“World: Asia-Pacific: US demands ‘killing fields’ trial”, BBC 1998-12-29↩︎

And the Concorde definitely did not succeed commercially: operating it could barely cover costs, its minimal profits never came close to the total R&D or opportunity costs, its last flight was in 2003_21ya (a shockingly low lifetime in an industry which typically tries to operate individual planes, much less entire designs, for decades), and as of 2017, the Concorde still has no successors in its niche or apparent upcoming successors despite great technological progress & global economic development and the notorious growth in wealth of the “1%”.↩︎

Arkes & Blumer 1985_39ya, “The psychology of sunk cost”:

The sunk cost effect is manifested in a greater tendency to continue an endeavor once an investment in money, effort, or time has been made. Evidence that the psychological justification for this behavior is predicated on the desire not to appear wasteful is presented. In a field study, customers who had initially paid more for a season subscription to a theater series attended more plays during the next 6 months, presumably because of their higher sunk cost in the season tickets. Several questionnaire studies corroborated and extended this finding. It is found that those who had incurred a sunk cost inflated their estimate of how likely a project was to succeed compared to the estimates of the same project by those who had not incurred a sunk cost. The basic sunk cost finding that people will throw good money after bad appears to be well described by prospect theory (D. Kahneman & A. Tversky, 1979, Econometrica, 47, 263–291). Only moderate support for the contention that personal involvement increases the sunk cost effect is presented. The sunk cost effect was not lessened by having taken prior courses in economics. Finally, the sunk cost effect cannot be fully subsumed under any of several social psychological theories.

As an example of the sunk cost effect, consider the following example [from Thaler 1980_44ya]. A man wins a contest sponsored by a local radio station. He is given a free ticket to a football game. Since he does not want to go alone, he persuades a friend to buy a ticket and go with him. As they prepare to go to the game, a terrible blizzard begins. The contest winner peers out his window over the arctic scene and announces that he is not going, because the pain of enduring the snowstorm would be greater than the enjoyment he would derive from watching the game. However, his friend protests, ‘I don’t want to waste the twelve dollars I paid for the ticket! I want to go!’ The friend who purchased the ticket is not behaving rationally according to traditional economic theory. Only incremental costs should influence decisions, not sunk costs. If the agony of sitting in a blinding snowstorm for 3 h is greater than the enjoyment one would derive from trying to see the game, then one should not go. The $12 has been paid whether one goes or not. It is a sunk cost. It should in no way influence the decision to go. But who among us is so rational?

Our final sample thus had eighteen no-discount, nineteen $2 discount, and seventeen $7 discount subjects. Since the ticket stubs were color coded, we were able to collect the stubs after each performance and determine how many persons in each group had attended each play…We performed a 3 (discount: none, $2, $7) x 2 (half of season) analysis of variance on the number of tickets used by each subject. The latter variable was a within-subjects factor. It was also the only significant source of variance, F(1,51) = 32.32, MS, = 1.81, (p < 0.OO). More tickets were used by each subject on the first five plays (3.57) than on the last five plays (2.09). We performed a priori tests on the number of tickets used by each of the three groups during the first half of the theater season. The no-discount group used significantly more tickets (4.11) than both the $2 discount group (3.32) and the $7 discount group (3.29), t = 1.79, 1.83, respectively, p’s < .05, one tailed. The groups did not use significantly different numbers of tickets during the last half of the theater season (2.28, 1 .S4, 2.18, for the no-discount, $2 discount, and $7 discount groups, respectively). Conclusion. Those who had purchased theater tickets at the normal price used more theater tickets during the first half of the season than those who purchased tickets at either of the two discounts. According to rational economic theory, after all subjects had their ticket booklet in hand, they should have been equally likely to attend the plays.

…A second feature of prospect theory pertinent to sunk costs is the certainty effect. This effect is manifested in two ways. First, absolutely certain gains (P = 1) are greatly overvalued. By this we mean that the value of certain gains is higher than what would be expected given an analysis of a person’s values of gains having a probability less than 1.0. Second, certain losses (P = 1.0) are greatly undervalued (ie. further from zero). The value is more negative than what would be expected given an analysis of a person’s values of losses having a probability less than 1.0. In other words, certainty magnifies both positive and negative values. Note that in question 3A the decision not to complete the plane results in a certain loss of the amount already invested. Since prospect theory states that certain losses are particularly aversive, we might predict that subjects would find the other option comparatively attractive. This is in fact what occurred. Whenever a sunk cost dilemma involves the choice of a certain loss (stop the waterway project) versus a long shot (maybe it will become profitable by the year 2500), the certainty effect favors the latter option.

…Fifty-nine students had taken at least one course; sixty-one had taken no such course. All of these students were administered the Experiment 1 questionnaire by a graduate student in psychology. A third group comprised 61 students currently enrolled in an economics course, who were administered the Experiment 1 questionnaire by their economics professor during an economics class. Approximately three fourths of the students in this group had also taken one prior economics course. All of the economics students had been exposed to the concept of sunk cost earlier that semester both in their textbook (Gwartney & Stroup, 1982_42ya, p. 125 [Microeconomics: Private and public choice]) and in their class lectures. Results. Table 1 contains the results. The x² analysis does not approach significance. Even when an economics teacher in an economics class hands out a sunk cost questionnaire to economics students, there is no more conformity to rational economic theory than in the other two groups. We conclude that general instruction in economics does not lessen the sunk cost effect. In a recent analysis of entrapment experiments, Northcraft & Wolf1984 concluded that continued investment in many of them does not necessarily represent an economically irrational behavior. For example, continued waiting for the bus will increase the probability that one’s waiting behavior will be rewarded. Therefore there is an eminently rational basis for continued patience. Hence this situation is not a pure demonstration of the sunk cost effect. However, we believe that some sunk cost situations do correspond to entrapment situations. The subjects who ‘owned’ the airline company would have endured continuing expenditures on the plane as they sought the eventual goal of financial rescue. This corresponds to the Brockner et al. entrapment situation. However, entrapment is irrelevant to the analysis of all our other studies. For example, people who paid more money last September for the season theater tickets are in no way trapped. They do not incur small continuous losses as they seek an eventual goal. Therefore we suggest that entrapment is relevant only to the subset of sunk cost situations in which continuing losses are endured in the hope of later rescue by a further investment.

According to Thomas 1981_43ya [Microeconomic applications: Understanding the American economy], one person who recognized it as an error was none other than Thomas A. Edison. In the 1880s Edison was not making much money on his great invention, the electric lamp. The problem was that his manufacturing plant was not operating at full capacity because he could not sell enough of his lamps. He then got the idea to boost his plant’s production to full capacity and sell each extra lamp below its total cost of production. His associates thought this was an exceedingly poor idea, but Edison did it anyway. By increasing his plant’s output, Edison would add only 2% to the cost of production while increasing production 25%. Edison was able to do this because so much of the manufacturing cost was sunk cost. It would be present whether or not he manufactured more bulbs. [the Europe price > marginal cost] Edison then sold the large number of extra lamps in Europe for much more than the small added manufacturing costs. Since production increases involved negligible new costs but substantial new income, Edison was wise to increase production. While Edison was able to place sunk costs in proper perspective in arriving at his decision, our research suggests that most of the rest of us find that very difficult to do.

Friedman et al 2006_18ya criticism of Arkes:

This is consistent with the sunk cost fallacy, but the evidence is not as strong as one might hope. The reported significance levels apparently assume that (apart from the excluded couples) all attendance choices are independent. The authors do not explain why they divided the season in half, nor do they report the significance levels for the entire season (or first quarter, etc.). The data show no significant difference between the small and large discount groups in the first half season nor among any of the groups in the second half season. We are not aware of any replication of this field experiment.

Davis’s complaint is a little odd, inasmuch as economics textbooks do apparently discuss sunk costs; Steele 1996_28ya gives examples back to 1910_114ya, or from “Do Sunk Costs Matter?”, McAfee et al 2007_17ya:

Introductory textbooks in economics present this as a basic principle and a deep truth of rational decision-making (Frank and Bernanke, 2006_18ya, p. 10, and Mankiw, 2004_20ya, p. 297).

Popularized discussions of Farmer & Geanakoplos 2009_15ya:

• “‘Discounting’ the future cost of climate change”, Science News

• “Einstein on Wall Street, Time-Money Continuum: Mark Buchanan”, WSJ column

  • “Discounting Details”, on Mark Buchanan’s blog “The Physics of Finance”

Some quotes from the paper:

Conventional economics supposes that agents value the present vs. the future using an exponential discounting function. In contrast, experiments with animals and humans suggest that agents are better described as hyperbolic discounters, whose discount function decays much more slowly at large times, as a power law. This is generally regarded as being time inconsistent or irrational. We show that when agents cannot be sure of their own future one-period discount rates, then hyperbolic discounting can become rational and exponential discounting irrational. This has important implications for environmental economics, as it implies a much larger weight for the far future.

…Why should we discount the future? Bohm-Bawerk (1889_135ya,1923_101ya) and Fisher (1930_94ya) argued that men were naturally impatient, perhaps owing to a failure of the imagination in conjuring the future as vividly as the present. Another justification for declining Ds (τ) in τ, given by Rae (1834_190ya,1905_119ya), is that people are mortal, so survival probabilities must enter the calculation of the benefits of future potential consumption. There are many possible reasons for discounting, as reviewed by Dasgupta (2004_20ya, 2008_16ya). Most economic analysis assumes exponential discounting Ds (τ) = D(τ) = exp(−rτ), as originally posited by Samuelson (1937_87ya) and put on an axiomatic foundation by Koopmans (1960_64ya). A natural justification for exponential discounting comes from financial economics and the opportunity cost of foregoing an investment. A dollar at time s can be placed in the bank to collect interest at rate r, and if the interest rate is constant, it will generate exp(r(t—s)) dollars at time t. A dollar at time t is therefore equivalent to exp(−r(t—s)) dollars at time s. Letting τ = t—s, this motivates the exponential discount function Ds (τ) = D(τ) = exp(−rτ), independent of s.

…For roughly the first eighty years the certainty equivalent discount function for the geometric random walk stays fairly close to the exponential, but afterward the two diverge substantially, with the geometric random walk giving a much larger weight to the future. A comparison using more realistic parameters is given in Table 1. For large times the difference is dramatic.

Farmer & Geanakoplos 2009_15ya: Table 1, comparing geometric random walk (GRW) vs exponential discounting over increasing time periods showing that GRW eventually decays much slower.

year	GRW	exponential
20	0.462	0.456
60	0.125	0.095
100	0.051	0.020
500	0.008	2 × 10−9
1000	0.005	4 × 10−18

…What this analysis makes clear, however, is that the long term behavior of valuations depends extremely sensitively on the interest rate model. The fact that the present value of actions that affect the far future can shift from a few percentage points to infinity when we move from a constant interest rate to a geometric random walk calls seriously into question many well regarded analyses of the economic consequences of global warming. … no fixed discount rate is really adequate—as our analysis makes abundantly clear, the proper discounting function is not an exponential.

For example, Staw 1981_43ya, “The Escalation of Commitment to a Course of Action”:

A second way to explain decisional errors is to attribute a breakdown in rationality to interpersonal elements such as social power or group dynamics. Pfeffer [1977_47ya] has, for example, outlined how and when power considerations are likely to outweigh more rational aspects of organizational decision making, and Janis [1972_52ya] has noted many problems in the decision making of policy groups. Cohesive groups may, according to Janis, suppress dissent, censor information, create illusions of invulnerability, and stereotype enemies. Any of these by-products of social interaction may, of course, hinder rational decision making and lead individuals or groups to decisional errors.

Wang & Keil2005; from the abstract:

Using meta-analysis, we analyzed the results of 20 sunk cost experiments and found: (1) a large effect size associated with sunk costs, (2) variability of effect sizes across experiments that was larger than pure subject-level sampling error, and (3) stronger effects in experiments involving IT projects as opposed to non-IT projects.

Background on why one might expect effects with IT in particular:

Although project escalation is a general phenomenon, IT project escalation has received considerable attention since Keil and his colleagues began studying the phenomenon (Keil, Mixon et al 1995_29ya). Survey data suggest that 30–40% of all IT projects involve some degree of project escalation (Keil, Mann, and Rai 2000_24ya). To study the role of sunk cost in software project escalation, Keil et al 1995_29ya conducted a series of lab experiments, in which sunk costs were manipulated at various levels, and subjects decided whether or not to continue an IT project facing negative prospects. This IT version of the sunk cost experiment was later replicated across cultures (Keil, Tan et al 2000_24ya), with group decision makers (Boonthanom 2003_21ya), and under different de-escalation situations (Heng, Tan et al 2003_21ya). These experiments demonstrated the sunk cost effect to be significant in IT project escalation.

The “real option” defense of sunk cost behavior has been suggested for software projects (Tiwana & Fichman2006)↩︎

“Diffusion of Responsibility: Effects on the Escalation Tendency”, Whyte 1991_33ya (see also Whyte1993):

In a laboratory study, the possibility was investigated that group decision making in the initial stages of an investment project might reduce the escalation tendency by diffusing responsibility for initiating a failing project. Support for this notion was found. Escalation effects occurred less frequently and were less severe among individuals described as participants in a group decision to initiate a failing course of action than among individuals described as personally responsible for the initial decision. Self-justification theory was found to be less relevant after group than after individual decisions. Because most decisions about important new policies in organizations are made by groups, these results indicate a gap in theorizing about the determinants of escalating commitment for an important category of escalation situations.

…The impact of personal responsibility on persistence in error has been replicated several times (eg. Bazerman, Beekun, & Schoorman, 1982_42ya; Caldwell & O’Reilly, 1982_42ya; Staw, 1976_48ya; Staw & Fox, 1977_47ya).

Both of them—but the first one on the Iraqi side, specifically Saddam Hussein; Bazerman & Neale 1992_32ya:

Similarly, it could be argued that in the Iraqi/Kuwait conflict, Iraq (Hussein) had the information necessary to rationally pursue a negotiated settlement. In fact, early on in the crisis, he was offered a package for settlement that was far better than anything that he could have expected through a continued conflict. The escalation literature accurately predicts that the initial “investment” incurred in invading Kuwait would lead Iraq to a further escalation of its commitment not to compromise on the return of Kuwait.

Kelly 2004_20ya:

The physicist Eugene Demler informs me that exactly parallel arguments were quite commonly made in the Soviet Union in the late 1980s in an attempt to justify continued Soviet involvement in Afghanistan.

Dawkins & Carlisle 1976_48ya sarcastically remark:

…The idea has been influential⁴, and it appeals to economic intuition. A government which has invested heavily in, for example, a supersonic airliner, is understandably reluctant to abandon it, even when sober judgement of future prospects suggests that it should do so. Similarly, a popular argument against American withdrawal from the Vietnam war was a retrospective one: ‘We cannot allow those boys to have died in vain’. Intuition says that previous investment commits one to future investment.

The former can be found in Bazerman, Giuliano, & Appelman, 1984, Davis & Bobko, 1986, & Staw, 1976 among other studies cited here. The latter is often called ‘self-justification’ or the ‘justification effect’ (eg. Brockner1992).

Self-justification is, of course, in many contexts a valuable trait to have; is the following an error, or business students demonstrating their precocious understanding of an invaluable bureaucratic in-fighting skill? Bazerman et al 1982_42ya, “Performance evaluation in a dynamic context: A laboratory study of the impact of prior commitment to the ratee” (see also Caldwell & O’Reilly, 1982_42ya; Staw, 1976_48ya; Staw & Fox, 1977_47ya):

A dynamic view of performance evaluation is proposed that argues that raters who are provided with negative performance data on a previously promoted employee will subsequently evaluate the employee more positively if they, rather than their predecessors, made the earlier promotion decision. A total of 298 business majors participated in the study. The experimental group made a promotion decision by choosing among three candidates, whereas the control group was told that the decision had been made by someone else. Both groups evaluated the promoted employee’s performance after reviewing 2 years of data. The hypothesized escalation of commitment effect was observed in that the experimental group consistently evaluated the employee more favorably, provided larger rewards, and made more optimistic projections of future performance than did the control group.

And it is difficult to judge from a distance when sunk cost has occurred: what exactly else are the Indians going to invest in? Remember our exponential discounting example. As long as various settlements are not running at an outright loss or are being subsidized, how steep an opportunity cost do they really face? From the paper:

By the end of the occupation in the late-A.D. 800s there is evidence of depletion of wood resources, pi on seeds, and animals (reviewed by Kohler 1992_32ya). Following the collapse of these villages, the Dolores area was never reoccupied in force by Puebloan farmers. A second similar case comes from nearby Sand Canyon Locality west of Cortez, Colorado, intensively studied by the Crow Canyon Archaeological Center over the last 15 years (Lipe 1992_32ya). Here the main occupation is several hundred years later than in Dolores, but the patterns of construction in hamlets versus villages are similar (fig. 4, bottom). The demise of the two villages contributing dated construction events to fig. 4 (bottom) coincides with the famous depopulation of the Four Corners region of the U.S. Southwest. There is strong evidence for declining availability of protein in general and large game animals in particular, and increased competition for the best agricultural land, during the terminal occupation (reviewed by Kohler 2000_24ya). We draw a final example from an intermediate period. The most famous Anasazi structures, the “great houses” of Chaco Canyon, may follow a similar pattern. Windes & Ford 1996_28ya show that early construction episodes (in the early A.D. 900s) in the canyon great houses typically coincide with periods of high potential agricultural productivity, but later construction continues in both good periods and bad, particularly in the poor period from ca. A.D. 1030^–₂₀1050_974ya.

Certainly there is strong evidence of diminishing marginal returns—evidence for the Tainter thesis—but diminishing marginal returns is not sunk cost fallacy. Given the general environment, and given that there was a ‘collapse’, arguably there was no opportunity cost to remaining there. How would the Indians have become better off if they abandoned their villages, given that there is little evidence that other places were better off in that period of great droughts and the observation that they would need to make substantial capital investments wherever they went?↩︎

Janssen & Scheffer 2004_20ya:

In fact, escalation of commitment is found in group decision making (Bazerman et al 1984). Members of a group strive for unanimity. A typical goal for political decisions within small-scale societies is to reach consensus (Boehm1996). Once unanimity is reached, the easiest way to protect it is to stay committed to the group’s decision (Bazerman et al 1984_40ya, Janis 1972_52ya [Victims of groupthink]). Thus, when the group is faced with a negative feedback, members will not suggest abandoning the earlier course of action, because this might disrupt the existing unanimity.

McAfee et al 2007_17ya:

But there are also examples of people who succeeded by not ignoring sunk costs. The same “we-owe-it-to-our-fallen-countrymen” logic that led Americans to stay the course in Vietnam also helped the war effort in World War II. More generally, many success stories involve people who at some time suffered great setbacks, but persevered when short-term odds were not in their favor because they “had already come too far to give up now.” Columbus did not give up when the shores of India did not appear after weeks at sea, and many on his crew were urging him to turn home (see Olson, 1967_57ya [The Northmen, Columbus and Cabot, 985–1503], for Columbus’ journal). Jeff Bezos, founder of Amazon.com, did not give up when Amazon’s loss totaled $1.4 billion in 2001, and many on Wall Street were speculating that the company would go broke (see Mendelson and Meza, 2001).

“Banking on Commitment: Intended and Unintended Consequences of an Organization’s Attempt to Attenuate Escalation of Commitment”, McNamara et al 2002_22ya:

The notion that decision makers tend to incorrectly consider previous expenditures when deliberating current utility-based decisions (Arkes & Blumer, 1985_39ya) has been used to explain fiascoes ranging from the prolonged involvement of the United States in the Vietnam War to the disastrous cost overrun during the construction of the Shoreham Nuclear Power Plant (Ross & Staw, 1993_31ya). In the Shoreham Nuclear Power Plant example, escalation of commitment meant billions of wasted dollars (Ross & Staw, 1993_31ya). In the Vietnam War, it may have cost thousands of lives…Kirby and Davis’s (1998_26ya) experimental study showed that increased monitoring could dampen the escalation of commitment. Staw, Barsade, and Koput’s (1997_27ya) field data on the banking industry led them to conclude that top manager turnover led to de-escalation of commitment at an aggregate level.

…So far, the results support the efficacy of changes in monitoring and decision responsibility as cures for the escalation of commitment bias. We now turn to the side effects of these treatments. Hypotheses 4 and 5 propose that the threat of increased monitoring and change in management responsibility increase the likelihood of a different form of undesirable decision commitment—the persistent underassessment of borrower risk. The results in column 3 of Table 2 support these hypotheses. Both the threat of increased monitoring and the threat of change in decision responsibility increase the likelihood of persistent underassessment of borrower risk (0.47, p < 0.01, and 0.50, p < 0.05, respectively). These findings support the view that decision makers are likely to fail to appropriately downgrade a borrower when, by doing so, they avoid an organizational intervention. We examined the change in investment commitment for borrowers whose risk was persistently underassessed and who faced either increased monitoring or change in decision responsibility if the decision makers had admitted that the risk needed downgrading. We found that decision makers did appear to exhibit escalation of commitment to these borrowers. The change in commitment (on average, over 30%) is significantly greater than 0 (t = 2.94, p < 0.01) and greater than the change in commitment to those borrowers who were correctly assessed as remaining at the same risk level (t = 2.58, p = 0.01). Combined, these findings suggest that although the organizational efforts to minimize undesirable decision commitment appeared successful at first glance, the threat of these interventions increased the likelihood that decision makers would persistently give overfavorable assessments of the risk of borrowers. In turn, the lending officers would then escalate their monetary commitment to these riskier borrowers.

On nuclear power plants as sunk cost fallacy, McAfee et al 2007_17ya:

According to evidence reported by De Bondt & Makhija1988, managers of many utility companies in the U.S. have been overly reluctant to terminate economically unviable nuclear plant projects. In the 1960s, the nuclear power industry promised “energy too cheap to meter.” But nuclear power later proved unsafe and uneconomical. As the U.S. nuclear power program was failing in the 1970s and 1980s, Public Service Commissions around the nation ordered prudency reviews. From these reviews, De Bondt and Makhija find evidence that the Commissions denied many utility companies even partial recovery of nuclear construction costs on the grounds that they had been mismanaging the nuclear construction projects in ways consistent with “throwing good money after bad.”…In most projects there is uncertainty, and restarting after stopping entails costs, making the option to continue valuable. This is certainly the case for nuclear power plants, for example. Shutting down a nuclear reactor requires dismantling or entombment, and the costs of restarting are extremely high. Moreover, the variance of energy prices has been quite large. The option of maintaining nuclear plants is therefore potentially valuable. Low returns from nuclear power in the 1970s and 1980s might have been a consequence of the large variance, suggesting a high option value of maintaining nuclear plants. This may in part explain the evidence (reported by De Bondt and Makhija, 1988_36ya) that managers of utilities at the time were so reluctant to shut down seemingly unprofitable plants.

Steven Pinker, The Better Angels of Our Nature 2011_13ya, pg 336:

In the case of a war of attrition, one can imagine a leader who has a changing willingness to suffer a cost over time, increasing as the conflict proceeds and his resolve toughens. His motto would be: ‘We fight on so that our boys shall not have died in vain.’ This mindset, known as loss aversion, the sunk-cost fallacy, and throwing good money after bad, is patently irrational, but it is surprisingly pervasive in human decision-making.⁶⁵ People stay in an abusive marriage because of the years they have already put into it, or sit through a bad movie because they have already paid for the ticket, or try to reverse a gambling loss by doubling their next bet, or pour money into a boondoggle because they’ve already poured so much money into it. Though psychologists don’t fully understand why people are suckers for sunk costs, a common explanation is that it signals a public commitment. The person is announcing: ‘When I make a decision, I’m not so weak, stupid, or indecisive that I can be easily talked out of it.’ In a contest of resolve like an attrition game, loss aversion could serve as a costly and hence credible signal that the contestant is not about to concede, preempting his opponent’s strategy of outlasting him just one more round.

It’s worth noting that there is at least one example of sunk cost (“entry licenses” [fees]) encouraging cooperation (“collusive price path”) in market agents: Offerman & Potter 2001_23ya, “Does Auctioning of Entry Licenses Induce Collusion? An Experimental Study”, who point out another case of how our sunk cost map may not correspond to the territory:

There is one caveat to the sunk cost argument, however. If the game for which the positions are allocated has multiple equilibria, an entry fee may affect the equilibrium that is being selected. Several experimental studies have demonstrated the force of this principle. For example, Cooper, DeJong, Forsythe & Ross1993, Van Huyck, Battalio & Beil1993, and Cachon and Camerer, (1996_28ya) study coordination games with multiple equilibria and find that an entry fee may induce players to coordinate on a different (Pareto superior) equilibrium.

McAfee et al 2007

Reputational Concerns. In team relationships, each participant’s willingness to invest depends on the investments of others. In such circumstances, a commitment to finishing projects even when they appear ex post unprofitable is valuable, because such a commitment induces more efficient ex ante investment. Thus, a reputation for “throwing good money after bad”—the classic sunk cost fallacy—can solve a coordination problem. In contrast to the desire for commitment, people might rationally want to conceal bad choices to appear more talented, which may lead them to make further investments, hoping to conceal their investments gone bad.

Kanodia, Bushman, and Dickhaut (1989_35ya), Prendergast & Stole1996, and Camerer & Weber 1999_25ya develop principal-agent models in which rational agents invest more if they have invested more in the past to protect their reputation for ability. We elucidate the general features of these models below and argue that concerns about reputation for ability are especially powerful in explaining apparent reactions to sunk costs by politicians. Carmichael and MacLeod (2003_21ya) [see also Carpenter & Matthews2003] develop a model in which agents initially make investments independently and are later matched in pairs, their match produces a surplus, and they bargain over it based on cultural norms of fair division. A fair division rule in which each agent’s surplus share is increasing in their sunk investment, and decreasing in the other’s sunk investment, is shown to be evolutionarily stable.

…If a member of an illegal price-fixing cartel seems likely to confess to the government in exchange for immunity from prosecution, the other cartel members may race to be first to confess, since only the first gets immunity (in Europe, such immunity is called “leniency”). Similarly, a spouse who loses faith in the long-term prospects of a marriage invests less in the relationship, thereby reducing the gains from partnership, potentially dooming the relationship. In both cases, beliefs about the future viability matter to the success of the relationship, and there is the potential for self-fulfilling optimistic and pessimistic beliefs.

In such a situation, individuals may rationally select others who stay in the relationship beyond the point of individual rationality, if such a commitment is possible. Indeed, ex ante it is rational to construct exit barriers like costly and difficult divorce laws, so as to reduce early exit. Such exit barriers might be behavioral as well as legal. If an individual can develop a reputation for sticking in a relationship beyond the break-even point, it would make that individual a more desirable partner and thus enhance the set of available partners, as well as encourage greater and longer lasting investment by the chosen partner. One way of creating such a reputation is to act as if one cares about sunk costs…We now formalize this concept using a simple two-period model that sets aside consideration of selection…That is, a slight possibility of breach is collectively harmful; both agents would be ex ante better off if they could prevent breach when V—ρ < 1, which holds as long as the reputation cost ρ of breaching is not too small. In this model, a tendency to stay in the relationship due to a large sunk investment would be beneficial to each party.

The qualifier is because hyperbolic discounting has been demonstrated in many primates, and a number of other biases, eg. Chen et al 2006_18ya, “How Basic Are Behavioral Biases? Evidence from Capuchin Monkey Trading Behavior”↩︎

See also Radford & Blakey 2000_24ya, “Intensity of nest defence is related to offspring sex ratio in the great tit Parus major”:

Nest-defence behavior of passerines is a form of parental investment. Parents are selected, therefore, to vary the intensity of their nest defence with respect to the value of their offspring. Great tit, Parus major, males were tested for their defence response to both a nest predator and playback of a great tit chick distress call. The results from the two trials were similar; males gave more alarm calls and made more perch changes if they had larger broods and if they had a greater proportion of sons in their brood. This is the first evidence for a relationship between nest-defence intensity and offspring sex ratio. Paternal quality, size, age and condition, lay date and chick condition did not significantly influence any of the measured nest-defence parameters.

…The most consistent pattern found in studies of avian nest defence has been an increase in the level of the parental response to predators from clutch initiation to £edging (eg. Biermann & Robertson1981; Regelmann & Curio1983; Montgomerie & Weatherhead1988; Wiklund 1990_34ya a). This supports the prediction from parental investment theory (Trivers 1972_52ya) that parents should risk more in defence of young that are more valuable to them. The intensity of nest defence is also expected to be positively correlated with brood size because the benefits of deterring a predator will increase with offspring number (Williams 1966_58ya; Wiklund 1990_34ya b).

Northcraft & Wolf 1984_40ya, “Dollars, sense, and sunk costs: A lifecycle model of research allocation decisions”. Academy of Management Review, 9, 225–234:

The decision maker also may treat the negative feedback as simply a learning experience-a cue to redirect efforts within a project rather than abandon it (Connolly, 1976_48ya).

…In some cases (Brockner, Shaw, & Rubin, 1979_45ya), the expected rate of return for further financial commitment even can be shown with a few assumptions to be increasing and (after a certain amount of investment) financially advisable, despite the claim that further resource commitment under the circumstances is psychologically rather than economically motivated…More to the point, the life cycle model clearly reveals the psychologist’s fallacy: continuing a project in the face of a financial setback is not always irrational (it depends on the stage in the project and the magnitude of the financial setback). Second, the life cycle model provides an insight into the manager’s preoccupation with a project’s financial past. It demonstrates how a project’s financial past can be used heuristically to understand the project’s future.

Friedman et al 2006_18ya:

…There are also several possible rational explanations for an apparent concern with sunk costs. Maintaining a reputation for finishing what you start may have sufficient value to compensate for the expected loss on an additional investment. The ‘real option’ value (eg. Dixit and Pindyck, 1994_30ya [Investment Under Uncertainty]) [cf. O’Brien & Folta2009, Tiwana & Fichman 2006_18ya] of continuing a project also may offset an expected loss. Agency problems in organizations may make it personally better for a manager to continue an unprofitable project than to cancel it and take the heat from its supporters (eg. Milgrom and Roberts, 1992_32ya [Economics, Organization, and Management]).

(Certainty effects seem to be supported by fMRI imaging.) One may ask why capital constraints aren’t solved—if the projects really are good profitable ideas—by resort to equity or debt? But those are always last resorts due to fundamental coordination & trust issues; McAfee et al 2007_17ya:

Abundant theoretical literature in corporate finance shows that imposing financial constraints on firm managers improves agency problems (see Stiglitz and Weiss, 1981_43ya, Myers and Majluf, 1984_40ya, Lewis and Sappington, 1989_35ya, and Hart and Moore, 1995_29ya). The theoretical conclusion finds overwhelming empirical support, and only a small fraction of business investment is funded by borrowing (see Fazzari and Athey, 1987_37ya, Fazzari and Peterson, 1993_31ya, and Love, 2003_21ya). When managers face financial constraints, sunk costs must influence firm investments simply because of budgets…Firms with financial constraints might rationally react to sunk costs by investing more in a project, rather than less, because the ability to undertake alternative investments declines in the level of sunk costs…Given limited resources, if the firm has already sunk more resources into the current project, then the value of the option to start a new project if it arises is lower relative the value of the option to continue the current project, because fewer resources are left over to bring any new project to fruition, and more resources have already been spent to bring the current project to fruition. Therefore, the firm’s incentive to continue investing in the current project is higher the more resources it has already sunk into the project.

• Stiglitz, Joseph E. and Weiss, Andrew, 1981_43ya. “Credit Rationing in Markets with Imperfect Information”, American Economic Review 71, 393–410.

• Myers, Stewart and Majluf, Nicholas S., 1984_40ya. “Corporate Financing and Investment Decisions when Firms Have Information that Investors Do Not Have,” Journal of Financial Economics 13, 187–221

• Lewis, Tracy and Sappington, David E. M., 1989_35ya. “Countervailing Incentives in Agency Problems,” Journal of Economic Theory 49, 294–313

• Hart, Oliver and Moore, John, 1995_29ya. “Debt and Seniority: An Analysis of the Role of Hard Claims in Constraining Management,” American Economic Review 85, 567–585

• Fazzari, Steven and Athey, Michael J., 1987_37ya. “Asymmetric Information, Financing Constraints, and Investment,” Review of Economics and Statistics 69, 481–487.

• Fazzari, Steven and Petersen, Bruce, 1993_31ya. “Working Capital and Fixed Investment: New Evidence on Financing Constraints,” RAND Journal of Economics 24, 328–342

• Love, Inessa, 2003_21ya. “Financial Development and Financing Constraints: International Evidence from the Structural Investment Model,” Review of Financial Studies 16, 765–791

The swans:

Although none of these terms are used, the same phenomena is also observed by Nolet et al 2001. In particular, tundra swans must expend more energy to “up-end” to feed on deep-water tuber patches than they do to “head-dip” to feed on shallow-water patches; however, contrary to the expectations of Nolet et al the swans feed for a longer time on each high-cost deep-water patch. In every context, the observation of the sunk-cost effect is an enigma because intuition suggests that this behavior is suboptimal. Here, we show how optimization of Eq. (3) predicts the sunk-cost effect for certain scenarios; a common element of every case is a large initial cost.

Klaczynski & Cottrell 2004_20ya:

Although considerable evidence indicates that adults commit the SC fallacy frequently, age differences in the propensity to honour sunk costs have been little studied. In their investigations of 7–15-year-olds (Study 1) and 5–12-year-olds (Study 2), Baron et al 1993_31ya found no relationship between age and SC decisions. By contrast, Klaczynski (2001_23yab) reported that the SC fallacy decreased from early adolescence to adulthood, although normative decisions were infrequent across ages. A third pattern of findings is reviewed by Arkes & Ayton 1999_25ya. Specifically, Arkes and Ayton argue that two studies (Krouse, 1986_38ya; Webley & Plaiser, 1998_26ya) indicate that younger children commit the SC fallacy less frequently than older children. Making sense of these conflicting findings is difficult because criticisms can be levied against each investigation. For instance, Arkes & Ayton 1999_25ya questioned the null findings of Baron et al 1993_31ya because sample sizes were small (eg. in Baron et al Study 2, n per age group ranged 7–17). The problems used by Krouse (1986_38ya) and Webley & Plaiser 1998_26ya were not, strictly speaking, SC problems (rather, they were problems of ‘mental accounting’; see Webley & Plaiser, 1998_26ya). Because Klaczynski (2001_23yab) did not include children in his sample, the age trends he reported are limited to adolescence. Thus, an interpretable montage of age trends in SC decisions cannot be created from prior research.

…An alternative proposition is based on the previously outlined theory of the role of metacognition in mediating interactions between analytic and heuristic processing. In this view, even young children have had ample opportunities to convert the ‘waste not’ heuristic from a conscious strategy to an automatically activated heuristic stored as a procedural memory. Evidence from children’s experiences with food (eg. Birch, Fisher, & Grimm-Thomas, 1999_25ya) provides some support for the argument that even preschoolers are frequently reinforced for not ‘wasting’ food. Mothers commonly extort their children to ‘clean up their plates’ even though they are sated and even though the nutritional effects of eating more than their bodies require are generally negative. If the ‘waste not’ heuristic is automatically activated in sunk cost situations for both children and adults, then one possibility is that no age differences in committing the fallacy should be expected. However, if activated heuristics are momentarily available for evaluation in working memory, then the superior metacognitive abilities of adolescents and adults should allow them to intercede in experiential processing before the heuristic is actually used. Although the evidence is clear that most adults do not take advantage of this opportunity for evaluation, the proportion of adolescents and adults who actively inhibit the ‘waste not’ heuristic should be greater than the same proportion of children.

eg. “Discounting of Delayed Rewards: A Life-Span Comparison”, Green et al 1994_30ya; abstract:

In this study, children, young adults, and older adults chose between immediate and delayed hypothetical monetary rewards. The amount of the delayed reward was held constant while its delay was varied. All three age groups showed delay discounting; that is, the amount of an immediate reward judged to be of equal value to the delayed reward decreased as a function of delay. The rate of discounting was highest for children and lowest for older adults, predicting a life-span developmental trend toward increased self-control. Discounting of delayed rewards by all three age groups was well described by a single function with age-sensitive parameters (all R²s > .94). Thus, even though there are quantitative age differences in delay discounting, the existence of an age-invariant form of discount function suggests that the process of choosing between rewards of different amounts and delays is qualitatively similar across the life span.

“The Bias Against Creativity: Why People Desire But Reject Creative Ideas”, Mueller et al 2011_13ya:

Uncertainty is an aversive state (Fiske & Taylor, 1991_33ya [Social cognition]; Heider, 1958_66ya [The psychology of interpersonal relations]) which people feel a strong motivation to diminish and avoid (Whitson & Galinsky, 2008).

“Searching for the Sunk Cost Fallacy”, Friedman et al 2007_17ya:

Subjects play a computer game in which they decide whether to keep digging for treasure on an island or to sink a cost (which will turn out to be either high or low) to move to another island. The research hypothesis is that subjects will stay longer on islands that were more costly to find. Nine treatment variables are considered, eg. alternative visual displays, whether the treasure value of an island is shown on arrival or discovered by trial and error, and alternative parameters for sunk costs. The data reveal a surprisingly small and erratic sunk cost effect that is generally insensitive to the proposed psychological drivers.

I cite Friedman 2006_18ya here so much because it’s unusual—as McAfee et al 2007_17ya puts it:

…Most of the existing empirical work has not controlled for changing hazards, option values, reputations for ability and commitment, and budget constraints. We are aware of only one study in which several of these factors are eliminated—Friedman et al 2006_18ya. In an experimental environment without option value or reputation considerations, the authors find only very small and non-statistically-significant sunk cost effects in the majority of their treatments, consistent with the rational theory presented here.

On the lessons of Garland et al 1990’s observation of non sunk cost fallacy, McAfee et al 2007_17ya:

While some projects have an increasing hazard, others appear to have a decreasing hazard. For example, curing cancer, originally expected to cost $1 billion (see Epstein, 1998), probably has a decreasing hazard; given initial failure, the odds of immediate success recede and the likely expenditures required to complete grow. Oil-exploration projects might also be characterized by decreasing hazards. Suppose a firm acquires a license to drill a number of wells in a fixed area. It decides to drill a well on a particular spot in the area. Suppose the well turns out to be dry. The costs of drilling the well are then sunk. But the dry well might indicate that the likelihood of striking oil on another spot in the area is low since the geophysical characteristics of surface rocks and terrain for the next spot are more or less the same as the ones for the previous spot that turned out to be dry. Thus, the firm might be rationally less likely to drill another well. In general, firms might be less willing to drill another well the more wells they had already found to be dry. This may in part explain the rapid “de-escalation” observed by Garland, Sandeford, and Rogers (1990_34ya) in their oil-exploration experiments.

Bornstein et al 1999:

Measurements and main results: Residents evaluated medical and non-medical situations that varied the amount of previous investment and whether the present decision maker was the same or different from the person who had made the initial investment. They rated reasons both for continuing the initial decision (eg. stay with the medication already in use) and for switching to anew alternative (eg. a different medication). There were two main findings: First, the residents’ ratings of whether to continue or switch medical treatments were not influenced by the amount of the initial investment (p’s > 0.05). Second, residents’ reasoning was more normative in medical than in non-medical situations, in which it paralleled that of undergraduates (p’s < 0.05).

Conclusion: Medical residents’ evaluation of treatment decisions reflected good reasoning, in that they were not influenced by the amount of time and/or money that had already been invested in treating a patient. However, the residents did demonstrate a sunk-cost effect in evaluating non-medical situations. Thus, any advantage in decision making that is conferred by medical training appears to be domain specific.

Some of this was replicated & generalized in Braverman & Blumenthal-Barby2012:

Specifically, we surveyed 389 health care providers in a large urban medical center in the United States during August 2009_15ya. We asked participants to make a treatment recommendation based on one of four hypothetical clinical scenarios that varied in the source and type of prior investment described. By comparing recommendations across scenarios, we found that providers did not demonstrate a sunk-cost effect; rather, they demonstrated a significant tendency to over-compensate for the effect. In addition, we found that more than one in ten health care providers recommended continuation of an ineffective treatment.

Staw 1981_43ya, “The Escalation of Commitment to a Course of Action”:

…However, when choosing to commit resources, subjects did not appear to persist unswervingly in the face of continued negative results or to ignore information about the possibility of future returns. These inconsistencies led to a third study [Staw & Ross, 1978_46ya] designed specifically to find out how individuals process information following negative versus positive feedback. In this third study, previous success/failure and causal information about a setback were both experimentally varied. Results showed that subjects invested more resources in a course of action when information pointed to an exogenous rather than endogenous cause of a setback, and this tendency was most pronounced when subjects had been given a previous failure rather than a success. The exogenous cause in this experiment was one that was both external to the program in which subjects invested and was unlikely to persist, whereas the endogenous cause was a problem central to the program and likely to persist.

“Continuing investment under conditions of failure: A laboratory study of the limits to escalation”, McCain 1986_38ya:

Brockner et al 1982, following Teger (1980_44ya), have also specifically suggested that entrapment involves two distinct stages. In the first stage subjects respond primarily to economic incentives, whereas self-justification supposedly governs the second. Brockner et al. found that cost salience significantly reduced entrapment early on but had little effect in later periods…Thus, a process that reflects efforts to learn both what caused the setbacks and the implications of that cause for future action may provide a better model of de-escalation.

…The findings of this study clearly showed that the escalation effect, defined by a difference between the allocations of high- and low-choice subjects, was limited to the initial stages of continuing investment. The findings were consistent with previous research (Staw & Fox, 1977) and support the contention that investment in failing projects involves two stages. Clearly, too, the availability of alternative investments limited the escalation effect. When subjects were given alternatives to the failing investment, the difference between the investments of the high- and low-choice groups disappeared. The results showed, as well, that high-choice subjects who displayed the escalation effect quit funding the failing investment sooner than comparable low-choice subjects, contrary to a commitment perspective. Similarly, the declining hazard rates observed here support a learning model more than they support the self-justification model…Some authors (eg. Northcraft & Wolf, 1984) have suggested that investors react differently to cost overruns than they react to revenue shortfalls, yet many escalation experiments do not clearly specify whether setbacks result from higher than expected costs or from lower than expected revenues. Clearly, if investors are sensitive to uncertainty, as the attributional model suggests, researchers must consider how subjects may respond to an inadequately specified investment context…

“Sunk and Opportunity Costs in Valuation and Bidding”, Phillips et al 1991, which also mentions another apparent instance of market agents initially committing sunk cost and then learning: Plott & Uhl 1981_43ya “Competitive Equilibrium with Middlemen: An Empirical Study”, Southern Economic Journal↩︎

In particular, “Pay-to-bid” penny auctions such as Swoopo have been called instances of sunk cost fallacy, or “escalation of commitment” by no less than Richard H. Thaler, and described by techie Jeff Atwood as being “as close to pure, distilled evil in a business plan as I’ve ever seen”. But as Wang & Xu 2012_12ya and Caldara2012 indicate, while people do lose money to the penny auctions, they eventually do learn that penny auctions are not good ideas and escape the trap. A previous analysis of penny auctions, Augenblick2009, omitted detailed survivorship data but still found some learning effects. And indeed, Swoopo has since shut down.↩︎

“Learning lessons from sunk costs”, Bornstein & Chapman 1995_29ya:

Study participants rated the quality of several arguments for continuing an original plan in sunk cost situations in order to (a) avoid wasting resources, (b) learn to make better decisions, (c) punish poor decision making, and (d) appear consistent. The lesson-learning argument was perceived as most appropriate when adult teachers taught lessons to others, the original decision was carelessly made, or if it consumed comparatively more resources. Ratings of the lesson-learning argument were higher for teacher-learner than for adult-alone situations, regardless of whether the learner was a child or an adult. The implications for improving decision making and judging whether the sunk cost effect is a bias are discussed…However, prospect theory does not predict an effect of variables such as whether the decision maker acted alone, the care with which the decision was made, or the nature of the relationship between teacher and learner. The other three responses were influenced by these variables.

What appears to be a bias in the laboratory may be functional behavior in a more realistic context (Funder, 1987_37ya; Hogarth, 1981_43ya), where a variety of justifications for the behavior can be considered. In general, ignoring sunk costs is an adaptive, cost-effective strategy. Yet what appears to be biased, irrational behavior—such as decreasing utility through attention to irretrievably wasted resources—can be described as ‘meta-rational’ (Jungermann, 1986_38ya), assuming the benefits of learning and implementing the lesson outweigh the costs of sticking to the original plan. However, it raises the interesting question of why continuing a failed plan is the best (or even a good) way to learn to make better decisions in the future. Perhaps one could both abandon the current unsuccessful plan and learn to think more carefully in future decisions.

See Arkes & Blumer 1985_39ya, which found no resistance in students who had were taken an economics course and most of whom had taken other economics courses; also good is Larrick et al 1993_31ya or Larrick et al 1990, “Teaching the use of cost-benefit reasoning in everyday life”:

It may be seen in Table 1 that economists’ reasoning on the university and international policy questions was more in line with cost-benefit rules than was that of biologists and humanists. This pattern was found for the net-benefit questions (p < 0.05), and for the opportunity cost questions (p < 0.05) and a trend was found for the sunk cost questions (p < 0.15)…Third, economists were more likely than biologists and humanists to report that they ignored sunk costs or attended to opportunity costs in their personal decisions. For instance, they were more likely to have dropped a research project because it was not proving worthwhile. (It is interesting to note that economists were not simply more likely to drop projects. All three disciplines gave the same answer on average to the question “have you ever dropped a research project because of a lack of funding?”) Finally, economists participated in a greater number of time-saving activities…The results show that training people only briefly on an economic principle significantly alters their solutions to hypothetical economic problems [including sunk cost]. Moreover, training effects generalize fully from a financial domain to a nonfinancial one and vice versa…The means for both indices showed that [quick-economics] trained subjects were ignoring sunk costs more than untrained subjects, but only the nine-item index based on the question “Have you bought one of the following items at some time and then not used it in the past month” approached significance. Trained subjects reported that they had paid for but not used 1.14 objects and activities compared to 0.84 for untrained subjects, f(78) = 1.64, p = 0.10.

On the other hand, Fennema & Perkins2008:

The results indicate that practicing Certified Public Accountants (CPAs), Masters of Business Administration students (MBAs) and undergraduate accounting students perform better than undergraduate psychology students. The level of training, as measured by the number of college courses in managerial accounting, was found to be positively correlated with performance, while the level of experience, as measured by years of financially-related work, was not. Justification was found to improve decisions only for those participants with significant work experience (MBAs and CPAs). Strategies used in this type of decision were examined with the surprising finding that economically rational decisions can be made even if sunk costs are not ignored.

For example, Heath1995 spends a page criticizing Brockner & Rubin1985’s setup of endowed subjects buying tickets in a lottery, pointing out they took subjects quitting ticket purchases after a long run of ticket-buying as evidence of sunk cost, even though if the expected value of the lottery was positive, the normative rational strategy is for the subject spend every penny of the endowment buying tickets! “Consider, for example, the average of $3.82 invested in the game with a $10.00 prize. In this game, the average subject quits at a point where the expected benefits from a marginal investment are three times what they were when the subject began investing.” [emphasis added]↩︎

The previously mentioned studies of sunk cost in children found minimal correlations with intelligence, when that was measured. For adults, see Strough et al 2008_16ya, previously cited. Also, Keith E. Stanovich, & West, R. F. (2008_16yab). “On the relative independence of thinking biases and cognitive ability”. Journal of Personality and Social Psychology, 94, 672–695 (pg 7–8):

Both cognitive ability groups displayed sunk-cost effects of roughly equal magnitude. For the high-SAT group, the mean in the no-sunk-cost condition was 6.90 and the mean in the sunk-cost condition was 5.08, whereas for the low-SAT group, the mean in the no-sunk-cost condition was 6.50 and the mean in the sunk-cost condition was 4.19. A 2 (cognitive ability) ϫ 2 (condition) ANOVA indicated a significant main effect of cognitive ability, F(1, 725) ϭ 8.40, MSE ϭ 9.13, p Ͻ .01, and a significant main effect of condition, F(1, 725) ϭ 84.9, MSE ϭ 9.13, p Ͻ .001. There was a slight tendency for the low-SAT participants to show a larger sunk-cost effect, but the Cognitive Ability ϫ Condition interaction did not attain statistical-significance, F(1, 725) ϭ 1.21, MSE ϭ 9.13. The interaction was also tested in a regression analyses in which SAT was treated as a continuous variable rather than as a dichotomous variable. The Form ϫ SAT cross product, when entered third in the equation, was not significant, F(1, 725) ϭ 0.32.

The sunk-cost effect thus represents another cognitive bias that is not strongly attenuated by cognitive ability. However, this is true only when it is assessed in a between-subjects context. Using a similar sunk-cost problem, Stanovich & West1999 did find an association with cognitive ability when participants responded in a within-subjects design.

And Parker & Fischhoff 2005_19ya:

The first two rows of Table 5 show strong correlations between five of the seven DMC component measures and respondents’ scores on the WISC-R vocabulary test and on Giancola et al.’s (1996_28ya) measure of ECF. Consistency in risk perception and resistance to sunk cost show little relationship to either of these general cognitive abilities.^8,9 [correlations: 0.12, 0.08]

Bruine de Bruin et al 2007_17ya, modifying Parker & Fischhoff2005’s test battery, improved the consistency of the sunk cost questions, and found similar small correlations with their 2 IQ measures, of 0.17 and 0.04. Larrick et al 1993_31ya recorded SAT/ACT scores (close proxies for IQ) and found some correlation, and noted that IQ was “positively related to recognition of economists’ position on various economic problems.”↩︎

“Sunk Costs in the NBA: Why Draft Order Affects Playing Time and Survival in Professional Basketball”:

A second problem is that much of the escalation literature, despite its intent to sources of commitment, has not directly explain nonrational challenged the assumptions of economic decision making. By and large, the escalation literature has demonstrated that psychological and social factors can influence resource allocation decisions, not that the rational assumptions of decision making are in error. A third weakness is that almost all the escalation literature is laboratory based. Aside from a few recent qualitative case studies (eg. Ross and Staw, 1986_38ya, 1993_31ya), escalation predictions have not been confirmed settings, using data that are or falsified in real organizational generated in their natural context. Therefore, despite the size of the escalation literature, it is still uncertain if to escalation effects can be generalized from the laboratory the field.

…Garland, Sandefur, and Rogers (1990_34ya) found a similar absence of sunk-cost effects in an experiment using an oil-drilling scenario. Prior expenditures on dry wells were not associated with continued drilling, perhaps because dry wells were so clearly seen as reducing rather than increasing the likelihood of future oil production. Thus it appears that sunk costs may only be influential on project decisions when they are linked to the perception (if not the reality) of progress on a course of action.

…Table 2 also shows that draft order was a significant predictor of minutes played over the entire five-year period. This effect was above and beyond any effects of a player’s performance, injury, or trade status. The regressions showed that every increment in the draft number decreased playing time by as much as 23 minutes in the second year (I, = −22.77, p < 0.001, one-tailed test). Likewise, being taken in the second rather than the first round of the draft meant 552 minutes less playing time during a player’s second year in the NBA.

Friedman et al 2006_18ya: “…Of course, it is hard to completely rule out other explanations based on unobserved components of performance or the coaches’ Bayesian priors.”↩︎

“Reinvestment decisions by entrepreneurs: Rational decision-making or escalation of commitment?”, McCarthy et al 1993_31ya:

The hypotheses were tested using data from a longitudinal study involving 1112_912ya firms. It was found that entrepreneurs who had started their firms and those who had expressed substantial over-confidence were significantly more likely to make the decision to expand. The hypotheses that those who had partners and those who expected to apply their skills would be more likely to expand were not supported. Furthermore, and consistent with previous research, these psychological escalation predictors seemed to exert a greater influence when feedback from the marketplace was negative. As expected, there was a declining influence in the third year as compared with the second. Consistent with the prior literature and the hypotheses, these psychological predictors did show a small, but systematic influence upon reinvestment decisions.

…Although the hypothesis regarding PARTNR was not supported, as noted, the zero-order correlation between PARTNR and NEWCAP2 is in the predicted direction (r = 0.06, p < 0.05, one-tailed). Thus, entrepreneurs with partners may be more likely to expand the asset base of their firms than they would be if they were sole owners. This has significant implications for entrepreneurial teams, in that the presence of partners does not inhibit the tendency to escalate, but in fact increases that tendency. This means that having partners is not insurance against the tendency to escalate. This is consistent with the research on escalation (Bazerman et al 1984_40ya).

…A puzzling finding was the lack of any relationship between financial indicators from the previous year and new capital invested in the business. In other words, there was no systematic relationship between sales growth and expansion of the asset base for these young firms. This may mean that many of these firms started with some excess capacity so that it was not necessary to add to facilities to support their early growth. It may also mean that management of working capital was erratic. On the other hand. the psychological factors predicted by escalation theory did, in two of four cases, show systematic relationships to additional investment.

…One final issue worth comment is the relatively small amount of variance accounted for by the models described in this study. The variance accounted for in this research is in line with the findings in similar studies of escalation. In a recent field study of the escalation bias, Schoorman (1988_36ya) reported that the escalation bias accounted for 6% of the variance in performance ratings. Schoorman (1988_36ya) noted in this article that the escalation variables were more powerful predictors of performance (at 6%) than a measure of ability used in a validated selection test for these same employees…Taken together these findings provide support for the view that escalation bias is a significant and common problem in decision-making among entrepreneurs. The characteristics of entrepreneurs and the nature of the decisions they are required to make leave them particularly vulnerable to escalation bias. Efforts to train entrepreneurs to guard against this bias may be very valuable.

“Reconceptualizing entrepreneurial exit: Divergent exit routes and their drivers”, Wennberg et al 2009_15ya:

…An alternative failure-avoidance strategy is to invest additional equity. We found that such reinvestments reduced the probability of all exit routes. While previous research on reinvestment also found that reinvestment was not related to well-defined performance levels (McCarthy et al 1993_31ya), it is interesting that it also reduced the odds of harvest sales and harvest liquidations. As a failure-avoidance strategy, reinvestment thus seems to be less effective than cost reduction. Cost reductions have direct effects on firm performance while reinvestments provide a temporary buffer for failing firms. As suggested, there might be disincentives to additional investments if tax laws punish entrepreneurs taking out money as salaries or dividends. If corroborated, this is an important finding for public policy makers.

See Cohen & Dupas2010 and Ashraf et al 2007; on the hypothetical (summary from Holla & Kremer2008, pg 11):

When they divide their sample into households that displayed a sunk-cost effect when responding to a hypothetical scenario posed to them by surveyors and those that did not, they find coefficients of much larger magnitude for the hypothetical-sunk-cost households, although these remain insignificant and cannot be statistically distinguished from the estimated effects for households that did not display this hypothetical sunk-cost effect. Ashraf et al 2007_17ya identify hypothetical-sunk-cost households from their answers to the following question posed during the follow-up survey: Suppose you bought a bottle of juice for 1,000 Kw. When you start to drink it, you realize you don’t really like the taste. Would you finish drinking it?

Friedman et al 2006_18ya:

Do Internet users respond to sunk time costs? Manley & Seltzer (1997_27ya) report that after a particular website imposed an access charge, the remaining users stayed longer. A rival explanation to the sunk cost fallacy is selection bias: the users with shortest stays when the site was free are those who stopped coming when they had to pay. Klein et al 1999 report that users stick around longer on their site after encountering delays while playing a game, but again selection bias is a possible alternative explanation. The issue is important in e-commerce because ‘stickier’ sites earn more advertising revenue. Schwartz (1999_25ya) [Digital Darwinism: 7 Breakthrough Business Strategies for Surviving in the Cutthroat Web Economy] reports that managers of the free Wall Street Journal site deliberately slowed the login process in the belief that users would then stay longer. One of us (Lukose) took a sample of 2000_24ya user logs from a website and found a significant positive correlation between residence time at the site and download latency. One alternative explanation is unobserved congestion on the web, and users may have been responding more to expected future time costs than to time costs already sunk. Also, good sites may be more popular because they are good, leading to (a) congestion and (b) more time spent on the site.

Friedman et al 2006_18ya:

…Barron et al 2001_23ya find that US firms are significantly more likely to terminate projects following the departure of top managers. This might reflect the new managers’ insensitivity to costs sunk by their predecessors, or it might simply reflect two aspects of the same broad realignment decision.

A “traditional maxim in Chinese statecraft”.↩︎

Dil Green’s description:

• ‘tactical optimism’: David Bohm’s term for the way in which humans overcome the (so far) inescapable assessment that; ‘in the long run, we’re all dead’. Specifically, within the building industry, rife with non-optimal ingrained conditions, you wouldn’t come to work if you weren’t an optimist. Builders who cease to have an optimistic outlook go and find other things to do.

It’s hard not to think of Arkes & Hutzel 2000_24ya, “The Role of Probability of Success Estimates in the Sunk Cost Effect”

The sunk cost effect is manifested in a tendency to continue an endeavor once an investment has been made. Arkes & Blumer 1985_39ya showed that a sunk cost increases one’s estimated probability that the endeavor will succeed [ p(s)]. Is this p(s) increase a cause of the sunk cost effect, a consequence of the effect, or both? In Experiment 1 participants read a scenario in which a sunk cost was or was not present. Half of each group read what the precise p(s) of the project would be, thereby discouraging p(s) inflation. Nevertheless these participants manifested the sunk cost effect, suggesting p(s) inflation is not necessary for the effect to occur. In Experiment 2 participants gave p(s) estimates before or after the investment decision. The latter group manifested higher p(s), suggesting that the inflated estimate is a consequence of the decision to invest.

The reasoning goes like this:

Another reason for honoring the sunk cost of the movie ticket (related to avoiding regret) is that you know yourself well enough to realize you often make mistakes. There are many irrational reasons why you would not want to see the movie after all. Maybe you’re unwilling to get up and go to the movie because you feel a little tired after eating too much. Maybe a friend who has already seen the movie discourages you to go, even though you know your tastes in movies don’t always match. Maybe you’re a little depressed and distracted by work/relationship/whatever problems. Etc.

For whatever reason, your past self chose to buy the ticket, and your present self does not want to see the movie. Your present self has more information. But this extra information is of dubious quality, and is not always relevant to the decision. But it still influences your state of mind, and you know that. How do you know which self is right? You don’t, until after you’ve seen the movie. The marginal costs, in terms of mental discomfort, of seeing the movie and not liking it, are usually smaller than the marginal benefit of staying home and thinking about what a great movie it could have been. The reasoning behind this trivial example can easily be adapted to sunk cost choices in situations that do matter.

And again:

People who take into account sunken costs in everyday decisions will make better decisions on average. My argument relies on the proposition that a person’s estimate of his own utility function is highly noisy. In other words, you don’t really know if going to the movie will make you happy or not, until you actually do it.

So if you’re in this movie-going situation, then you have at least two pieces of data. Your current self has produced an estimate that says the utility of going to the movie is negative. But your former self produced an estimate that says the utility is substantially positive—enough so that he was willing to fork over $10. So maybe you average out the estimates: if you currently value the movie at −$5, then the average value is still positive and you should go. The real question is how confident you are in your current estimate, and whether that confidence is justified by real new information.

“Real artists ship”, as the saying goes, and don’t give into the temptation to rewrite entire systems unless starting over or starting an entirely new system is truly necessary (particularly given planning fallacy). One might call using “sunk cost fallacy” to justify abandoning partial projects for new projects the “‘sunk cost fallacy’ fallacy”:

I have a problem with never finishing things that I want to work on. I get enthusiastic about them for a while, but then find something else to work on. This problem seems to be powered partially by my sunk costs fallacy hooks. When faced with the choice of finishing my current project or starting this shiny new project, my sunk costs hook activates and says “evaluate future expected utility and ignore sunk costs”. The new project looks very shiny compared to the old project, enough that it looks like a better thing to work on than the rest of the current project. The trouble is that this always seems to be the case. It seems weird that the awesomeness of my project ideas would have exponential growth over time, so there must be something else here.

Johnicholas:

…Sometimes it can be hard to maintain a good balance among multiple activities. For example, it is important to notice new good ideas. However, I tend to spend too much time pursuing novelty, and not enough time working on the best idea that I’ve found so far. There is a tradition of browser games (see Kingdom Of Loathing) that enforce a kind of balance using a virtual currency of ‘turns’. You accumulate turns slowly in real time, and essentially every action within the game uses up turns. This enforces not spending too much time playing the game (and increases the perceived value of the game via forced artificial scarcity, of course). If I gave myself ‘explore dollars’ for doing non-exploration (so-called exploit) tasks, and charged myself for doing exploration tasks (like reading arXiv or Wikipedia), I could enforce a balance. If I were also prone to the opposite problem (‘A few months in the lab can often save whole hours in the library.’), then I might use two currencies; exploring costs explore points but rewards with exploit points, and exploiting costs exploit points but rewards with explore points. (Virtual currencies are ubiquitous in games, and they can be used for many purposes; I expect to find them able to be placed across from many different failure modes.)

Mass Driver:

…I have the same problem at work; although, by mainstream society’s standards, I am a reasonably successful professional, I can’t really sit down and write a great essay when I’m too hot, or, at least, it seems like I would be more productive if I stopped writing for 5 minutes and cranked up the A/C or changed into shorts. An hour later, it seems like I would be more productive if I stopped writing for 20 minutes and ate lunch. Later that afternoon, it seems like I would be more productive if I stopped for a few minutes and read an interesting article on general science. These things happen even in an ideal working environment, when I’m by myself in a place I’m familiar with. If I have coworkers, or if I’m in a new town, there are even more distractions. If I have to learn who to ask for help with learning to use the new software so that I can research the data that I need to write a report, then I might spend 6 hours preparing to spend 1 hour writing a report.

All this worries me for two reasons: (1) I might be failing to actually optimize for my goals if I only spend 10–20% of my time directly performing target actions like ‘write essay’ or ‘kayak with friends’, and (2) even if I am successfully optimizing, it sucks that the way to achieve the results that I want is to let my attention dwell on the most efficient ways to, say, brush my teeth. I don’t just want to go kayaking, I want to think about kayaking. Thinking about driving to the river seems like a waste of cognitive ‘time’ to me.

Daniel Meade:

…I’ve had so many ‘projects’ over the past few years I’ve lost count. Has any one of them seen the light of day? Well yes, but that failed miserably. The point is, it’s all too easy to get distracted and move onto something else, perhaps it’s that hurdle where you’re just not sure what to do next or how to do it, so instead of finding a way to tackle it head on, you take the easy way out and start something new. I’m quick to blame my failings on the lack of capital, that I ‘need’ to get my projects off the ground. And that justifies my avoidance. Of course, it doesn’t, far from it. But I just don’t know how to push through, not right now any way.

We all have heard of businesses engaged in sunk cost, but that doesn’t tell us anything unless we know to what degree, if any, they engage in the opposite behavior, switching too much; from “How DigiCash Blew Everything”, Next! Magazine:

It all started out quite nicely. The brand new company sold a smart card for closed systems which was a cash-cow for years. It was at this time that the first irritants appeared. Even if you are a brilliant scientist, that doesn’t mean you are a good manager. David Chaum was a control freak, someone who couldn’t delegate anything to anyone else, and insisted upon watching over everybody’s shoulders. “That resulted in slowing down research,” explains an ex-DigiCash employee who wished to remain anonymous. “We had a lot of half-finished product. He continuously changed his mind about where things were headed.”

Initiation rituals may increase commitment as they become more unpleasant or demanding (Aronson & Mills1959); and we can see attempts in our daily lives with consumerism—buying gym memberships or exercise equipment; McAfee et al 2007_17ya:

Anecdotal evidence suggests that individuals may even exploit their own reactions to sunk expenditures to their advantage. Steele (1996_28ya, p. 610) [cf. Elster 2000_24ya, Ulysses Unbound: Studies in Rationality, Precommitment and Constraints, and Kelly2004] and Walton (2002_22ya, p. 479) recount stories of individuals who buy exercise machines or gym memberships that cost in the thousands of dollars, even though they are reluctant to spend this much money, reasoning that if they do, it will make them exercise, which is good for their health. A reaction to sunk costs that assists in commitment is often helpful.

Or prepaying for lessons, or buying excessively expensive writing tools:

‘Are Moleskines really worth the cost compared to Mead? If so, why?’

Plenty of people seem to swear by them. But here’s the thing—it’s not so much the cost (in absolute sums, it’s not that large). It’s whether you use it. You obviously sweat over costs; perhaps this sweating can be a cudgel to force you to write down whatever. The more a moleskine isn’t worth buying, the more you will find yourself compelled to use it. Then wouldn’t you be better off in the end?

Geoffrey Miller mocks this logic (pg 122 of Spent 2011_13ya):

All experienced fitness machine salespeople are well aware that this is the fate of most of their products. What they are really selling consumers is the delusion that the sunk costs of buying the machines will force them to exercise conscientiously. (The consumers know that they could have already been jogging for months around their neighborhood parks in their old running shoes, but they also know that their access to the parks and shoes has not, empirically, been sufficient to induce regular exercise.) So, the consumer thinks: ‘If I invest $3,900 in this PreCor EFX5.33 elliptical trainer, it will (1) call forth regular aerobic activity from my flawed and unworthy body, through the techno-fetishistic magic of its build quality, and (2) save me money in the long run by reducing medical expenses.’ The salesperson meanwhile thinks: ‘20% commission!’ and the manufacturer thinks: ‘We can safely offer a ten-year warranty; because the average machine only gets used seventeen times in the first two months after purchase.’ Everybody’s happy, except for most consumers, and they don’t complain because they think it’s all their fault that they’re failing to use the machine.’ The few conscientious consumers who do use the equipment regularly enjoy many benefits: efficient muscle building and fat burning through the low perceived exertion of the PreCor’s smooth elliptical movement; a lean body that elicits lust and respect; a self-satisfied glow of moral superiority.

Grit is a slightly narrower version of Conscientiousness; from “Grit: Perseverance and Passion for Long-Term Goals”:

…We define grit as perseverance and passion for long-term goals. Grit entails working strenuously toward challenges, maintaining effort and interest over years despite failure, adversity, and plateaus in progress. The gritty individual approaches achievement as a marathon; his or her advantage is stamina. Whereas disappointment or boredom signals to others that it is time to change trajectory and cut losses, the gritty individual stays the course. Our hypothesis that grit is essential to high achievement evolved during interviews with professionals in investment banking, painting, journalism, academia, medicine, and law. Asked what quality distinguishes star performers in their respective fields, these individuals cited grit or a close synonym as often as talent. In fact, many were awed by the achievements of peers who did not at first seem as gifted as others but whose sustained commitment to their ambitions was exceptional. Likewise, many noted with surprise that prodigiously gifted peers did not end up in the upper echelons of their field.

More than 100 years prior to our work on grit, Galton (1892_132ya) collected biographical information on eminent judges, statesmen, scientists, poets, musicians, painters, wrestlers, and others. Ability alone, he concluded, did not bring about success in any field. Rather, he believed high achievers to be triply blessed by ‘ability combined with zeal and with capacity for hard labour’ (p. 33). Similar conclusions were reached by Cox (1926_98ya) in an analysis of the biographies of 301 eminent creators and leaders drawn from a larger sample compiled by J. M. Cattell (1903_121ya). Estimated IQ and Cattell’s rank order of eminence were only moderately related (r = ϭ.16) when reliability of data was controlled for. Rating geniuses on 67 character traits derived from Webb (1915_109ya), Cox concluded that holding constant estimated IQ, the following traits evident in childhood predicted lifetime achievement: ‘persistence of motive and effort, confidence in their abilities, and great strength or force of character’ (p. 218).

…However, in the Terman longitudinal study of mentally gifted children, the most accomplished men were only 5 points higher in IQ than the least accomplished men (Terman & Oden, 1947_77ya). To be sure, restriction on range of IQ partly accounted for the slightness of this gap, but there was sufficient variance in IQ (SD ϭ 10.6, compared with SD ϭ 16 in the general population) in the sample to have expected a much greater difference. More predictive than IQ of whether a mentally gifted Terman subject grew up to be an accomplished professor, lawyer, or doctor were particular noncognitive qualities: ‘Perseverance, Self-Confidence, and Integration toward goals’ (Terman & Oden, 1947_77ya, p. 351). Terman and Oden, who were close collaborators of Cox, encouraged further inquiry into why intelligence does not always translate into achievement: ‘Why this is so, what circumstances affect the fruition of human talent, are questions of such transcendent importance that they should be investigated by every method that promises the slightest reduction of our present ignorance’ (p. 352).

…The cross-sectional design of Study 1 limits our ability to draw strong causal inferences about the observed positive association between grit and age. Our intuition is that grit grows with age and that one learns from experience that quitting plans, shifting goals, and starting over repeatedly are not good strategies for success. In fact, a strong desire for novelty and a low threshold for frustration may be adaptive earlier in life: Moving on from dead-end pursuits is essential to the discovery of more promising paths. However, as Ericsson & Charness 1994_30ya demonstrated, excellence takes time, and discovery must at some point give way to development. Alternatively, McCrae et al 1999_25ya speculated that maturational changes in personality, at least through middle adulthood, might be genetically programmed. From an evolutionary psychology perspective, certain traits may not be as beneficial when seeking mates as when providing for and raising a family. A third possibility is that the observed association between grit and age is a consequence of cohort effects. It may be that each successive generation of Americans, for social and cultural reasons, has grown up less gritty than the one before (cf. Twenge, Zhang, & Im, 2004_20ya).

“Factors Affecting Entrapment in Escalating Conflicts: The Importance of Timing”, Brockner et al 1982

All subjects were given an initial monetary stake and had the opportunity to win more by taking part in an entrapping investment situation. In Experiment 1, half the subjects were provided with a payoff chart that made salient the costs associated with investing (High-cost salience condition) whereas half were not (Low-cost salience condition). Moreover, for half of the subjects the payoff chart was introduced before they were asked to invest (Early condition) whereas for the other half it was introduced after they had invested a considerable portion of their resources (Late condition). Entrapment was lower in the High salience-Early than in the Low Salience-Early condition. However, there was no difference between groups in the Late condition. In Experiment 2, the perceived presence of an audience interacted with personality variables related to face-saving to effect entrapment. When the audience was described as ‘experts in decision making,’ subjects high in public self-consciousness (or social anxiety) became less entrapped than those low on these dimensions. When the audience consisted of individuals who ‘wished simply to observe the experimental procedure,’ however, high public self-consciousness (or social anxiety) individuals were…more entrapped than lows. Moreover, these interaction effects occurred when the audience was introduced late, but not early, into the entrapment situation. Taken together, these (and other) findings suggest that economic factors are more influential determinants of behavior in the earlier stages of an entrapping conflict, whereas face-saving variables are more potent in the later phases.

…For example, individuals may ‘throw good money after bad’ in repairing an old car, remain for an excessively long period of time in unsatisfying jobs or romantic relationships, or decide to escalate the arms race (even in the face of information suggesting the impracticality of all these actions) because of their belief that they have ‘too much invested to quit’ (Teger, 1980_44ya).

“Face-saving and entrapment”, Brockner 1981_43ya:

Entrapping conflicts are those in which individuals: (1) have made substantial, unrealized investments in pursuit of some goal, and (2) feel compelled to justify these expenditures with continued investments, even if the likelihood of goal attainment is low. It was hypothesized that entrapment (ie. amount invested) would be influenced by the relative importance individuals attach to the costs and rewards associated with continued investments. Two experiments tested the notion that entrapment would be more pronounced when costs were rendered less important (and/or rewards were made more important). In Experiment 1, half of the subjects were instructed beforehand of the virtues of investing conservatively (Cautious condition), whereas half were informed of the advantages of investing a considerable amount (Risky condition). Investments were more than twice as great in the Risky condition. Moreover, consistent with a face-saving analysis, (1) the instructions had a greater effect on subjects with high rather than low social anxiety, and (2) individuals with high social anxiety who participated in front of a large audience were more influenced by the instructions than were individuals with low social anxiety who participated in front of a small audience. In the second experiment, the importance of costs and rewards were varied in a 2 × 2 design. As predicted, subjects invested statistically-significantly more when cost importance was low rather than high. Contrary to expectation, reward importance had no effect. Questionnaire data from this study also suggested that entrapment was at least partially mediated by the participants’ concern over the way they thought they would be evaluated. Theoretical implications are discussed.

Disagreeing with Brockner 1981_43ya on the social concern part; “Factors Affecting Entrapment: Justification Needs, Face Concerns, and Personal Networks”, Karavanov & Cai 2007_17ya:

The current investigation did not support the findings from previous studies that suggest that justification processes and face concerns lead to entrapment. This study found that only internal self-justification and other-positive face concerns are related to entrapment, but instead of contributing to entrapment, these aspects prevent individuals from becoming entrapped. Personal networks were demonstrated to have positive effect on both self- and other-positive face concerns, providing empirical support for the value of using personal networks as a predictor of face goals. However, personal networks did not contribute to entrapment.

Heath takes the use of ‘budget accounting’—which can lead to reduced total return, as it did for subjects in his experiments, who if they stuck it out and escalated commitments earned $7.35 versus the budget-users at $4.84—as often conflicting with normative standards. My own perspective is to wonder how much budget making resembles writing down one’s justification for a particular probabilistic prediction, a strategy which reduces hindsight bias when one’s predictions are ultimately falsified.↩︎

See, for example, Pala et al 2007 which investigated whether causal loop diagrams helped people prevent sunk cost more than being given “a list of important factors”. They didn’t.↩︎