(Behavioural Science) #15 Sunk Cost Fallacy

 

Principle · Loss / Risk category

Sunk cost fallacy

The tendency to continue investing time, money, or effort into a failing course of action because of what has already been spent — rather than evaluating the decision purely on future costs and future benefits. Sunk costs are irrecoverable; rational decision-making should ignore them entirely. Humans systematically cannot.

~50%

of people escalate commitment even after clear failure signals

$billions

lost annually in corporate escalation-of-commitment decisions

1985

Arkes & Blumer landmark paper formally naming the fallacy

Robust

one of the most replicated findings in behavioral economics

1. What it is and the science behind it

A sunk cost is any past expenditure of money, time, effort, or emotion that cannot be recovered regardless of what you decide next. The economically rational rule is simple: sunk costs are irrelevant to forward-looking decisions. Only future costs and future benefits should matter. And yet, across hundreds of studies and every domain of life, people consistently violate this rule — dragging past investment into decisions about the future.

The fallacy is driven by a cluster of overlapping psychological mechanisms, not a single cause. Disentangling them matters because different mechanisms suggest different counter-designs.

Why it happens — four mechanisms

Loss aversionCutting losses means realizing a loss — which feels roughly twice as painful as the equivalent gain. Continuing to invest delays that moment of painful recognition, even at greater total cost.

Waste aversionPeople have a deep aversion to "wasting" resources. Abandoning a half-finished project feels like making the prior investment meaningless, even though it already is — the project's status cannot retroactively change what was spent.

Ego protectionContinuing signals confidence in the original decision. Quitting signals it was a mistake, threatening self-image and social reputation. Escalation of commitment is partly motivated reasoning — convincing yourself the decision was right all along.

Completion pressureThe Zeigarnik effect: unfinished tasks create cognitive tension. Abandoning something 80% complete feels worse than abandoning something 20% complete, even when the remaining effort and value are identical.

Key studies

The theater ticket experiment

Arkes & Blumer, 1985

Participants were told they had bought a $100 ski trip ticket, then later discovered a $50 trip to a better resort on the same day. The majority chose the worse trip — because they had paid more for it. In a variation, people who received the more expensive ticket as a gift showed no such preference. The sunk cost effect was triggered purely by financial outlay, not by the quality of the option. This paper is the foundational empirical work establishing the effect in controlled settings.

Majority chose worse option when they had paid more

Escalation of commitment in organizations

Staw, 1976 — the "knee deep in the Big Muddy" study

MBA students acting as financial decision-makers were given a failing investment scenario. Those who had personally made the initial investment decision (high personal responsibility) allocated significantly more additional funds than those who inherited the decision from another manager — even with identical financial information in front of both groups. The effect was amplified when outcomes were negative and when the decision-maker was publicly identified. Responsibility for the initial choice is a key amplifier of the fallacy.

Personal responsibility dramatically increases escalation

Concorde fallacy in animals

Dawkins & Carlisle, 1976; replicated across species

Named after the Franco-British supersonic aircraft project — continued long after it was clear it would never be commercially viable, largely because of national pride in prior investment. Behavioral ecologists found the same pattern in non-human animals: female digger wasps continue investing in nests in proportion to prior effort rather than future reproductive value. This suggests the sunk cost bias has deep evolutionary roots, not just cultural ones — possibly because in natural environments, effort expenditure was a reliable proxy for future value.

Observed in animals — likely an evolutionary artifact

NBA salary and playing time

Staw & Hoang, 1995

Analyzing NBA rosters over multiple seasons, coaches gave more playing time to highly-drafted players — even controlling for actual performance — especially in early career seasons. Higher draft picks (representing larger organizational investment) were kept on rosters longer and benched less often despite equivalent or inferior play. The effect was robust across coaches and teams, suggesting institutional sunk cost bias is not eliminated by professional expertise or competitive market pressure.

Draft investment predicted playing time beyond performance

2. Real application examples

Business

Product development escalation

Engineering teams continue building features that early user testing has already invalidated — "we've already spent six months on this." Product managers who inherit a project from another team are significantly more willing to kill it than the original team.

Business

M&A integration failures

Companies overpay for acquisitions and then continue pouring resources into integrations that are clearly not generating value — because admitting failure would mean writing down the acquisition price. The larger the premium paid, the stronger the escalation tendency.

Business

Advertising spend inertia

Marketing teams continue running campaigns that analytics show are underperforming because the creative assets were expensive to produce. Prior production cost — a sunk cost — influences media spend decisions that should only be based on current ROI data.

Public policy

Infrastructure megaprojects

The Concorde, the Big Dig, the Channel Tunnel — all continued past the point of rational investment because stopping felt like "wasting" prior public expenditure. Bent Flyvbjerg's research shows megaprojects average 44% cost overrun in real terms, driven partly by political sunk-cost entrenchment.

Public policy

Military escalation

"We can't leave now — too many have died" is a textbook sunk cost argument. Casualty counts — irrecoverable losses — become political arguments for continued engagement even when strategic goals are no longer achievable. McNamara later acknowledged this pattern explicitly in Vietnam.

Public policy

Welfare program continuation

Programs that have been running for decades receive continued funding even when outcome evaluations are poor — because of institutional inertia and the political cost of admitting prior spending was ineffective. Sunsetting clauses are one structural counter-design.

Personal habit

Staying in bad relationships

"We've been together for 7 years — I can't just walk away." The years invested are gone regardless of the decision made today. The relevant question is only: does continuing this relationship produce a better future than ending it? Sunk time makes that calculation nearly impossible to make clearly.

Personal habit

Finishing bad meals / books / films

People finish films they are not enjoying because they have already watched half of it. They finish food on their plate because they paid for it. Each is a small, daily sunk cost error — the calories, the time, the experience cannot be recovered by continuing.

Personal habit

Holding losing investments

Retail investors famously hold losing stocks far too long — the disposition effect. Selling feels like "making the loss real." The original purchase price is irrelevant to whether the stock will recover; only future expected value matters. Yet the purchase price dominates individual investor decisions.

3. Design guidance — when and how to use it

Unlike most principles in this series, the sunk cost fallacy is primarily something to design against — in yourself, in your organization, and in the systems you build. But it can also be deliberately leveraged to drive positive behaviors. Both directions are covered below.

The two design modes

Counter-design

Removing the bias from decisions

For organizations making investment, product, or policy decisions — build processes that structurally separate the future decision from the past expenditure. The goal is to make it cognitively easier to cut losses when cutting losses is the right call.

Leverage design

Harnessing it for positive behavior

For consumer products, health programs, and habit design — use upfront investment to create commitment and follow-through. Getting people to invest early (money, effort, identity) increases the probability they continue with a positive behavior.

When counter-design is needed

Use counter-design when

Decisions involve large prior expenditures and ongoing investment options — M&A, product development, infrastructure, hiring.

Use counter-design when

The original decision-maker is still in place. Personal responsibility for the initial choice is the strongest amplifier of escalation.

Use counter-design when

The organization has a pattern of cost overruns or "zombie projects" — initiatives that limp on with minimal resources long past usefulness.

Be cautious when

Applying the rational-actor model too rigidly — sometimes "sunk costs" carry legitimate information (e.g., prior learning) that genuinely informs future decisions.

Step-by-step counter-design process

1. Reframe every continuation decision as a fresh start — ask "If we hadn't already invested in this, would we start today given what we now know?" If the answer is no, the prior investment is doing the work — not the future case.
2. Assign a "project devil's advocate" with explicit authority to argue for cancellation. Research shows that separating the role of evaluator from the role of original champion significantly reduces escalation. This person's job is to make the case for cutting losses, regardless of their personal view.
3. Pre-mortem before major continuation decisions — ask the team to imagine the project has failed and write down why. Activating failure scenarios in advance reduces motivated reasoning and makes exit easier to contemplate.
4. Use rotating decision-makers for continuation reviews. Staw's research shows that decision-makers who did not make the original investment are dramatically less prone to escalation. Build this into governance cycles for large projects.
5. Set explicit kill criteria at launch — decide in advance what evidence would cause you to stop. "If we don't reach X by date Y, we will discontinue" removes the continuation decision from the moment of high emotional investment in prior work.
6. Celebrate intelligent exits publicly. Organizations that stigmatize project cancellation create cultural sunk cost traps. If stopping a failing project is punished socially, no structural process will overcome it. Leaders who visibly and positively frame cancellations as good decisions break the cultural driver of escalation.

Leveraging it for positive behavior change

The same mechanism that keeps people in bad situations can keep them in good ones. A workout you paid for is more likely to be attended than a free one. A course you publicly enrolled in is more likely to be completed. Strategic early investment — of money, time, public commitment, or identity — creates a sunk cost that works in the desired direction.

This is why onboarding flows that ask users to set up a profile, answer questions, or make a small investment significantly increase long-term retention. The user's prior effort becomes a reason to stay — the sunk cost effect working as a retention mechanism.

Before and after — decision framing

Project continuation review — organizational

Weak (sunk-cost-contaminated)

"We've spent $4M and 18 months on this — we need to see it through to get a return on that investment."

Strong (forward-looking)

"Forgetting what we've spent so far: given what we know today, would we commit $1.5M and 6 months to complete this? What's the expected return from here?"

Personal investment decision

Weak

"I've already put so much into this stock — I can't sell at a loss."

Strong

"If I didn't already own this stock and had this cash today, would I buy it at the current price? If not, why am I holding it?"

Health program — leveraging sunk cost for adherence

Weak onboarding

"Sign up for free and try our app. No commitment needed."

Strong onboarding

"Before you begin, let's set up your health profile — it takes 8 minutes and personalizes everything. People who complete setup are 4× more likely to reach their goal."

Critical nuance — sunk costs are not always irrational

The normative rule "ignore sunk costs entirely" is correct in pure financial decision theory, but real-world sunk costs sometimes carry legitimate signal. Prior investment often represents learning, relationship-building, or strategic positioning that genuinely has forward value — just not the value of "recovering" the cash. The practical skill is separating the legitimate forward value of past work (expertise gained, partnerships formed, infrastructure built) from the illegitimate motivator (I don't want to feel like I wasted money). The reframe is: "What real assets did that investment create that I can deploy going forward?" — not "How do I recoup what I spent?"