginal analysis is the natural partner to sunk-cost discipline.

See How The World Works | Marginal Analysis.

The Sunk-Cost Audit

State the current decision. Continue, stop, pause, sell, switch or redesign?
List all past costs. Money, time, effort, reputation, political capital.
Mark what is unrecoverable. Those are sunk.
List surviving assets. Skills, equipment, data, resale value, contracts, intellectual property.
List future exit costs. Penalties, transition, restoration, redundancy, trust damage.
List future continuation costs. Money, time, risk, attention and opportunity cost.
Estimate future continuation benefits. Use current evidence, not the original business case.
Compare with the best alternative. What could the next resources do elsewhere?
Run the replacement test. Would you start or buy this project today at the price of its remaining costs?
Check planning fallacy. Is “nearly finished” based on realistic remaining work?
Check confirmation bias. Are bad signals receiving tougher scrutiny than good ones?
Check loss aversion. Does stopping feel like making the loss real?
Check reputation incentives. Who personally loses status if the project stops?
Check default continuation. Is the project continuing because nobody must reapprove it?
Use independent review for large commitments.
Set kill criteria before the next tranche.

When the Sunk-Cost Lens Fails

The lens fails when people say “sunk cost” to dismiss every reason connected to the past.

Past investment can create present assets, capabilities, liabilities and reputational consequences.

Those matter.

It fails when stopping is assumed to be costless.

Exit can have real future costs.

It fails when continuing a good project is called irrational merely because much has already been invested.

The correct question is still future value versus future cost.

And it fails when sunk-cost language is used to avoid learning from the original mistake.

The past should not control the current allocation.

It should absolutely inform the post-mortem.

A Better Question Than “How Can We Walk Away After Spending So Much?”

Ask:

If the past spending vanished from the story but the current assets, liabilities, risks and opportunities remained exactly the same, what would we choose now?

How Sunk Costs Connect to the Rest of the World

Opportunity cost: the future alternative forgone is relevant; the unrecoverable past cost is not.
Marginal analysis: compare the next unit of cost with the next unit of benefit.
Loss aversion: stopping can make the past loss feel psychologically sharper.
Confirmation bias: invested decision-makers can protect the project by filtering evidence.
Planning fallacy: remaining costs are often underestimated, making continuation look artificially attractive.
Irreversibility: future irreversible consequences matter even though past spending is sunk.
Defaults: continuation can become automatic unless projects face fresh approval.
Bargaining: specific sunk investments can weaken outside options.
Hold-up: relationship-specific investment changes future bargaining power.
Mechanism design: kill criteria, stage gates and independent review can reduce escalation.
Trust: cancellation can create real future reputational costs that should be counted explicitly.
Risk: remaining uncertainty belongs in the forward calculation.

Frequently Asked Questions

What is a sunk cost?

A sunk cost is a past cost that has already been incurred and cannot be recovered or changed by the current decision.

What is the sunk-cost fallacy?

It is the tendency to continue an activity partly because money, time or effort has already been invested, even when forward-looking costs and benefits favour stopping or switching.

Should sunk costs always be ignored?

The unrecoverable expenditure itself should normally be irrelevant to the current choice. But present assets, capabilities, liabilities, contractual penalties, resale value, learning and future reputational consequences created by the past can be highly relevant.

How is sunk cost different from opportunity cost?

Sunk cost is what has already been spent and cannot be recovered. Opportunity cost is the value of the best alternative you give up by choosing what to do next.

Is escalation of commitment the same as sunk-cost bias?

No. Escalation of commitment is broader. Sunk costs can contribute, but reputation, self-justification, incentives, uncertainty, politics and organisational identity can also drive escalation.

Research Basis and Further Reading

OpenStax, “How Individuals Make Choices Based on Their Budget Constraint”, for the forward-looking economics of sunk costs and opportunity cost.
OpenStax, “Identify Relevant Information for Decision-Making”, for managerial treatment of sunk costs as decision-irrelevant historical expenditures.
Hal R. Arkes and Catherine Blumer, “The Psychology of Sunk Cost”, the classic 1985 experimental and field study of increased continuation after investments of money, effort or time.

The Larger Idea

The past deserves respect.

It deserves accounting.

It deserves learning.

It does not deserve a permanent vote.

Money already spent cannot be unspent by spending more.

Time already used cannot be recovered by using more time badly.

A wrong decision does not become right because reversing it would make the error visible.

The disciplined decision-maker keeps the lesson and releases the expenditure.

Sunk-cost thinking asks the future to rescue the past. Better judgment asks the future what it is worth on its own terms.

How The World Works | Sunk Costs — Why Money Already Spent Can Keep Bad Decisions Alive