Invested
89 € for the ticket
Non-refundable
New situation
Match day, steady rain
Staying in would be nicer
Decision
“I already paid for it”
Evidence
The classic is a field study in a theater. Arkes and Blumer (1985) sold season tickets at randomly assigned prices. Those who had paid the full price attended more plays in the first half of the season: 4.11 against 3.32 and 3.29 tickets. In the second half, six to nine months after purchase, the difference had gone.
It carries less than it appears to. Fifty-four people were analysed, the discount was not a dependable source of variance in the overall analysis, and the group differences come from one-tailed planned comparisons (t = 1.79 and 1.83 — values that sit just at the conventional threshold).
The effect does replicate, but small. In the Many Labs project a sunk-cost scenario (Oppenheimer et al., 2009) appeared again in the aggregate across 36 samples (d = 0.27, original 0.23) — yet only half of the individual samples reached a dependable result on their own (Klein et al., 2014). For scale: 0.2 counts as a small effect.
Limits
The evidence concerns money and scenarios. Whether a few minutes of form-filling effort work similarly is open — Arkes and Blumer name non-monetary cases such as relationships or bad films only as examples in their introduction, not as data. Knowing about it offered no protection: students who had taken economics and covered the concept that term decided exactly the same way. For checkouts the transfer is not demonstrated.