Last March a reader wrote in with a position she had held for six years, down 40%, and the question she asked was about the company. The question was really about the money she had already put in. That distinction has a name in decision theory, and once you can see it, you notice it everywhere money meets memory.
What the term actually covers
A sunk cost is money already spent that no decision can recover. The textbook advice is to ignore it when choosing what to do next. The mind declines. Sunk cost reasoning is one of the most reliably reproduced biases in decision research, and investing is where it gets expensive, because the market prices your holdings every day and invites the comparison.
The clean historical case has a name too. The Concorde supersonic airliner, a joint British and French project approved in 1962, was forecast at roughly £70 million to develop. By the time it flew in 1969 the figure was past £1.3 billion, and both governments kept funding a plane each knew would never clear its cost. Retiring the project would have meant admitting the loss on paper. The Concorde fallacy is now shorthand for any commitment kept alive mainly to justify what it has already consumed. The aircraft itself served admirably until October 2003. The accounting lesson stands regardless.
How it shows up in a portfolio
The behavioral finance name for the pattern is the disposition effect. In a 1985 paper, Hersh Shefrin and Meir Statman documented that investors sell winners too early and hold losers too long, and decades of account data since have confirmed it. A gain feels like a verdict you get to collect. A loss feels like a verdict you can postpone.
Three shapes it takes in practice:
Averaging down on a thesis that has already failed its own milestones. Adding to a position is a new decision with new money, and the old decision doesn’t get a vote in it.
Refusing to rebalance because selling means realizing. The portfolio drifts toward whatever lost most recently, which is the opposite of what rebalancing is for.
Holding a concentrated single stock through a decade of underperformance because “it has to come back”. It has to do nothing of the kind. The company owes you no recovery date.
The working test
The question that cuts through it: if this holding were cash in your account this morning, would you buy it at today’s price? If the answer is no, you are holding it for the money you already spent, and the market has no idea what you paid.
A useful companion question is opportunity cost: what else would this capital do over the next five years? An index fund is a real alternative with a real history. The cash under your mattress is a real alternative. Sleep is a real alternative. Every month you hold a position out of memory rather than conviction, you pay the spread between them.
The honest exceptions
Tax-loss harvesting is a legitimate reason to keep a loser for a while longer. Offsetting realized gains is worth real money, and the calendar decides it. So does a written rebalancing band that says you won’t trade inside a range, because discipline has to survive its boring years to be discipline.
Those are rules set in advance. The sunk cost test targets the other case, where the reason is a number from years ago that no living decision can change.
What the model does for you
The test won’t make the sale. It names the force pushing against it. Shefrin and Statman’s paper is forty years old and the tendency is unfazed by knowing about it. What changes is the moment of recognition: you catch the question about the company, notice it is a question about your purchase price, and answer the question that the market is actually asking. Money already spent is gone either way. The portfolio only contains what you hold from today.