Investor psychology Lesson 4 of 15
Articles · Behavioral finance
Sunk cost fallacy and averaging down: when to stop adding
Why money already lost keeps investors in a position, what the sunk cost experiments found and how well they replicate, and what averaging down did in one real stock.
On this page
A sunk cost is money, time or effort already spent that no decision can bring back. The sunk cost fallacy is letting it decide what you do next: staying in a position because of what you have put in, not because of what you expect to get out.
In investing it has a characteristic move, averaging down, which means buying more of a holding that has fallen in order to lower the average price paid. Averaging down is not always a mistake, and this lesson is careful about the difference. It covers what the classic experiments found, how well they have held up, what averaging down does to the arithmetic of a position, and one real stock where the arithmetic played out over five years.
The principle
Economics is unusually clear on this point. A decision can only affect the future, so only future costs and benefits belong in it. If you paid $10,000 for shares now worth $6,000, the $4,000 difference is the same in every branch of the decision tree. You have $6,000 of something, and the only live question is where $6,000 is best kept from today.
People do not decide this way, and Richard Thaler was among the first economists to treat that as a fact to be explained. His 1980 paper used the example of a family that paid for basketball tickets and then drives through a snowstorm to use them, though they would have stayed home had the tickets been free.
What the experiments found
Theater tickets
Hal Arkes and Catherine Blumer ran the best-known test in the 1982–83 season of the Ohio University Theater. The first 60 people who came to buy season tickets were randomly sold them at the normal price of $15, at a $2 discount or at a $7 discount. The tickets were color-coded, so attendance could be counted.
Over the first half of the season, those who had paid full price used 4.11 tickets on average. The two discount groups used 3.32 and 3.29. Everyone held the same tickets to the same plays. The only difference was what they had paid, which was gone either way. In the second half of the season the difference faded.
In a questionnaire study from the same paper, people imagined they had bought a $100 ski trip to Michigan and a $50 ski trip to Wisconsin, expected to enjoy Wisconsin more, and then found the trips were on the same weekend with no refunds. 54% (33 of 61) chose the Michigan trip, the one they expected to enjoy less, because it had cost more.
Escalation
Barry Staw's 1976 experiment put the fallacy in a business setting. Each of 240 business students played a financial executive who had to allocate $10 million of research funding to one of two divisions, and then, after seeing five years of results, divide a further $20 million between them. Some were told their first choice had worked out and some that it had failed. For half the students, the first choice had been made by a predecessor.
The students who had made the first choice themselves and been told it had failed put the most new money into the same division, $13.07 million of the $20 million on average. That was well above every other group; those who inherited a predecessor's choice put in about $9 million whether it had failed or not. Personal responsibility for the original decision was what drove the escalation. Staw called it "knee-deep in the big muddy".
How well it holds up
The sunk cost effect is one of the classic findings that has been retested at scale, and the verdict is "real, modest".
- The Many Labs project in 2014 ran a version in 36 laboratories with more than 6,000 participants: would you be more likely to go to a football game on a freezing day if you had paid for the ticket than if it had been free? The effect replicated across the pooled sample, with a small effect size (a standardized difference of about 0.3), well below that of the strongest effects in the project.
- Daniel Friedman and colleagues built a computer task with real money specifically to isolate the fallacy and reported a "surprisingly small" effect that did not respond to the psychological triggers proposed for it.
A fair summary is that sunk costs tilt decisions rather than dictate them, and that the tilt is strongest when the earlier decision was your own and visibly went wrong. That is exactly the situation of an investor looking at a falling stock they chose.
From the laboratory to the brokerage account
Terrance Odean's study of 10,000 brokerage accounts, better known for the disposition effect, also counted additional purchases. Investors bought more shares of 13.5% of the positions they held at a loss, against 9.4% of the positions they held at a gain. They were about 45% more likely to add to a loser than to a winner.
Buying more of a loser is where three biases meet. Loss aversion makes the loss painful to realize. The purchase price is the anchor that defines the loss. And the sunk cost supplies the reason to act: a lower average cost makes "getting back to even" look closer.
What averaging down does to the numbers
An investor buys 100 shares at $50, a $5,000 position. The stock falls to $30. She buys another 100 shares for $3,000.
| Before adding | After adding | |
|---|---|---|
| Shares | 100 | 200 |
| Total paid | $5,000 | $8,000 |
| Average cost | $50 | $40 |
| Price needed to break even | $50 (+67%) | $40 (+33%) |
| Loss if the stock falls to $20 | $3,000 | $4,000 |
| Value at risk | $3,000 | $6,000 |
The break-even price fell from $50 to $40, which is the part that feels like progress. The position doubled, which is the part that is easy to miss. Nothing about the company changed in this table. If the first purchase was a mistake about the company, the second purchase is the same mistake at twice the size.
Averaging down changes one number that matters for psychology (the distance to break-even) and one that matters for wealth (the dollars exposed). The fallacy is choosing by the first.
A real example: PayPal from 2021
PayPal (PYPL) closed June 2021 at $291.48, its highest month-end close. The figure follows an investor who buys $2,000 at that close and then, by a fixed rule, buys another $2,000 each time the month-end price is 25% or more below the previous purchase.
| Purchase | Month | Price | Average cost after | Total put in |
|---|---|---|---|---|
| 1 | Jun 2021 | $291.48 | $291.48 | $2,000 |
| 2 | Nov 2021 | $184.89 | $226.26 | $4,000 |
| 3 | Feb 2022 | $111.93 | $168.79 | $6,000 |
| 4 | Jun 2022 | $69.84 | $124.64 | $8,000 |
| 5 | Oct 2023 | $51.80 | $97.28 | $10,000 |
At the end of October 2026 the stock closed at $56.95, 80% below the starting price. The five purchases, $10,000 in total, were worth $5,854: a loss of $4,146, or 41%. The average cost had fallen to $97.28, and the stock still needed to rise 71% to reach it.
The comparison that matters: the first purchase alone was worth $391, a loss of $1,609. Averaging down made the percentage loss far smaller (41% against 80%) and the dollar loss two and a half times larger. Both statements are true, and which one an investor quotes says a lot about what they are protecting.
This is one stock chosen because it fell and did not come back, which is its own kind of selection. Many stocks that fall this far recover, and for those the same rule looks brilliant. That is the difficulty. On the day of each purchase, the investor in a stock that will recover and the investor in a stock that will not are looking at the same chart.
When is adding to a loser reasonable?
The fallacy is in the reason, not the action. Adding at a lower price is defensible when:
- The plan came first. Buying in stages was decided before the first purchase, with the total size fixed in advance. Scheduled investing in a broad index fund, dollar-cost averaging, is the common case.
- The holding is diversified. A fund of thousands of companies cannot go to zero the way one company can. A single stock can.
- The thesis has been re-examined, and still holds. Something specific explains the fall and leaves the reason for owning intact. "It is cheaper now" is a price, not a thesis.
- The total stays within a size limit. See position sizing. In the PayPal example there was no limit, and the position grew fivefold while the stock fell.
It is the fallacy when the reason is to lower the average cost, to get back to even sooner, or to avoid admitting that the first purchase was wrong.
What helps
No study shows a technique that removes the sunk cost effect, and the effect itself is small enough that claims of a cure should be doubted. Three practices follow from what the experiments found.
- The cash test. "If I held this position's value in cash today, would I buy this, and this much of it?" It removes the past payment from the question. It is the same test used against the disposition effect.
- Decide the maximum size before the first purchase. A ceiling in dollars or as a share of the portfolio turns "should I add?" into arithmetic.
- Let someone else, or an earlier you, make the call. In Staw's experiment, students who had inherited the losing decision did not escalate. A written rule made before the loss, or a second opinion from someone with no stake in the original choice, stands in for that predecessor.
How Finabase helps
A portfolio shows each position's average cost beside its cost and market value. The useful column for this lesson is market value: it is the amount you are deciding about, where average cost is only the record of the past. The PayPal chart on the 5Y or MAX range shows the fall used in the example. In paper trading you can add to a position and watch the average cost fall while the cost and market value rise, which is the table from this lesson in live form.
Frequently asked questions
What is the sunk cost fallacy in investing?
It is holding or adding to an investment because of what you have already paid or lost, when that amount cannot be recovered by any choice. The classic thought is "I'm down too much to sell now".
Is averaging down a good strategy?
It depends on why and into what. Staged buying of a diversified fund under a plan made in advance is a normal way to invest. Buying more of a single falling stock in order to lower the break-even price increases the money at risk without changing the company's prospects.
What is the difference between averaging down and dollar-cost averaging?
Dollar-cost averaging invests fixed amounts on a schedule whatever the price does. Averaging down buys more specifically because the price fell below what you paid. The first is driven by the calendar and the second by the loss.
How do I know if I am holding a stock because of sunk costs?
Ask whether you would buy it today, at today's price, with cash. If the honest answer is no, and you still do not want to sell, the purchase price is doing the deciding.
Summary
A sunk cost cannot be changed by any decision, yet it shapes many. Theater-goers who paid full price attended more plays, students escalated most when the failed decision was their own, and large replications confirm a real but small effect. In brokerage data investors added to 13.5% of their losers against 9.4% of their winners. Averaging down lowers the break-even price and enlarges the position: in the PayPal example it shrank the percentage loss from 80% to 41% and grew the dollar loss from $1,609 to $4,146. Adding to a position is reasonable under a plan, a size limit and a thesis that has been re-examined; it is the fallacy when the aim is to get back to even.
Check your understanding
5 questions on this lesson. Each answer comes with the reason, and nothing is scored.
An investor paid $10,000 for shares now worth $6,000. Which amount is the subject of the decision to hold or sell?
- The $10,000 paid
- The $4,000 lost
- The $6,000 the position is worth today
Show the answer
C. The $6,000 the position is worth today The $4,000 is gone in every branch of the decision. Holding is choosing to keep $6,000 in this stock rather than anywhere else.
In Staw’s experiment, which group put the most new money into the failing division?
- Those who inherited the original choice from a predecessor
- Those who had made the original choice themselves and were told it had failed
- Those whose original choice had succeeded
Show the answer
B. Those who had made the original choice themselves and were told it had failed Personal responsibility for a decision that went wrong drove the escalation: $13.07 million of $20 million, against about $9 million for those who inherited the choice.
An investor owns 100 shares bought at $50. At $30 she buys 100 more. What has changed?
- Her break-even price fell to $40 and her dollars at risk doubled from $3,000 to $6,000
- Her loss has been reduced
- The company’s prospects improved
Show the answer
A. Her break-even price fell to $40 and her dollars at risk doubled from $3,000 to $6,000 Averaging down moves the break-even closer and enlarges the position. The existing $2,000 loss is unchanged, and the company is the same company.
In the PayPal example, five purchases lost 41% while the first purchase alone lost 80%. Which statement is also true?
- Averaging down therefore reduced the investor’s loss
- The dollar loss was about $4,146 with averaging down against $1,609 without
- The stock recovered to the average cost
Show the answer
B. The dollar loss was about $4,146 with averaging down against $1,609 without A smaller percentage on a five times larger stake is a bigger dollar loss. Quoting only the percentage is a way of protecting the original decision.
Which reason for buying more of a fallen holding is not the sunk cost fallacy?
- To bring the average cost down
- To get back to even sooner
- Because staged purchases up to a fixed total were planned before the first one, and the reason for owning it has been re-checked
Show the answer
C. Because staged purchases up to a fixed total were planned before the first one, and the reason for owning it has been re-checked The fallacy is in the reason. A plan and a size limit made before the loss, plus a thesis that still holds, do not depend on what was already paid.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open Paper trading (the sample account when signed out, or your own). Pick the position with the lowest Gain %. Read its Avg cost, Price and Market value. Work out what would happen if you bought the same number of shares again at today’s Price: the new average cost, and the new market value. Then ask the cash test: with that market value in cash, would you buy this position today?
Open Paper tradingWhat you should find
The new average cost lands halfway between Avg cost and Price, and the market value doubles. Only the second number is money you can still lose. If your answer to the cash test depended on Avg cost, the past payment was deciding. Market value and your view of the company are the inputs that belong in the decision.
Sources and further reading
- Arkes and Blumer — The psychology of sunk cost, Organizational Behavior and Human Decision Processes 35(1), 1985
- Staw — Knee-deep in the big muddy: a study of escalating commitment to a chosen course of action, Organizational Behavior and Human Performance 16(1), 1976
- Thaler — Toward a positive theory of consumer choice, Journal of Economic Behavior and Organization 1(1), 1980
- Friedman, Pommerenke, Lukose, Milam and Huberman — Searching for the sunk cost fallacy, Experimental Economics 10(1), 2007
- Klein et al. — Investigating variation in replicability: a “many labs” replication project, Social Psychology 45(3), 2014
- Odean — Are investors reluctant to realize their losses? Journal of Finance 53(5), 1998
- Finabase chart data — PayPal Holdings (PYPL) monthly closes, June 2021 to October 2026
