The transfer is the fourth one. The first was small, a test, and the profit appeared on the dashboard the next morning exactly as promised. The second was larger. The third was most of the savings account. Now the platform says the account is frozen for a tax audit and a payment will release it. He knows what this looks like. He also knows that if he stops now, everything sent so far is simply gone, and if he sends this one, there is a version of the story where it comes back.
He sends it.
Hal Arkes and Catherine Blumer at Ohio University published the foundational study in 1985. The best-known part was run on the university's own theatre subscribers. People buying a season ticket were, without knowing it, sold it at one of three prices: full price, a small discount, or a large discount. The tickets were otherwise identical. Over the first half of the season, the people who had paid full price attended more performances than the people who had paid less.
The money was gone either way. A rational account of the decision to go out on a cold night would weigh only the evening ahead. But the people who had paid more felt they had more to lose by staying home, and so they went. Arkes and Blumer called this the sunk cost effect: a greater tendency to continue an endeavour once an investment of money, effort or time has been made.
They paired the field study with a scenario that has been reused ever since. A company has spent most of a development budget on a new aircraft when a competitor announces a better, cheaper version. Should the company spend the remainder to finish? Most people said yes. Asked the same question without the prior spending, most said no. The prior spending changed the answer even though it could not change the outcome.
Barry Staw at the University of Illinois had described a related pattern in 1976 under the name escalation of commitment: managers who had made a losing investment decision, when given more money, put more of it into the failing option than managers who had inherited the decision. Responsibility for the original choice made people defend it.
The broad explanation combines two things. One is loss aversion, the finding from Daniel Kahneman and Amos Tversky's prospect theory that losses weigh more heavily than equivalent gains. Stopping converts an open position into a realised loss. Continuing keeps it open. The other is self-justification: quitting means accepting that the earlier decisions were wrong, and people will pay to avoid that admission.
The effect is among the most robust in the decision literature. It was one of the findings the Many Labs project successfully reproduced across many sites in 2014, at a time when much of social psychology was failing to replicate. In 2018 Brian Sweis and colleagues at the University of Minnesota reported it in mice and rats as well as humans, using a foraging task: the longer an animal had waited for a reward, the less willing it was to abandon the wait. Whatever this is, it is old.
The person sending the fourth transfer is not being stupid. He is in the grip of a mechanism that works on theatre subscribers and rodents. Fraud is designed to exploit it.
Long-form investment scams open with small deposits and visible returns precisely so that a sunk cost exists before any large demand is made. The frozen account, the tax, the release fee, are the aircraft scenario in person: you have already spent this much, surely you will spend a little more to finish. Recovery scams, in which a second fraudster contacts a victim offering to retrieve the lost money for an upfront fee, work on the same loss. The victim wants the earlier money back more than he wants to keep the new money safe.
The Office of Fair Trading's 2009 study of scam victims, led by Stephen Lea at the University of Exeter, noted that many victims had continued responding after they suspected something was wrong, and that the prospect of recovering earlier losses was part of what kept them engaged.
Caryl Rusbult's investment model, first published in 1980, holds that commitment to a relationship depends on three things: how satisfying it is, how good the alternatives are, and how much has been invested in it. Investments are years, shared property, children, mutual friends, a version of the self that only exists inside the relationship. They cannot be recovered by leaving. They function as sunk costs.
Rusbult and John Martz tested the model in 1995 with women at a domestic violence shelter. The ones more likely to return to an abusive partner were not the ones who reported more satisfaction. They were the ones with more invested and fewer alternatives. Time in the relationship worked against leaving.
This is the honest answer to the question people ask survivors: why did you stay so long. The length of the stay was part of the reason. Each year made the next year cheaper to endure than the exit, because the exit was priced in years already spent.
The research does not say that continuing is always wrong. Sometimes the extra payment does finish the aircraft. What it says is that the amount already spent carries no information about whether it will, and that people act as if it does. The useful question is the one Arkes and Blumer's participants failed to ask: if I were arriving at this situation now, with nothing behind me, would I put money in.