RATIONAL CHOICE IN AN
UNCERTAIN WORLD, ROBYN M. DAWES,
Carnegie Mellon University Under the General Editorship of JEROME KAGAN Harvard
University Harcourt Brace College Publishers 1988
“My colleagues, we have
come too far to waver now, we have accomplished too much to give up now, we
have struggled too long to fail now.” Speaker of the House Thomas P.
O'Neill, Ir., arguing in favor of tax reform, September 25,1986
2.1 CONFUSION ABOUT
CURRENT MONETARY ASSETS: SUNK COSTS
Would you rather pay $100 to
be where you wanted to be or where you didn't want to be?[1]
The answer to this question seems obvious: Now let's rephrase the question.
You and your companion have
driven half-way to a resort. Responding to a reduced-rate advertisement, you
have made a nonrefundable $100 deposit to spend the weekend there. Both you and
your companion feel slightly bad physically and out of sorts psychologically. Your
assessment of the situation is that you and your companion would have a much more
pleasurable weekend at home. Your
companion says it is "too bad" you have reserved the room because you
both would much rather spend the time at home, but you can't afford to waste
$100. You agree. Further, you both agree that given the way you both feel, it
is extraordinarily unlikely you will have a better time at the resort than you
would at home. Do you drive on or turn back? If you drive on, you are behaving
as if you prefer paying $100 to be where you don't want to be than to be where
you want to be.
Look at the problem another
way. The moment you paid the $100 your net assets decreased by $100. That decrease occurred several days
before your drive half-way to the resort. Is the fact that your net assets have
decreased by $100 sufficient reason for deciding to spend the weekend at a
place you don't want to be?
The point is, you reiterate,
that if you turn back you will have wasted the $100; waste not, want not. Perhaps you are slightly obese due to
the same reasoning. Once you have paid for your food, you feel compelled to eat
it all in order to avoid wasting it - even though the outcome of that
particular policy is to decrease your dining pleasure and make you fat.
If the $100 could be
refunded, you would certainly return home; otherwise you must remain a slave to
the check you wrote (or to your credit card number). What does that imply? Once
the deposit was made, you had a certain net asset level. Now, if you could only
get a refund and thereby increase your net assets by $100, you would be willing
to do what you prefer to do - otherwise not. That's rational?
The $100 you have already
paid is technically termed a sunk
cost. Rationally, sunk costs should
not effect decisions about the
future.
When you behave as if your
nonrefundable deposit is equivalent to a current investment, you are honoring a
sunk cost. You certainly would not send the resort a check for $100 for the
privilege of staying home. That, too, would be irrational, because you would
simultaneously conclude that $100 is worth more to you than nothing, a price
for which you could remain at home. You are not paying now for the Opportunity to go to the resort, although at
the time you did so it might have
been a very wise decision. Now, however, the only choice available that avoids
the contradictions specified earlier is the one you judge to be the more
valuable - turning back. Honoring sunk costs is irrational. (I'm excluding the
possibility that you have motives other than enjoying yourself for going to the resort, or that you wish to
create the impression you are at the resort when you actually are not. The
information presented in the examples in this book is to be taken as the total
information available to the decision maker. Naturally, if there is other
information, or if there are other reasons for engaging in a behavior that are
not specified in the examples, then the choice might be different.)
People honor sunk costs:
"To terminate a project
in which $1.1 billion has been invested represents an unconscionable
mishandling of taxpayers'
dollars.” - Senator Jeremiah Denton, November 4, 1981
"Completing
Tennessee-Tombigbee is not a waste of taxpayers' dollars. Terminating the
project at this late stage of development would, however, represent a serious
waste of funds already invested.” – Senator James Sasser, November
4, 1981[2]
Both senators were responding
to critics who had pointed out that the total value of the Tennessee-Tombigbee
Waterway Project, if completed, would be less than the amount of money yet
to be spent completing it. Of course,
these statements may be mere rationalizations for continued expense that will
create employment in the state. The point, however, is that both senators
believed that the arguments presented were compelling - or they would not have
made them. (I'm assuming that politicians do not present public arguments they
believe the listeners will categorize as inane or stupid.) I am criticizing the
logic of the argument, not the possibility of ulterior motives for presenting
it.
The irrationality is clear.
The federal budget deficit was $1.1 billion more than it would have been had
the project not been started. That is used as a justification for spending
money to create something worth less than the money yet to be spent. Clearly,
like lost lives, dollars must not be spent in vain. Limiting concern to the future consequences of choices is the best way to avoid
honoring sunk costs. Conversely, honoring sunk costs violates the first
criterion of rationality - that decisions should be based only on future consequences.
As illustrated by the example, such violation yields contradictions.
2.2 THE SUNK COST PROBLEM:
OTHER ASSETS
If the irrational
"payment" of sunk costs were limited to money, it might be possible
to educate business and laypeople to ignore them Irrationality fails, for
reasons we discussed in Chapter 1; and the rational abandonment of sunk costs
often succeeds - as, for example, when Lockheed stock jumped 7 ¾ points
in 1981 the day after it finally abandoned its thirteen-year-old Tristar L1011
program into which it had sunk $2.5 billion.[3]
Concern with sunk costs is
pervasive in other domains of our lives as well. We do not wish to invalidate the past, particularly our own pasts, by ignoring
sunk costs. It is not just vanity that keeps us from admitting that we have
invested our past resources - monetary and psychic - in a venture that we
should now abandon. We wish to view our lives as cohesive. People who turn back
due to discomfort without being affected by the fact they have already paid the
resort $100 may be regarded not just as financially wasteful, but as disordered
or silly. Businesspeople who abandon an unprofitable line may be regarded not
as wise, but as vacillating - certainly not the "dynamic" type we all
admire.
For example, Ford Motor Company wisely abandoned the Edsel as not suitable to American taste and later marketed the popular Mustang. In the 1964 presidential elections, however, the Republican candidate, Barry Goldwater publicly chided the former president of Ford, then the Secretary of Defense, for having first promoted and then abandoned the Edsel - even though it could equally well be maintained that the Edsel venture provided Ford with invaluable information that led to the tremendous success of the Mustang. This same Secretary of Defense showed a much greater commitment to a sunk cost in Southeast Asia - as did the subsequent Secretary of State, who wrote: "We could not simply walk away from an enterprise involving two administrations, five allied countries, and thirty-one thousand dead as if we were switching off a television channel."[4] The kindest interpretation of these commitments is that because the leaders of other nations honor sunk costs, the United States would have suffered a severe blow to its prestige had it failed to do so.
If, indeed, abandonment of a
sunk cost negatively affects reputation, then it may be wise not to do it.
Reputational damage, however, leads to future problems, and it is only with respect to such future
consequences that honoring sunk costs should make a difference. If the future
cost to reputation outweighs the cost of paying a sunk cost, it is rational to
pay it.
Consider, now, our beliefs
about addictions. "Knowledgeable" people confidently maintain that it
is impossible to stop the course
of an addiction until the addict has reached "the bottom of tbe
bottle." Why? To my knowledge, no studies, have demonstrated this. In
fact, studies that are conducted examine people who are already in the advanced stages of addiction - perhaps comparing
them to people who claim not to be addicted at all - so it is logically
impossible to estimate the number of people who stop somewhere along the way.
Of course it is possible to study people prospectively; such a study would begin with people who are not
addicted but might become so (anyone - or perhaps people presumed to have a
genetic disposition to an addiction); then, and only then, would it be possible
to see whether people stop before reaching "the bottom," and how many
people do so. Such studies are rare. The fact that such studies are
"classics" when
they are done - for example, those by George Vaillant - attests to their
rarity. (Asserting that no one stops half-way is like the assertion discussed
in Chapter 1 that child abuse will not stop without intervention. Given that
these colleagues' evidence is based entirely on contact with a
"clinical" population - that is, with people who haven't stopped - it
is logically impossible to reach this conclusion.)
So why the belief? Consider
the problem from the point of view of the person wishing to kick an addiction.
By definition, the addiction has been ongoing and has led to distress and
suffering - or the person would not wish to stop. If, however, it is possible
to stop now, why was it not possible to stop earlier, for example, before
trouble arose? That's a reasonable question, which remains unanswered unless being at rock bottom is a necessary precondition for
stopping. Conversely, if it was not possible to stop earlier, why should it be
possible now? Again, without the belief, that question is perfectly reasonable
- and perhaps haunting.
Consider, now, the assessment
of the addicted person who does not believe in the necessity of reaching the
bottom but who has some hope of quitting:
1. I can quit now.
2. I could have quit earlier,
but didn't.
3. My addiction has caused me
- and perhaps others – significant distress, or worse.
This is a very painful set of
beliefs. From a sunk cost perspective, the past has been "wasted" if
the addict does now indeed quit. One possible way of justifying such costs is
to conclude that one is simply a victim, couldn't quit, and can't (a not very
petty or pretty consistency).
In contrast, the assessment:
1. I can quit now.
2. I couldn't quit earlier
because I hadn't reached rock bottom.
3. My addiction has caused me
- and perhaps others – significant distress, or worse.
provides the addict with no
sunk costs. In fact, previous addicted behavior is now assessed as a benefit, because it put the addict in a position to quit.
("Thank God, I'm an alcoholic!") I am not maintaining that the
"bottom of the bottle" hypothesis is false, just that the evidence
for it is not there, and that it is explicable in terms of our irrational need
to justify sunk costs (and may therefore be "therapeutic" even if
false).
Of course, there is an
alternative assessment:
1. The past cannot be
changed.
2. Today is the first day of
the rest of my life.
3. Trying to quit can't hurt
and might help.
A final explanation for our
devotion to sunk costs lies in our belief not just in coherent behavior but in committed behavior. ("Love is not love which alters when
it alteration finds." ------ Shakespeare.) If I have made sacrifices for a
reason, I should continue doing so. To behave otherwise would be cowardly. But
while love may not be love that alters when it alteration finds, behavior that
doesn't is just plain stupid.
2.3 THE RATIONALITY OF
CONSIDERING ONLY THE FUTURE
Emphasis on merely ignoring
sunk costs has arisen only with modern decision theory, which in turn is based
on probabilistic thinking that arose in the Italian Renaissance. This thinking
is based on the idea that probabilities can be assessed properly only with
reference to future events. For
example, consider a fair coin that has been tossed four times and is to be
tossed a fifth time. The probability of its landing heads is ½. The
pattern of previous results is irrelevant because they have already occurred
and do not affect the way in which the coin is handled when it is tossed for
the fifth time. For example, four previous heads do not make a fifth head
unlikely even though in general four heads and a tail is an outcome five times
more likely than five heads. Given four heads have already occurred, a fifth
head is as likely as a tail.
That the idea of limiting
such probability assessment to future possibilities was not intuitively obvious
prior to the Italian Renaissance can be inferred from answers proposed to a
famous problem in Fra Luca Paccioli's Summa de arithmetica, geometrica,
proportioni e proportionalita, published in 1494. The problem: "A and B
are playing a fair game of balla. They agree to continue until one has won a
total of six rounds. The game actually stops when A has won five rounds and B
three rounds. How should the stakes be divided?"[5]
Paccioli's answer: 5:3.
One problem with this answer
- dividing the stake proportionally to the number of games won - is that it
implies A should get the entire stake whether he has won one, two, three, four,
or five games in a row, although he clearly is in a much better position the
more simultaneous games he has won. Further, it implies that there would be a
sudden discontinuity in the rule for dividing the stake if A were to move from
leading 5:4 to winning 6:4 (in which case the stake is split 1:0). Moreover, it
implies that A is more deserving when he is ahead 2 to 1 than ahead 5 to 3,
even though it is clear that he has a much better chance of winning six games
from the
latter lead. Proposals based
on the differences between the number of games won have, mutatis mutandis, the same problems.
It was not until 64 years
later that G. F. Peverone proposed a solution that doesn't have any of the
problems listed above (or others). According to Peverone's solution:
1. The more consecutive games
won the higher the proportion of the stake;
2. The same rule is applied
once a player has won six games as is applied earlier; and
3. A player ahead 5 to 3
receives a higher proportion of the stake than does a player ahead 2 to 1.
The solution is based on two
principles. First, where p is the
probability that A will be the first person to win six games, p is the proportion of the stake that should be given
to A. Second, p is computed by
analyzing the number of games remaining before A or B wins a total of six.
Peverone actually miscomputed p.
The correct computation begins by noting that when A is ahead 5 to 3, the only
way B can win six first is to win three consecutive games. Since the game is
fair, that probability is (1/2)x(1/2)x(1/2), or 1/8. Hence, A's probability of
winning is 7/8; A should receive 7/8ths of the stake. (Peverone concluded
5/6ths.)
Similar calculations can be
used to determine A's proportion of the stake when A has won five
consecutive games, when A is
ahead 2 to 1, and so on. When, of course, A has won six, he has the probability
of 1 of having won, and hence, receives the whole stake.
In general, the past is
relevant, only for estimating current probabilities and the desirability of
future states. It is rational to conclude that a coin that has landed heads in
19 of 20 previous flips is probably biased, and that therefore the probability
it lands heads on the 21st flip is greater than 1/2. It is not rational to
estimate the probability of a head on the 21st toss by assigning a probability
to the entire pattern of results including those that have already occurred.
(Again, the probability of five straight heads when tossing a fair coin is
1/32; in contrast, the probability of a fifth head given four heads in the past is 1/2.) Rational
estimation of probabilities and rational decision making resulting from this
estimation are based on a very clear demarcation between the past and the
future.
That people do not naturally
make this demarcation in their own lives is evidenced by the recent success
(circa 1958 on) of attribution theory
in social psychology. Briefly,
attribution theory examines the conclusions people reach about their own and
other people's characteristics and the effects of these conclusions on future judgment
and choice - as one early attribution theorist put it, "what we say to
ourselves about ourselves and what that does to ourselves"
(and to others). For example,
one finding of this theory is the "fundamental attribution error," by
which we attribute others' behavior largely to personality factors and our own
behavior largely to situational factors to which we respond.[6]
Thus, for example, a candidate for parole who explains the situational factors
leading to his crime may be judged either to be Iying or to be lacking insight
into his character defects, even though he is truthfully explaining the reasons
for the crime as he views them. In contrast, a candidate who duplicitously
claims to have seen the reasons as residing in his own past antisocial characteristics
- which have now changed (e.g., through a religious conversion) - may be
believed.
The relevance of attribution
theory is that attributions ("schemas") are not of the form
"When
confronted with this type of situation in the past, I have (he has) done X;
therefore, it is probable that in the future in this situation ...."
but of the form
“When
confronted with this type of situation in the past, I have (he has) done X;
therefore, I am (he is) the type of person who will ...."
This distinction is not
trivial. If, for example, I have acted weakly in a situation in the past, it is
rational to conclude that I
have a high probability of acting weakly again in a similar situation in the
future. But that is a half-full/half-empty glass, because this probability
assessment may in fact forewarn me to take precautions against acting weakly. If, in contrast, I conclude that since
I have acted weakly in the past I am a weak person, then I have woven myself a
verbal straightjacket that may actually increase the probability I behave in a
way I wish not to behave.
"Whoever
observes himself arrests his own development. A caterpillar that wanted to know
itself well would never become a butterfly." Andre Gide, Les Nouvilles
Nourritures, 1935.
“Life
could not continue without throwing the past into the past, liberating the
present from its burdens." Paul Tillich, The Eternal Now
Attributions of stable
characteristics - which forge an inappropriately strong link between past and
future - can easily become self-fulfilling prophecies. (Not all prophecies are
self-fulfilling; there must be a mechanism. Some prophecies may even be self-negating - for example, the prophecy
of reckless drivers that "nothing bad can happen to me.")
The point here is that
attribution theory describes some significant aspects of human behavior quite
well. We continually make attributions about ourselves and others. For example,
when asked to describe themselves, first-graders in one U. S. school system
used roughly 86% action verbs (describing what they do) and only 14%
state-of-being verbs (describing what they are). The percentage of being verbs
used increased steadily over grade levels until the ratio was 50-50 by the
senior year of high school.[7]
The unfortunate thing about such attributions is that they not only reflect
perceived sunk costs ("I am; therefore I must continue"), but that
they are highly malleable as well. Social psychology experimenters have been
able to change subjects' attributions (at least temporarily) with minimal manipulations
such as mumbling "good job," "hmmm," "that's not quite
as good as average," and the like. More dramatic manipulations have
convinced male subjects that they were more attracted to some Playboy
"playmates" than to others by giving them phony feedback about their
heart rates as they viewed the magazine pictures,[8]
or that they have "homosexual tendencies" by having a needle that
supposedly indicates unconscious libidinal tendencies deflect sharply when they
view a photograph of a nude man. (Such experimentation is now considered
unethical.)
One possible implication is
that people with negative self-attributions (for example, "I am a
vulnerable, sick, weak,
incompetent person") should not be encouraged to develop positive ones
("I am an invulnerable, well, strong, competent person," at which
point the changed person is elected President and starts a nuclear war), but to
abandon the self-attribution process altogether." Shut up to yourself
about yourself. What you say might be a trap, and it probably isn't true
anyway." Given that we have been constantly bombarded with verbal
descriptions - usually evaluative - of ourselves throughout our lives, that's a
hard prescription.
In summary, rational
decisions are based on an assessment of future possibilities and probabilities.
The past is relevant only insofar as it provides information about possible and
probable futures. Rational decision making demands the abandonment of sunk
costs, unless such abandonment creates future problems outweighing the benefits
of abandonment. Today really is the first day of the rest of our lives.
Before concluding this
chapter, it is necessary to point out again that any "loss of face"
involved in abandoning sunk costs may be a real cost for the decision maker. In
fact, such a change in reputation is a future cost that one might choose to
bear in abandoning a sunk cost. The auto maker who abandons the Edsel may be
derided for making a "gutless" decision. The alcoholic who quits
drinking without finding God may be reviled by drinking companions and viewed
with suspicion by the "therapeutic establishment." Such costs are
perfectly reasonable costs to consider in determining whether or not to abandon
a particular course of action. The sunk cost per se should not be a factor,
however. So long as other people believe in honoring sunk costs, the person who
does not may be regarded as somewhat aberrant. (A student once stated that
after taking a course on decision making she found it more difficult to discuss
decisions with her friends, and sometimes her friends thought her something of
a kook.) The
point is to deal with the
possibility of mindless social disapproval as a potential consequence of your
decisions. (A suggestion for doing so is given at the end of Chapter 3.) And as
will be argued in Chapter 11, our concern with derogation for abandoning sunk
costs may be grossly exaggerated anyway. It is not rational to honor sunk costs
mindlessly.
* * * * * * * *
To end this chapter on a
clearly upbeat note, I will point out that social problems that arise after
abandoning a sunk cost can be
ameliorated by a type of framing. The framing consists of explaining that one
is not forsaking a project or enterprise, but rather wisely refusing "to
throw good money after bad." Rationally, that is exactly what is involved
in abandoning a sunk cost, and of course it involves forsaking a project or
enterprise. Using this phrase, moreover, tends to enhance the credibility of
the speaker, who is then relieved of the necessity to explain in any detail the
irrationality of honoring such sunk costs.
This "good money after
bad" framing focuses the listener's attention on the present as the status
quo and phrases the abandonment of a sunk cost as the avoidance of a sure loss (which is good). In contrast, honoring
a sunk cost involves framing a past state as the status quo and abandoning it
as the acceptance of a sure loss (which is bad). As pointed out throughout this
chapter, frames are malleable. The person who abandons a sunk cost benefits
from behaving rationally and if the present is effectively framed as the status
quo, he or she also enjoys the approval of others. Remember that President
Kennedy achieved the height of his popularity shortly after he abandoned the
Bay of Pigs venture.
[1] I am indebted to my daughter Molly Dawes, who was 15 years old at the time, for this phrasing of the
question.
[2] See Arkes,
H. R., and Blumer, C. (1985). The psychology of sunk cost. Organizational
Behavior and
Human Performance, 35, 129-140.
[3] Referenced in Staw, B. M. (1982). Counterforces to change. In Goodman, P S. (ed.). Change in Organizations. San Francisco: Jossey-Bass, 87-121.
[4] Referenced in Howard, M. (1984). The Causes of War. Cambridge: Harvard University Press, 232.
[5] The history
of this problem is discussed in Chapter 4 of David, N. (1962), Games, Gods,
and Gambling. London: Charles Griffin.
[6]Jones, E. E., and Nisbett, R. E. (1972). The actor and the observer:| Divergent perceptions of the causes of behavior. In Jones, E. E.; Kanouse, D. E.; Kelley, H. H.; Nisbett, R. E.; Valins, S.; and Weiner, B. (eds.), Attribution: Perceiving the Causes of Behavior. Morristown, N.J.: General Learning Press.
[7]McGuire, W. J. (1984). Search for the self: Going beyond self-esteem and the reactive self. In Zucker, R. A.; Aronoff, J.; and Rabin, A. I. (eds.), Personality and the Prediction of Behavior. New York: Academic Press 73-120.
[8]Valins, S. A. (1966). Cognitive effects of false heart rate feedback. Journal of Personality and Social Psychology, 7, 345-350. In fact, male heart rate decreases (and peripheral blood volume increases) with sexual arousal, but - most people don't know that, and the subjects were presented with phony feedback that their heart rate increased more when they viewed some pictures than when they viewed others.