Escalation Of Commitment: Why Businesses Keep Funding Projects They Should Have Killed

Ask a room of business owners whether past spending should influence a decision about future spending, and every hand goes up for "no." Ask the same room whether they have ever continued funding something primarily because of how much had already gone into it, and the hands go up again. Both answers are honest. The sunk cost principle is not hard to understand; it is hard to obey, and understanding it provides remarkably little protection.

The Principle, And Why It Fails

The normative rule is unambiguous. A rational decision-maker evaluating whether to continue a project should consider only incremental future costs and incremental future benefits. Money already spent is gone; it cannot be recovered by any choice available now, and it therefore carries no information about whether the next dollar is well spent. The only relevant question is whether the forward-looking return on the next increment of investment exceeds its opportunity cost.

Human beings do not reason this way. Arkes and Blumer's foundational 1985 work demonstrated the effect across a series of experiments, including the well-known ski trip scenario in which participants who had paid more for a less-attractive trip chose to take it over a more-attractive cheaper one, purely because the non-refundable outlay was larger[1]. The participants were not confused about which trip they would enjoy more. They chose the worse experience to avoid the feeling of waste.

That last clause is the mechanism. The sunk cost fallacy is not primarily an error of arithmetic. It is an aversion to the psychological experience of having wasted something, and continuing to invest defers that experience indefinitely. Abandonment forces the waste to be recognized, in the accounting sense and the emotional one, on a specific date, by a specific person.

What The Research Actually Shows

Barry Staw's 1976 study introduced the term "escalation of commitment" to describe the organizational version, and found something that should trouble anyone designing a governance process: participants who had personally made an initial resource allocation decision that turned out badly subsequently allocated more to that same failing course than participants who inherited the decision from someone else[2]. Responsibility for the original choice increased, rather than decreased, willingness to escalate.

This finding has been replicated and extended extensively over five decades. Subsequent work established that escalation intensifies when the decision-maker's own competence is publicly implicated, when the project has an identifiable champion, and when information about performance is ambiguous enough to permit optimistic interpretation. Meta-analytic reviews have generally supported the robustness of the core effect while also documenting substantial variation in its magnitude across contexts[3].

Two practical implications follow directly. First, the person best positioned to evaluate whether a project should continue, the one who knows it most intimately, is also the person most susceptible to escalation, precisely because of that involvement. Second, no amount of analytical sophistication substitutes for structural separation, because the bias operates on the interpretation of evidence rather than on the arithmetic performed with it.

The Four Drivers Of Escalation

It is useful to separate the distinct forces, because different countermeasures address different drivers.

Loss aversion and waste avoidance. The prospect-theory finding that losses register roughly twice as intensely as equivalent gains means that crystallizing a write-off is disproportionately painful relative to the accounting reality[4]. Continuing keeps the loss unrealized and therefore, psychologically, provisional.

Self-justification. Staw's core insight: abandoning implies the original decision was wrong, which threatens the decision-maker's self-concept and public standing. Escalation defends the earlier judgment rather than the current investment[2].

Optimistic reinterpretation of ambiguous data. Failing projects rarely produce unambiguous failure signals. They produce mixed signals, and a committed champion reads them favourably. This is motivated reasoning rather than deception, and it is largely invisible from the inside.

Completion bias. Partially finished things exert a pull disproportionate to their value. A half-built system, a partially renovated location, a nearly finished custom development all generate pressure to finish that is about closure rather than return.

Canadian Mechanics That Make It Worse

Several features of the Canadian operating environment amplify escalation, and they are worth naming because they are frequently mistaken for legitimate financial reasoning.

Capital cost allowance psychology. Owners frequently resist disposing of underperforming capital assets on the grounds that they have "not finished depreciating it." This confuses tax mechanics with economics. The undepreciated capital cost of an asset is not a reason to keep operating it; on disposition, the tax system generally deals with the remaining balance through terminal loss or recapture rules. Whether continuing to run an unprofitable asset makes sense depends on forward cash flows, not on how much CCA remains unclaimed. But the language of "we haven't written it off yet" makes retention feel financially disciplined when it may be the opposite.

Personal guarantees. Most Canadian small business lending involves a personal guarantee from the owner. This creates a genuine, non-psychological reason for owners to resist any action that crystallizes a loss, because the loss may reach their personal balance sheet directly. The complication is that a real constraint and a behavioural bias here point in the same direction, which makes them extremely difficult to disentangle. An owner refusing to close a failing division may be escalating irrationally, or may be rationally avoiding a covenant breach. Both look identical from outside.

The absence of independent governance. Most Canadian private companies have no independent board. Staw's finding implies that the single most effective structural defence, separating the continue/abandon decision from the person who made the original commitment, is structurally unavailable in the majority of the businesses that need it most.

Grant and credit program entanglement. Projects partly funded by SR&ED claims, provincial grants, or program financing acquire an additional layer of commitment. Owners become reluctant to abandon work that generated favourable tax treatment, reasoning, incorrectly, that abandoning somehow forfeits value already claimed.

A Worked Case: The Second Location

A service business opens a second location. Eighteen months in, it has consumed roughly $310,000 in leaseholds, equipment, and accumulated operating losses. It is currently losing about $6,000 a month. The lease has 30 months remaining at $4,200 monthly, with a personal guarantee.

The escalation framing: "We've put $310,000 into this. We can't walk away now, we'd lose everything we've invested." The forward-looking framing asks an entirely different question: over the next 30 months, what does continuing cost relative to closing?

Continuing costs roughly $6,000 monthly in operating losses, about $180,000 over the remaining term, assuming performance does not change. Closing costs the remaining lease obligation of roughly $126,000 unless the space can be sublet or the landlord will negotiate, plus any equipment disposition loss and wind-down costs. The $310,000 appears in neither calculation, because it is unrecoverable under either choice.

On these figures, closing is roughly $54,000 better before considering sublet potential, redeployment of management attention, or the option value of not being locked in. The number changes entirely if there is a credible path to breakeven, and that is exactly the question that deserves scrutiny. But note what the escalation framing did: it made $310,000, the one figure that is irrelevant, the dominant number in the conversation, and left the two figures that actually determine the answer unexamined for eighteen months.

These figures are illustrative and the arithmetic in any real case will turn on lease terms, sublet markets, guarantee exposure, and tax treatment of the disposition. The structural point is what transfers: the decision changed the moment the irrelevant number was removed from the table.

The Mirror Image: Inaction Inertia

Escalation has a less-discussed opposite that afflicts the same businesses, and treating one without the other produces lopsided decisions. Inaction inertia describes the tendency to forgo a currently attractive opportunity because a better version was previously missed[6]. An owner who declined to buy a competitor at four times earnings two years ago frequently refuses at five times today, not because five times is unattractive on current facts, but because it is worse than a reference point that no longer exists.

The structural similarity is exact. In escalation, an irrelevant past number, money spent, dominates a forward decision. In inaction inertia, an irrelevant past number, a price no longer available, does the same. Both replace the question "is this a good use of capital now?" with a comparison to a historical figure that carries no information about present value.

Owners tend to exhibit both, in different domains, and often simultaneously: escalating on projects they started while refusing opportunities they passed on. The common remedy is the same discipline, evaluating each decision against current forward economics and a stated hurdle rate rather than against history.

The Cost Nobody Books: Management Attention

Financial analysis of continuation decisions almost always understates the true cost, because the largest input is rarely measured. A struggling project consumes management attention out of all proportion to its size. The second location losing $6,000 a month in the earlier example was, on the owner's own estimate, absorbing something closer to 40% of his working time, against roughly 15% of revenue.

That asymmetry is characteristic rather than exceptional. Problems are attention-intensive; functioning operations are not. The consequence is that the genuine cost of continuing includes whatever the owner's displaced attention would have produced applied to the core business, a figure that never appears in a variance report and is therefore excluded from exactly the analysis where it matters most.

There is no precise way to quantify this, and inventing a number risks false precision. But it can be surfaced qualitatively and deliberately: asking, at each review, roughly what share of senior attention the project is consuming relative to its share of revenue or profit. Where that ratio is badly skewed and has been for several periods, the financial case for continuation is being calculated on an incomplete cost base, and usually a substantially understated one.

The Linguistic Tells

Escalation announces itself in language before it appears in numbers. Certain phrases reliably indicate that sunk costs are driving the discussion rather than forward economics:

• "We're too far in to stop now." Distance travelled is not evidence about distance remaining.
• "We just need one more push." Particularly when this is not the first such push, and when "one more" is undefined in size or duration.
• "If we stop now, all of that was wasted." The waste already occurred. Stopping recognizes it; continuing adds to it.
• "It's finally starting to turn." Worth examining against the specific metric and threshold defined in advance, if one was defined at all.
• "We can't let it fail after everything the team put in." Real and humane, and a reason to handle the wind-down well, not a reason to fund it further.

Why De-Escalation Is So Rare

Organizations rarely kill projects cleanly, and the reasons are as much social as analytical. Abandonment requires someone to say, on the record, that a previously approved course of action should stop, which in most private companies means contradicting the owner. It requires accepting a visible write-off in a specific period rather than an invisible drift across many. And it typically has no champion: every project has someone whose role, reputation, or enthusiasm is tied to its continuation, while the case for stopping is usually made by nobody in particular.

This asymmetry is structural, not cultural. It is why explicit kill criteria set in advance work so much better than case-by-case judgment: they create a designated moment at which stopping is the default rather than an act of dissent, and they assign the decision to a rule rather than to a person who must absorb the social cost of making it.

The Governance That Actually Works

The interventions supported by the research are procedural, and they share a common feature: they operate on the decision structure rather than on the decision-maker's reasoning.

Pre-commit kill criteria before funding. At approval, specify what performance by what date would mean the project stops, in writing, with numbers. This is the single highest-leverage intervention, because it is made before anyone is committed and before the ambiguity that enables motivated reinterpretation exists.

Separate the reviewer from the approver. Staw's finding argues directly for this. Where an independent board is unavailable, the substitute is an external reviewer, a fractional CFO, an advisory board, or a peer group, whose standing does not depend on the original decision.

Reframe every review as a fresh decision. The operative question is never "should we continue?" but "if we had this opportunity today, at this stage, with this track record, and none of the money already spent, would we invest the remaining amount?" This reframing is simple, and reliably changes answers.

Track the counterfactual explicitly. Requiring that every continuation decision name what else the incremental capital could fund makes opportunity cost concrete rather than abstract. Escalation thrives when the alternative is unnamed.

Separate the person from the project in how outcomes are discussed. Where killing a project is treated as a personal failure, escalation is rational self-protection. Organizations that treat well-run terminations as evidence of good judgment get earlier, cheaper stops.

When Persistence Is Correct

The most common misapplication of sunk cost reasoning is using it to justify abandoning things prematurely. Not every struggling project is an escalation trap, and the fallacy label is sometimes deployed as a rhetorical device by people who simply want to stop.

Persistence is defensible when there is a specific, testable mechanism by which performance improves, not merely hope; when the project is genuinely early relative to its expected maturation curve, and that curve was articulated in advance rather than retrofitted; when the incremental investment required is small relative to a credible remaining upside; and when the original thesis remains intact and the shortfall traces to execution issues that have been identified and addressed.

The distinguishing question is whether the case for continuing rests on evidence that would persuade someone with no history with the project. If the argument requires the listener to already care about what has been invested, it is escalation. If it stands on forward evidence alone, it is judgment.

The Owner-Managed Complication

Everything above is harder in an owner-managed business, and it is worth being direct about why. In a large organization, escalation is a governance problem with governance solutions. In an owner-operated company, the person who conceived the project, approved the capital, championed it internally, guaranteed the debt, and must now decide whether to kill it, is one person. Every driver of escalation, self-justification, loss aversion, completion bias, motivated reasoning, concentrates in a single individual with no structural counterweight.

Additionally, the owner's identity is often bound up in the venture in ways that have no analogue in corporate settings. A failed division at a large firm is a line item. A failed second location for an owner-operator is frequently experienced as a personal verdict, discussed in their community, visible to their employees and family. The rational case for closing may be clear and still be genuinely hard to act on, and treating that difficulty as mere irrationality misunderstands the situation.

This is the strongest practical argument for external input in owner-managed businesses: not because owners lack analytical capability, but because the structural separation that makes de-escalation possible cannot be generated from inside a single head.

A Note For Students

For anyone studying finance or management, escalation of commitment is worth understanding as more than an exam item, because it illustrates a general and uncomfortable point about the relationship between knowledge and behaviour. The sunk cost principle is taught in essentially every introductory finance course. It is among the simplest normative rules in the discipline. And decades of evidence show that teaching it produces remarkably little behavioural protection, including among people who can state it flawlessly.

The lesson generalizes. A substantial portion of professional value in advisory work comes not from knowing rules that clients do not know, most owners understand sunk costs perfectly well, but from occupying a structural position that makes acting on those rules possible. The advisor's advantage in a continuation decision is frequently not superior analysis. It is simply not having been the person who approved the original spend.

The Limits Of This Analysis

Several caveats deserve stating. The foundational experimental work was largely conducted in laboratory settings with students and modest stakes, and the extension to high-stakes business decisions, while widely accepted and supported by field studies, is an inference rather than a direct measurement. Some scholars have argued that apparent escalation in real projects is better explained by genuine uncertainty, complexity, and information asymmetry than by bias, a critique that has been raised forcefully in the megaproject literature[5].

There is also a real risk of hindsight bias in diagnosing escalation. Projects that eventually succeeded after difficult periods look like admirable persistence; identical decisions that failed look like escalation. The label is much easier to apply after the outcome is known, which means it should be applied cautiously to decisions still in progress, and applied to the process rather than the outcome wherever possible.

Finally, the countermeasures described here have their own costs. Rigid kill criteria can terminate projects that a more flexible reading would have saved. External reviewers lack context. The goal is better-calibrated decisions, not mechanical ones.

Related Reading

Behavioural Finance The Corporation Is Not A Piggy Bank: Mental Accounting And The Canadian Owner-Operator The companion bias: why economically identical dollars get treated as entirely different money. Fractional CFO & Capital Allocation WACC In Volatile Rate Environments The hurdle rate that continuation decisions should actually be tested against. Fractional CFO & Exit Planning EBITDA Optimization Before An Exit Why underperforming divisions cost multiples, not just their own losses, at sale.

Frequently Asked Questions

What exactly is the difference between the sunk cost fallacy and escalation of commitment?

The sunk cost fallacy is the individual-level error of letting unrecoverable past spending influence a forward-looking decision. Escalation of commitment, a term introduced by Barry Staw in 1976, describes the organizational pattern in which decision-makers increase investment in a failing course of action, and is amplified specifically when the decision-maker was responsible for the original commitment.

Doesn't remaining CCA give me a financial reason to keep an underperforming asset?

Generally no. Undepreciated capital cost reflects tax treatment of past expenditure, and the tax system deals with the remaining balance on disposition through terminal loss or recapture rules. Whether to keep operating an asset depends on its forward cash flows relative to alternatives. The specific tax outcome on disposition is fact-dependent and worth confirming with your accountant, but "we haven't finished depreciating it" is not by itself a reason to continue.

How do I set kill criteria without being unfair to a project that just needs time?

Set them at approval, before anyone is committed, and tie them to the project's own articulated thesis: what performance, by what date, would indicate the thesis is working. Criteria set in advance against a stated maturation curve are far fairer than judgments made later under pressure, precisely because they are defined before motivated reasoning has anything to work with.

Is every struggling project an example of escalation?

No, and treating it that way is its own error. Persistence is defensible where there is a specific testable mechanism for improvement, the project is genuinely early against a curve defined in advance, the incremental cost is modest relative to credible upside, and the original thesis remains intact. The test is whether the argument for continuing would persuade someone with no history with the project.

What can an owner-operator do without an independent board?

The practical substitutes are an external reviewer whose standing does not depend on the original decision, a fractional CFO, an advisory board, or a structured peer group, combined with written kill criteria set at approval. The point is structural separation, which cannot be generated from within a single decision-maker regardless of their analytical ability.

Does knowing about the sunk cost fallacy protect me from it?

Only weakly. The effect has proven robust to training and explicit warning across decades of research, because it operates on how ambiguous evidence is interpreted rather than on the arithmetic performed with it. Procedural defences consistently outperform awareness.

References

1. Arkes, H. R., & Blumer, C. (1985). The Psychology of Sunk Cost. Organizational Behavior and Human Decision Processes, 35(1), 124-140.
2. Staw, B. M. (1976). Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action. Organizational Behavior and Human Performance, 16(1), 27-44.
3. Sleesman, D. J., Conlon, D. E., McNamara, G., & Miles, J. E. (2012). Cleaning Up the Big Muddy: A Meta-Analytic Review of the Determinants of Escalation of Commitment. Academy of Management Journal, 55(3), 541-562.
4. Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk. Econometrica, 47(2), 263-291.
5. Love, P. E. D., Ika, L. A., & Ahiaga-Dagbui, D. D. (2019). On de-bunking "fake news" in the post-truth era: Why does the Planning Fallacy explanation for project overruns fall short? Transportation Research Part A, 126, 397-408.
6. Tykocinski, O. E., Pittman, T. S., & Tuttle, E. M. (1995). Inaction Inertia: Foregoing Future Benefits as a Result of an Initial Failure to Act. Journal of Personality and Social Psychology, 68(5), 793-803.
7. Lovallo, D., & Kahneman, D. (2003). Delusions of Success: How Optimism Undermines Executives' Decisions. Harvard Business Review, 81(7), 56-63.

This article discusses general behavioural patterns documented in academic research and is provided for informational purposes only. The worked example uses illustrative figures and is not a template for any specific decision. Capital allocation, asset disposition, and lease termination decisions carry tax and legal consequences that are highly fact-specific; confirm your own position with qualified advisors.