Tuesday, 6 November 2012

The Sunk Cost Effect

Ah, the Concorde; the joint development program of the British and French governments that pushed ahead even when the economic benefits of the project were no longer possible. It was designed to be a passenger aircraft capable of supersonic flight but its lasting legacy resides mostly in game theory, where it has been adopted as a description of irrational behavior - the Concorde fallacy. More generally, the process behind the fallacy is known as the sunk cost effect.

As the Concorde example suggests, the problematic behavior in question is when a person continues to engage in a behavior due to their initial investment, even though the payoff is no longer available. In common parlance, this could be described as "knowing when to cut your losses"; or, as a famous philosopher once remarked, "You got to know when you hold 'em, know when to fold 'em, know when you walk away and know when to run". It was either Descartes or Kenny Rogers, I can never remember. 

It is mostly of interest to researchers because these behaviors violate our optimality predictions and instead of engaging in behaviors which maximise returns, there seems to be a consistent deviation towards sub-optimal responding. Initially it was believed to be an irrational approach that was unique to humans (and perhaps even limited to adult humans), which led to the hypothesis suggesting that the phenomenon was a product of higher-order thinking - specifically, the overgeneralisation of a rule like "Don't waste"1. Recent research, however, suggests that this might not be true2, 3.

For example, Kacelnick and Marsh4 looked at the preferences of starlings in a two phase task where they initially had to respond differently on two possible schedules - a high effort schedule (flying 16 times over a 1m distance) and a low effort schedule (flying 4 times over a 1m distance) that were signaled by different colours. In the second phase, the two alternatives had the same effort requirement but they found that the subjects would consistently prefer the alternative that had the same colour as the high effort schedule. The results were interpreted in terms of the sunk cost fallacy by arguing that the level of investment involved with the high effort schedule produced a greater perceived value of that alternative.


Even though we create optimality models for behavior, the point is not to assume that animals will act optimally but rather to use it as a comparison for how their behavior differs from our predictions which allows us to identify causal factors and variables causing their behavior. In this case, it was Macaskill and Hackenberg who recently attempted to uncover the underlying cause in their paper: "The Sunk Cost Effect with Pigeons: Some Determinants of Decisions about Persistence". They ran a series of related experiments where the requirements for persistence and escape were varied, and conditions where the response requirements were either signaled or unsignaled. 

The authors argue that their results suggest that there are two basic rules that determine whether an organism would commit the sunk cost fallacy: 1) the decision whether to escape, and 2) the decision when to escape. That is, the ability to be able to identify when to escape influenced the decision of whether to escape. So in the first two experiments they performed, which were essentially replications of prior studies, they found that the choice to persist was a function of the ratio of the mean global response requirements for 'escaping' to 'persisting'. The third and fourth experiments introduced a signal which indicated the best time to stop persisting. When the ratio of 'escaping' to 'persisting' was close to 1 (hard to distinguish), or when there was no signal (low discriminability), the subjects were more likely to commit the sunk cost fallacy. 

Interestingly, the authors also noted that overall there was a general tendency towards persisting regardless of the ratios and signals, and in one experiment they observed what they called the "reverse sunk cost fallacy" where the subjects would choose to escape when persisting was the best option. In agreement with similar findings5, they suggest that this was a demonstration of 'impulsivity' as choosing to escape meant a shorter delay to reinforcement (despite the reinforcement being less than what they would have received if they had persisted).


Is this research an example of the sunk cost fallacy; universities throwing good money after bad? I don't believe so. In addition to providing behavioral research with more precision and information on the factors underpinning the sunk cost fallacy (like disconfirming the suggestion that it is a result of verbal "don't waste" rules), I think this study also gives us more information about the decision making processes that control the behaviors of both humans and other animals. 

Perhaps on a more practical level it could also help us consider ways to improve our own behavior. When faced with difficult decisions concerning the investments we have made in a particular venture, it could be helpful to try to quantify and clarify the advantages and disadvantages of persisting and escaping our choices. By reducing the ambiguity of the benefits of each option, we improve our ability to identify the best time to escape and the best time to persist. This is presumably the logic behind writing up a "pros and cons" list and arguably this study lends some credence to the practice.


1. Arkes H.A, Ayton P. (1999) The sunk cost and Concorde effects: are humans less rational than lower animals. Psychological Bulletin. 5:591–600.

2. De la Piedad X, Field D, Rachlin H. (2006) The influence of prior choices on current choice. Journal of the Experimental Analysis of Behavior. 85:3–21.

3. Navarro A.D, Fantino E. (2005) The sunk cost effect in humans and pigeons. Journal of the Experimental Analysis of Behavior. 85:1–13.

4. Kacelnik A, Marsh B. (2002) Cost can increase preference in starlings. Animal Behavior. 63:245–250.

5. Avila-Santibanez R, Gonzalez-Montiel J.C, Miranda-Hernandez P, Guzman-Gonzalez M.D. (2010). Stimuli effects on optimal behavior in a sunk cost situation with pigeons. Mexican Journal of Behavior Analysis. 36:17–29.


  1. Nice review. I've often wondered, when hearing about this type of phenomenon, if a simpler operant mechanism might not work as an explanation. Presumably in the pigeons such a thing could be ruled out, but I don't know about the inspirational events like the Concord.

    For example, no one tries to invoke the sunk cost fallacy to explain why people keep playing slots. They do so because the reinforcement of winning (and possibly the incidental reinforcements in the course of playing) is stronger than the punishments (or more properly, extinction) for losing. There is also the complication that the people getting reinforced might not have been the people eventually punished... as in the mortgage industry where people got bonuses for selling mortgages, but the risks went to others.

    Interesting phenomenon either way!

    1. "I've often wondered, when hearing about this type of phenomenon, if a simpler operant mechanism might not work as an explanation. Presumably in the pigeons such a thing could be ruled out, but I don't know about the inspirational events like the Concord."

      Yes definitely, and that's essentially what the Macaskill and Hackenberg article does - it explains the phenomenon through the utilisation of operant principles. It's always difficult trying to apply basic research to large-scale real world situations but scientific studies like these give us reason to suspect specific mechanisms that are in play.

      I think the reason for using it as an explanation for some behaviors but not others is a result of differing levels of explanation. Things like the sunk cost effect are higher level descriptions of the irrational behavior that is occurring, but it's no more or less "true" than the operant principles underpinning it. In a similar sense, self-control experiments often highlight the irrational behavior of "preference reversal" (where the choice between two alternatives does not stay constant over time) but preference reversal is simply an emergent effect of the variables 'magnitude' and 'delay' interacting in simple decision making algorithms.

      The only problem with different levels of explanation seems to be when a higher order description is used as its own explanation. For example, positing that the impulsivity in ADHD is caused by an impaired executive function - this doesn't really tell us anything useful when we define the executive function as something which governs self-control and the ability to resist impulsive urges.

  2. I started to type out a reply, but it was getting long, and it didn't have anything to do with the sunk-cost effect, so I posted on my blog instead: http://fixingpsychology.blogspot.com/2012/11/tautology-part-1-cognitive-psychology.html