Sunk cost

principle · origin: study · evidence: supported

In short

A sunk cost is an investment you can’t get back, whatever you decide from now on: a ticket already paid for, months put into a project, money spent on an old car. Economic theory says such costs shouldn’t matter, because a decision only changes the future. Arkes and Blumer (1985) defined the sunk-cost effect as “a greater tendency to continue an endeavor once an investment in money, effort, or time has been made.” The effect is real on average, but its size varies a lot from one situation to another.

What it says

The original study. Arkes and Blumer (1985) showed the effect in several experiments. We could not access the original article (it is with Elsevier, not open access and with no public abstract), so we describe it through the meta-analysis by Roth, Robbert and Straus (2015). In one scenario, participants had paid for two ski trips, a more expensive one and a cheaper one, which by mistake fell on the same weekend. Although the scenario made them prefer the cheaper trip, the higher price paid made participants more likely to choose the expensive one. As an explanation, Arkes and Blumer proposed the wish not to appear wasteful.

Two kinds of situations. Roth et al. (2015) split the studies into two groups:

• utilization decisions: you have already paid for something (a ticket, a membership) and decide whether, or how much, to use it;
• progress decisions: you have started a project, new information shows it is going badly, and you decide whether to invest more. In management this is called escalation of commitment.

The meta-analysis. Roth et al. (2015) pooled 100 effect sizes from dozens of studies. All of them are about money already spent, not time or effort.

The overall effect is moderate: d = 0.50. It is 0.58 for utilization decisions and 0.44 for progress decisions, and the difference between the two is not statistically significant. But the individual studies differ a lot, so the average doesn’t tell you how strong the effect will be in a given situation.

Three results from the same meta-analysis:

• Time matters, but differently. In utilization decisions, the effect weakens as time passes since the payment: money paid a few months ago “hurts” less. In progress decisions, the effect was instead stronger when a long time passed between decisions.
• Knowing economics doesn’t seem to protect you. Studies with participants trained in economic decision-making did not find a significantly smaller effect.
• Age. In utilization decisions, older participants showed a smaller effect.

Not only in humans. Sweis et al. (2018) built parallel tasks for mice, rats and humans: the subject waited for a reward (food, or a video clip) and could quit at any time. In all three species, the longer they had already waited, the more likely they were to wait until the end. But the effect appeared only after the subject had accepted the offer, not while they were still deciding whether to accept it.

Example

You have paid for a yearly gym membership. By March you realise you don’t like it and would rather run outside. The money for the membership is spent either way, whether you keep going or not. The useful question isn’t “how do I not waste the money?” but “what do I want to do with my exercise time from now on?”. If the answer is running, going to the gym just to “get your money’s worth” adds a cost rather than recovering one.

How to apply it

The steps below are a practical way to apply the idea, proposed by us based on the studies cited.

1. Ask the question without the past. “If I hadn’t invested anything so far, would I start this today, with what I know now?” If the answer is no, the past investment is no reason to continue.
2. Separate the cost from the decision. The money spent is lost in both options. Compare only what comes next: what still has to be invested and what you gain in each option.
3. Look at what continuing costs you. Every hour put into continuing can’t be put into something else. That is the opportunity cost.
4. Decide in advance when you will stop. In long projects, the effect was stronger over time. A threshold set at the start (“if I don’t have X by June, I stop”) is easier to keep than a decision made once you have already invested a lot.
5. Use the effect in your favour, deliberately. Roth et al. (2015) note that paying in advance can act as a commitment: people go to the gym more often right after paying. It is a trap when it keeps you in a bad choice, but it can help when it keeps you in a good one.

Limits and nuances

• It doesn’t show up everywhere. Friedman et al. (2007) tested the effect in a computer game where participants searched islands for treasure and paid to reach a new island. They found a “surprisingly small” effect, which did not change with the psychological factors they tested. Ashraf, Berry and Shapiro (2010) sold a water purification solution at different prices, door to door, in Zambia, and found no consistent evidence that those who paid more used it more.
• Hypothetical scenarios. Roth et al. (2015) note that most studies use imagined scenarios, often with students. In field studies, where the cost is not made salient, the effects are smaller.
• Not all persistence is an error. Continuing a project can be rational if you have since learned that it works, if quitting has real costs, or if reputation matters. It is an error only when the past investment, by itself, is the reason.
• The effect mixes with others. In projects, the wish to finish what you started or to justify an earlier decision produce the same behaviour. Roth et al. (2015) say their meta-analysis cannot fully separate them.
• The original study was not read directly. The description of Arkes and Blumer (1985) above comes from Roth et al. (2015). For Friedman et al. (2007), Ashraf et al. (2010) and Sweis et al. (2018) we read the abstracts, not the full articles.

Sources

• Hal R. Arkes, Catherine Blumer (1985). The psychology of sunk cost
• Stefan Roth, Thomas Robbert, Lennart Straus (2015). On the sunk-cost effect in economic decision-making: a meta-analytic review
• Daniel Friedman, Kai Pommerenke, Rajan Lukose, Garrett Milam, Bernardo A. Huberman (2007). Searching for the sunk cost fallacy
• Nava Ashraf, James Berry, Jesse M. Shapiro (2010). Can higher prices stimulate product use? Evidence from a field experiment in Zambia
• Brian M. Sweis, Samantha V. Abram, Brandy J. Schmidt, Kelsey D. Seeland, Angus W. MacDonald, Mark J. Thomas, A. David Redish (2018). Sensitivity to "sunk costs" in mice, rats, and humans

See also: Cost of action vs cost of inaction, Opportunity cost