Sunk Cost Fallacy Personal Finance Decisions: A $6,300 Case Study in Letting Go
In 1985, two psychologists asked people to imagine they were an airline president who had already sunk $9 million into a plane a competitor had just beaten to market. Eighty-five percent said to spend the last $1 million anyway. Told the same story without the $9 million, only about 17% wanted to fund it.
That gap is the engine behind sunk cost fallacy personal finance decisions, and I recognized it in myself while standing in a repair shop holding a $1,900 estimate for a car I had already put $4,400 into over fourteen months. What follows is that decision start to finish: the actual numbers, the research explaining why the wrong answer felt like the responsible one, and a five-step process you can run the next time you catch yourself protecting money that is already gone.
The $6,300 question: how fourteen months of repairs added up
The car was a sedan a little past the U.S. average. According to S&P Global Mobility, the average age of light vehicles on American roads hit a record 12.8 years in 2025, with passenger cars specifically averaging 14.5 years. Mine was in that range, paid off, and — in the phrase every owner of an old car repeats to themselves — “basically fine.”
Here is what “basically fine” actually cost:
| Month | Repair | Cost |
|---|---|---|
| 1 | Alternator and serpentine belt | $610 |
| 4 | Front brakes, rotors, one caliper | $840 |
| 7 | Water pump and timing service | $1,450 |
| 11 | Catalytic converter | $1,500 |
| 14 | Transmission service and rear main seal (quoted) | $1,900 |
| Total | Fourteen months | $6,300 |
Every single one of those repairs was defensible on its own. The alternator was cheaper than a new car payment. The water pump protected the engine. The converter was legally required to pass inspection. That is exactly the shape of the trap: no individual decision looks irrational, and the running total never gets audited. Most sunk cost fallacy personal finance mistakes are built out of individually reasonable choices, which is why they survive scrutiny one at a time.
What I actually said out loud in the shop was, “I just put fifteen hundred into the exhaust — I’m not walking away now.” That sentence contains the entire fallacy. The $1,500 was gone the moment the converter was bolted on. It had no bearing whatsoever on whether the next $1,900 was a good use of money.
Why sunk cost fallacy personal finance decisions feel like discipline
The reason this bias is so durable is that it wears the costume of a virtue. Walking away from something you have invested in reads as wasteful, flaky, or undisciplined. Continuing reads as committed. Our brains file “I already paid for it” under responsibility rather than under irrelevance.
Hal Arkes and Catherine Blumer documented this in their 1985 paper in Organizational Behavior and Human Decision Processes. Beyond the airplane scenario, they ran a field experiment on real buyers of season tickets to a university theater. Roughly a third of buyers paid the full $15 price, a third got a $13 price, and a third got $8 — assigned at random, after the person had already decided to buy. Over the following six months, the full-price group attended noticeably more performances than the discount groups, roughly 25% more in the first half of the season. Same seats, same plays, same people, different sunk cost. The only variable that moved attendance was money that could not be recovered either way.
Two other biases reinforce it. Loss aversion makes abandoning the project feel like booking a $4,400 loss rather than avoiding a $1,900 one — the same asymmetry that makes budget cuts feel disproportionately painful even when the math is neutral. And the endowment effect inflates what you think the thing is worth simply because you own it; the same mechanism shows up in everyday possessions people refuse to sell at market price. Add the ordinary gravitational pull of doing nothing, which is why default options quietly win most financial decisions, and the sunk cost fallacy rarely operates alone.
Deciding whether to repair the car you have or finance a replacement?
The math I should have run in the first month
The fix for a sunk cost is structural, not emotional: build a table with two columns, one for figures that belong in the decision and one for figures that do not. Anything already spent goes in the second column and then gets ignored. Here is that table for my car, using the numbers I had at month fourteen.
| Figure | Amount | Belongs in the decision? |
|---|---|---|
| Repairs already paid, months 1–11 | $4,400 | No — unrecoverable |
| “But the converter was brand new” | $1,500 | No — already counted above |
| Quoted repair in front of me | $1,900 | Yes — future cash out |
| Expected further repairs, next 24 months | ~$2,400 | Yes |
| Resale value after doing the repair | ~$3,800 | Yes |
| Resale value sold as-is, today | ~$1,600 | Yes |
Read only the green rows and the question becomes almost boring. Spending $1,900 lifted the car’s value by roughly $2,200 and bought two more years of driving, against an expected $2,400 of further repairs. It was close to a coin flip — which is a completely different conclusion from “obviously I have to fix it,” and the only reason it ever felt obvious was the red rows.
The $2,400 estimate was not a guess pulled from nowhere. AAA’s 2025 Your Driving Costs analysis puts maintenance, repair, and tire costs at 11.04 cents per mile, up from 10.13 cents the prior year. At 15,000 miles a year that is roughly $1,650 annually for a new vehicle under warranty. For a car past 150,000 miles with a fourteen-month repair history, doubling that baseline is conservative, not pessimistic.
The same trap, with worse consequences, inside a portfolio
A car is a bounded problem: the worst case is a few thousand dollars and an annoying Saturday. Sunk cost fallacy personal finance errors inside an investment account compound instead.
The retail version sounds like this: “I’m down 40% on this position, so I can’t sell it now.” That sentence treats the purchase price as a fact about the asset, when it is only a fact about your history. The market does not know or care what you paid. The only relevant question is whether that dollar, today, is better deployed here or somewhere else — and the answer never depends on your cost basis.
Morningstar’s Mind the Gap 2025 study quantifies what all this reshuffling costs. Over the ten years ending December 31, 2024, the average dollar invested in U.S. mutual funds and ETFs earned 7.0% annually, while the funds themselves returned 8.2% — a gap of about 1.2 percentage points a year, or roughly 15% of the total return, given up to badly timed buying and selling. Not all of that gap is sunk cost reasoning, but “hold the loser until it comes back, sell the winner to lock in the gain” is a textbook contributor, and it is the exact inverse of what the tax code rewards.
Which is the genuinely irritating part: the IRS will pay you to abandon a sunk cost. Realized capital losses offset capital gains dollar for dollar, and up to $3,000 of net loss per year can be deducted against ordinary income — $1,500 if married filing separately — with anything unused carrying forward indefinitely. That $3,000 cap has not been adjusted for inflation since 1978, so it is not generous, but it is free. Refusing to sell a losing position “until it recovers” declines a real, immediate tax benefit in exchange for a psychological one. Our walkthrough of whether tax-loss harvesting is worth the effort on a small portfolio covers the mechanics and the wash-sale rule.
Sunk costs also distort research. Once you have spent forty hours studying a company, those forty hours start defending the conclusion — which is how effort turns into conviction, the pattern behind why more research often makes investors more certain and less accurate.
I started tracking my own positions by “what would I buy today with this cash” rather than by cost basis a few years back, mostly out of engineering habit — if you can’t state the current decision without referencing history, the model is wrong. I hold index funds in tax-advantaged accounts and do all of this myself without an advisor, so there is nobody to talk me out of a bad rationalization; the reframe had to be built into the spreadsheet instead. The honest result: it changed maybe three decisions in four years. But two of them were expensive.
Five steps to break sunk cost fallacy personal finance decisions
- Say the sentence out loud and listen for the tense. If your justification is in the past tense — “I already paid,” “I’ve put so much into it,” “I’ve been doing this for three years” — it is not a reason. Reasons are in the future tense. This one filter catches most of it.
- Run the stranger test. Ask: if I acquired this asset today, for free, with zero history, would I spend $1,900 on it? Would I buy this stock at today’s price with fresh cash? If the answer is no, you are not deciding whether to continue — you are deciding whether to keep buying.
- Write the two-column table before you get emotional. Future costs and future values on one side, everything already spent on the other. Physically separating them on paper does more work than any amount of resolve, because the bias operates on attention, not intelligence. Knowing about a bias does not disarm it.
- Set the abandon-threshold in advance. Before the first repair, before the first share, write down the condition that ends it: “if cumulative repairs pass 60% of the car’s value in twelve months, it goes”; “if the thesis I wrote down is falsified, I sell regardless of price.” A rule written when you have nothing at stake is worth ten deliberations made when you do.
- Count the time, not just the money. The sunk-time effect is real, and time is the resource you cannot carry forward. A subscription you don’t use, a certification half-finished, a side project on life support — the correct question is never “how far in am I,” it is “what does the next hour buy.”
For the record, I did the $1,900 repair. Given the numbers it was a defensible call — but I made it for the wrong reason, and I only know it was defensible because I built the table afterward, which is precisely backwards. The same failure mode shows up in remodels and projects that run long past their estimates, a pattern covered in our piece on why a $30,000 kitchen closes at $41,000.
Key takeaways
- Money already spent is irrelevant to the next decision, every time, without exception. If it cannot be recovered, it cannot inform anything.
- Arkes and Blumer’s 1985 experiments showed 85% would fund a doomed project when a $9 million sunk cost was mentioned, versus roughly 17% when it was not.
- Real theater ticket buyers who paid full price attended about 25% more performances than randomly discounted buyers — identical tickets, different unrecoverable spend.
- Splitting figures into “future” and “already gone” columns beats willpower, because this bias operates on what you notice, not what you know.
- In a portfolio, refusing to sell at a loss also forfeits a real tax deduction: up to $3,000 of net capital loss per year against ordinary income, carried forward indefinitely.
- Write your exit condition before you have anything invested. That is the only moment you will ever be objective about it.
