4.

1 The neoclassical model Assumptions In the previous chapter we have examined

how people form attitudes, values, prefer ences, and finally make choices. In terms
of the standard model described in equation (1.1), these aspects are largely related
to component (4), involving utilities, and, when it comes to choices, component (1)
involving maximization. However, various assump tions were made at that stage
regarding the options and outcomes of these options in the decision-making
process. The factor we want to focus on in this chapter is the certainty of these
outcomes. Therefore the component we now need to examine is (3), related to
probabilities or beliefs. As far as beliefs are concerned, the main assumptions in the
neoclassical model (NM) are that decision-makers have perfect as opposed to
bounded rationality, and that they are Bayesian probability estimators. Let us
explain each in turn: 1 Perfect rationality This means that people not only have all
the relevant information pertaining to a deci sion, but have the cognitive resources
to process it instantly, accurately and costlessly. If this is not the case, and it is
obviously unlikely in most real-life situations, then we can say that there is bounded
rationality. This term was introduced by Simon (1955), who was the first researcher
to emphasize its implications for decision-making. The most general implication is
that we tend to use heuristics in many decision-mak ing situations; these are
‘methods for arriving at satisfactory solutions with modest amounts of computation’
(Simon, 1990). The term heuristic was originally introduced in psychology to refer to
simple processes that replace complex algorithms (Newell and Simon, 1972), and
has become extended now to include any decision rules that we implement as
shortcuts to simplify and or accelerate the decision-making pro cess. A good
example is to never order the lowest-price or highest-price items on the menu in a
restaurant. This might imply that the decision-maker believes that nei ther of these
items represents good value. As we shall see, there are a large number of heuristics
described in the behavioral literature, and indeed some have proposed that there
are too many. Shah and Oppenheimer (2008) propose that there is much
redundancy in the field of heuristics, with different names for similar and
overlapping concepts, too much domain-specificity, and insufficient attention paid
to the overrid ing principle that heuristics are effort-reducing mechanisms. We shall
return to these issues at various points throughout the following chapters. The most
significant implication of using heuristics is that they often result in biases, meaning
systematic errors. In terms of beliefs these errors are factual; biases can also occur
in terms of preferences, where the errors may result in non-optimal choice. 118
beliefs anD exPeCtations CH•4 2 Bayesian probability estimation This means that
people are able to estimate probabilities correctly, given the relevant information,
and in particular are able to update them correctly given a sequence of prior
outcomes. The interpretation and significance of this will be examined in the section
on the law of small numbers, along with deviations in estimation or biases, but a
simple example will suffice at this stage. When a coin is tossed several times and
comes up heads each time, a correct Bayesian updater will still estimate the
probabil ity of heads on the next coin toss as being 0.5, since the prior outcomes
have no effect on the next outcome in this situation. However, many people tend to
incorrectly assume that the prior outcomes do affect the probability of the next
outcome here (as it would in other situations), and estimate the probability of the
next outcome being a head as less than 0.5. This is an example of a ‘mean-
reverting’ regime resulting in the ‘gambler’s fallacy’. Both of these terms are
explained in the third section related to the law of small numbers. The Bayes
formula in general terms is as follows: P(A|B)= P(B|A) P(A) P(B) (4.1) This formula
can be used to estimate probabilities of the truth or falsehood of events that are not
random, but where the truth is unknown, such as when a die has been thrown
within a cup – there is a definite outcome, but until the cup is removed we do not
know what it is. Bayes’ theorem updates or modifies probabilities, given new pieces
of evidence, in the following way: P(H|E)= P(E|H) P(H) P(E) where • H represents a
specific hypothesis, which may or may not be some null hypothesis. • E represents
the evidence that has been observed. • P(H) is called the prior probability of H that
was inferred before new evidence became available. • P(E|H) is called the
conditional probability of seeing the evidence E if the hypothesis H happens to be
true. It is also called a likelihood function when it is considered as a function of H for
fixed E. • P(E) is called the marginal probability of E: the a priori probability of
witness ing the new evidence E under all possible hypotheses. It can be calculated
as the sum of the product of all probabilities of any complete set of mutually
exclusive hypotheses and corresponding conditional probabilities: P(E) = P(E|Hi )
P(Hi ) • P(H|E) is called the posterior probability of H given E and is the new esti
mate of the probability that the hypothesis H is true, taking the evidence E into
account. The factor P(E|H) / P(E) represents the impact that the evidence has on the
belief in the hypothesis. The interpretation of this factor, and an example of an
application of all of the above concepts, is given in the discussion of base rate bias
in the next section. 119 PT• II FOUNDATIONS 4.2 Probability estimation The types of
deviation described in this section relate to rational Bayesian updating. There are
various aspects of bias here. Before discussing these, there is a general point to be
clarified. It is often assumed by researchers that a bias implies faulty reasoning.
This is not necessarily true, since systematic errors in probability estimation can
occur because of random noise interfering with information processing (Costello and
Watts, 2014). However, many researchers support the existence of dual process
theories popu larized by Kahneman and Frederick (2002, 2005), where cognitive
processes involve two separate systems, frequently referred to as system 1 and
system 2 (Evans 2011; Evans and Stanovich, 2013). System 1 relates to intuitive
aspects, dominated by perception, where beliefs and judgments are formed quickly,
on the basis of heuristics. It is useful to think of these as being shortcuts to decision-
making, often based on rules-of-thumb. System 2 relates to reasoning, a slower,
more conscious and calculating process. The main con sequence of this duality is
that system 1 can lead to biases in judgment, which system 2 fails to correct in
many cases. As an example, consider the following problem: a bat and a ball
together cost £11, and the bat costs £10 more than the ball. How much does the
ball cost? A majority of people quickly answer £1. This is an incorrect answer, which
system 2 frequently fails to adjust. We will now consider a number of common
heuristics and their consequent biases. The availability heuristic People are often
lousy at estimating probabilities of events occurring, especially rare ones. They
overestimate the probability of dying in plane crashes, or in pregnancy, or suffering
from violent crime. An often-quoted example of overestimating low probabilities con
cerns playing the lottery. The California lottery, one of the biggest in the world,
requires matching six numbers between 1 and 51 in order to win the main prize.
The odds against doing this are over 18 million to one. In other words, if one played
this lottery twice a week, one could expect to win about every 175,000 years. It was
found by Kahneman, Slovic and Tversky (1982) that people overestimated the odds
of winning by over 1,000%. In many of their papers Kahneman and Tversky have
suggested that people use an availability heuristic when estimating probabilities.
This means that people believe that events are more frequent or more probable if
examples of it are easier to remember. In general, this heuristic works reasonably
well because it is easier to recall examples of events that happen more frequently.
A type of availability bias occurs when people judge the probability of a category in
a population occurring. Research has found that a category size bias exists in which
‘consumers perceive an outcome as more likely to occur when it is categorized with
many rather than few alternative possibilities, even when the grouping criterion is
irrelevant and the objective probability of each outcome is identical’ (Isaac and
Brough, 2014). For example, participants in one study irrationally predicted being
more likely to win a lottery if their ticket color matched many (vs. few) of the other
gamblers’ tickets, and wagered nearly 25% more as a result. The main source of
error with the availability heuristic is salience; this factor features in other types of
bias also, but the main effect here is that events that have been well pub licized or
are prominent in people’s memories tend to be estimated as having exaggerated
probabilities. Thus it has been found that there is an increased purchase of
earthquake insurance following a recent quake, in spite of the fact that such events
will be less likely to recur in the short-term future. This error may be compounded
by the effect proposed by Vosgerau (2010), related to misattribution due to arousal,
sometimes referred to as the affect heuristic (Pachur, Hertwig and Steinmann,
2012). 120 beliefs anD exPeCtations CH•4 Both of these heuristics may be relevant
in decision-making when high stakes are involved compared with low stakes. One
would expect rational decision-makers to use more system 2 processes and rely
less on heuristics when stakes are higher. However, a study by Freling, Saini and
Yang (2012) has found contrary evidence. They propose that when stakes are
higher people rely more on anecdotal evidence than objective statistical
information, since it tends to have more emotional impact and salience, and also
tends to be easier cognitively in terms of processing. The combination of the
availability and affect heuristics can have serious impacts on government policy, so
that governments may devote too few resources to big problems and too many to
small problems (Sunstein, 2002). Thus a government may engage in a knee-jerk
reaction in response to public fears relating to a contagious disease or terrorism
threat, but spend insufficient funds on key infrastructure or disease prevention. The
representativeness heuristic In general the representativeness heuristic refers to
the phenomenon that global judg ment of a category is determined primarily by the
relevant properties of a prototype (Kahneman and Tversky, 1972, 1973; Tversky
and Kahneman, 1971, 1983). This means that people have the tendency to evaluate
the likelihood that a subject belongs to a certain category based on the degree to
which the subject resembles a typical item in the category. Although this strategy
may be effective in certain circumstances, the basic principles of probability and set
theory are often ignored by people in making judgments involving
representativeness. An illustration of this phenomenon is where respondents are
given a description of a personality of a woman, Linda, who has the characteristics
of a typical feminist. The majority of respondents rank the statement ‘Linda is a
bank teller’ as less likely than the conjunctive statement ‘Linda is a bank teller and
an active mem ber of the feminist movement’ (Tversky and Kahneman, 1983). In
this case the strong representativeness of feminism overcomes the basic probability
rule that P(A and B) can never be higher than P(A). The difficulties that people have
in reasoning related to connectives and conditionals has been observed in several
studies (Johnson-Laird et al., 1992, 2000). A different example of the representation
heuristic, which is applicable to financial markets, concerns investor behavior and
reactions to new information. According to a model developed by Gennaioli, Shleifer
and Vishny (2015) ‘investors overreact to a series of good news, because such a
series is representative of a good state. A few bad news do not change investor
minds because the good state is still representative, but enough bad news leads to
a radical change in beliefs and a financial crisis.’ At this point the bad state then
becomes the norm, and a few good news may be ignored. The authors claim that
this model generates ‘debt over-issuance, “this time is different” beliefs, neglect of
tail risks, under- and overreaction to information, boom-bust cycles, and excess
volatility of prices in a unified psychological model of expectations.’ Essentially, the
model adds a behavioral component to the Minsky hypothesis regarding the
relationship between debt and financial crises. It is helpful to give a further example
here, since it illustrates not only the rep resentation heuristic but also the base rate
bias, discussed further below. A woman, Jane, is described as having the following
characteristics: she likes aromatherapy, new age music, reads her horoscope
regularly and belongs to a spirituality group. When people are asked whether they
think Jane is more likely to be a holistic healer or a schoolteacher, they regularly
answer that she is more likely to be a holistic healer, since her personal
characteristics more closely match those of a holistic healer than a schoolteacher.
121 PT• II FOUNDATIONS Base rate bias The reason the response above is incorrect
is that it ignores the fact that the pro portion of holistic healers in the population is
much smaller than the proportion of schoolteachers. This is an example of base rate
bias, a common phenomenon again resulting from salience. In general terms base
rate bias means ignoring general infor mation relating to a population and focusing
on specific information which is more salient. A more complex example involving
conditional probabilities is given by Casscells, Schoenberger and Graboys (1978),
and relates to the problem of ‘false positives’. This involves a situation where a
person takes a medical test, maybe for a disease like HIV, where there is a very low
probability (in most circumstances) of having the disease, say one in a thousand.
However, there is a chance of a false prediction; the test may only be 95% accurate.
Under these circumstances people tend to ignore the rarity of the phenom enon
(disease) in the population, referred to as the base rate, and wildly overestimate the
probability of actually being sick. Even the majority of Harvard Medical School
doctors failed to get the right answer. For every 1,000 patients tested, one will be
actually sick while there will be 50 false positives. Thus there is only a 1 in 51
chance of a positive result meaning that the patient is actually sick. This example
can be explained in more detail using Bayes’ theorem. For simplicity, it is assumed
initially that if the patient has the disease the test returns a positive result 100% of
the time, meaning that there are no false negatives. Let A represent the condition in
which the patient has the disease, and B represent the evidence of a positive test
result. Then, the probability that the patient actually has the disease given the
positive test result is P(A|B)= = P(B|A)P(A) P(B|A)P(A) + P(B|not A)P(not A) 1 ×
0.001 1 × 0.001 + 0.05 × 0.999 = 0.0196 This means that the probability that a
positive result is a false positive is about 1 − 0.0196 = 0.98, or 98%. If, more
realistically, there is also a chance of the test return ing a false negative, this would
mean that P(B|A) < 1, and this would modify the result slightly. The difference
would be small, assuming that the chance of a false negative is low; for example, if
the probability of a negative result given the person has the disease is 0.99, then
P(A|B) = 0.0194. In general terms, if it is likely that the evidence E (a positive test)
would be observed when the hypothesis under consideration (the person is sick) is
true, but, when no hypoth esis is assumed, it is inherently unlikely that E would
have been the outcome of the observation, then the factor P(E|H) / P(E) will be
large. Multiplying the prior probability of the hypothesis, P(H), by this factor would
result in a larger posterior probability of the hypothesis given the evidence.
However, if P(H), the base rate, is very low, the posterior probability will still tend to
be low. Thus the consequence of base rate bias, meaning ignoring the base rate, is
that we tend to overestimate the probability of being sick, given a positive test.
Conversely, if it is unlikely that the evidence E would be observed if the hypoth esis
under consideration is true, but a priori likely that E would be observed, then the
factor would reduce the posterior probability for H. Under Bayesian inference,
Bayes’ theorem therefore measures how much new evidence should modify a belief
in a hypothesis. 122 beliefs anD exPeCtations CH•4 The ‘law of small numbers’ The
main error here is when people apply principles that apply to infinite populations to
small samples. We will examine the model described by Rabin (2002b). This model
examines the situation where people are observing a sequence of signals from a
process that involves independent and identically distributed (iid) random variables.
This means that each random variable has the same probability distribution as the
others and all are mutually independent. A simple example is a sequence of coin
tosses, where the probabil ity distribution is 0.5 for a head and 0.5 for a tail for each
toss, and the outcome of each toss has no effect on the outcome of any other toss.
The model assumes that people believe, incorrectly, that the signals are drawn from
an urn of finite size without replacement, whereas the correct assumption in this
case is that there is replacement after each draw from the urn. We now need to
examine the consequences of this incorrect assumption. 1 The ‘gambler’s fallacy’
effect This effect derives its name from the observation that gamblers frequently
expect a certain slot machine or a number that has not won in a while to be ‘due’ to
win. We find that that the effect occurs when the distribution of signals is known, as
it is with the coin toss situation. If an urn contains ten balls, five representing Up
and five representing Down, and one ball is drawn at a time with replacement, this
experiment is identical to tossing a coin. Thus if three successive draws all result in
an Up out come (equivalent to three heads in a row), then the rational person will
estimate the probability of an Up on the next draw as 0.5. However, if the person
believes that the balls are not being replaced, this means that there is only two Up
balls left in the urn out of seven balls in total, so they will estimate the probability of
the next draw being Up as only 2/7 or about 0.286, with the probability of Down
being 0.714. This is an example of the representativeness heuristic, in that the
sequence Up, Up, Up, Down is judged as being more representative of the
population than the sequence Up, Up, Up, Up. We shall encounter other examples of
this heuristic in the following chap ters. The ‘gambler’s fallacy’ is sometimes
referred to as the ‘law of averages’, in this case meaning that the number of Ups
should on average be the same as the number of Downs, given there is a 50%
chance of each event occurring. There is a variety of empirical evidence supporting
the existence of the ‘gambler’s fallacy’ effect. For example, New Jersey’s pick-three-
numbers game is a pari-mutuel betting system; this means that the fewer people
bet on a number, the higher is the expected payout. It has been found that the
amount of money bet on a particular number falls sharply after the number is
drawn, and only gradually returns to normal after several months (Clotfelter and
Cook, 1993; Terrell, 1994). There is an interesting explanation for this apparently
irrational phenomenon in terms of evolutionary psychology (Pinker, 1997). It is
proposed that in our past evolu tionary environment there was often good reason to
believe that a series of common outcomes would be likely to be broken at some
point. This was particularly true for meteorological events, like rain or sunshine. Of
course the expected length of the series would depend on the circumstances, but,
just as a cloud eventually blows past the sun, at some point the probability
becomes higher that in the next time period the sun will come out again. We shall
see that many of the biases that people have are based on evolutionary adaptations
or factors in our past. In the examples people are incorrectly inferring from a
sequence of identical signals, like three Ups, that the next signal or outcome will be
of a different type. However, there are many situations where people make exactly
the opposite infer ence, that the sequence will continue. This contradictory finding
is now described and explained. 123 PT• II FOUNDATIONS 2 The ‘hot hand’ effect
This effect derives its name from the mistaken belief among basketball players and
fans that a player’s chance of hitting a shot is greater following a hit than follow ing
a miss on the previous shot (Gilovich, Vallone and Tversky, 1985). Although it
appears that this ‘overinference’ is the opposite of the ‘gambler’s fallacy’, it is actu
ally a complementary effect, again involving a misapplication of the assumption of
non-replacement. The effect arises when there is uncertainty regarding the
distribution of signals, for example whether a stock price will go up or down in any
particular time period. It is instructive here to follow the example given by
DellaVigna (2009), involving a mutual fund with a manager of uncertain ability. This
time the situation involves two urns, each with ten balls; the well-managed fund has
seven Up balls and three Down balls, meaning the fund goes up in value seven
times out of ten, while the poorly man aged fund has three Up balls and seven
Down balls, meaning it only goes up three times out of ten. There is a prior
probability of 0.5 that the fund is well managed and a probability of 0.5 that the
fund is poorly managed, so that before we observe any draw of a ball from an urn it
is equally probable that the fund is well or poorly managed. Balls are then drawn in
sequence from an urn, but the investor does not know which urn they are drawn
from. After observing a sequence of three Up balls the investor has to compute the
probability that the urn drawn from was the one with seven Up and three Down
balls, which is equivalent to estimating the probability that the fund is well
managed after it has gone up three times in succession. The rational investor will
implement Bayes’ theorem to solve the problem, on the assumption that the balls
are replaced after each draw. Repeating the Bayes formula in Equation (4.1): P(A|
B)= P(B|A) P(A) P(B) Thus the rational investor computes the probability that the
mutual fund is well man aged as: P(Well|UUU) = P(UUU|Well) × .5 [P(UUU|Well) ×.5
+ P(UUU|Poor) × .5] This equals 0.73/(0.73 + 0.33) = 0.927. However, if the
investor behaves according to the law of small numbers and assumes that there is
no replacement after each draw, the Bayesian expression becomes: P(Well|UUU) =
(7/10 × 6/9 × 5/8)/[(7/10 × 6/9 × 5/8) + (3/10 × 2/9 × 1/8)] = 0.972. Thus we can
see that this type of investor will overinfer about the ability of the mutual fund
manager after three good performances. When the rational investor forecasts the
performance of the fund in the next period, they will calculate the probability of an
Up performance as 0.927 × 0.7 + (1 − 0.927) × 0.3 = 0.671. On the other hand,
the law-of-small-numbers investor, assuming they believe that the urn is
replenished after three periods, estimates the probability of an Up performance as
0.972 × 0.7 + (1 − 0.972) × 0.3 = 0.689, representing a perceived more probable
outcome. There are various studies relating to financial markets which provide
evidence for the ‘hot hand’ effect. Benartzi (2001) found that the degree to which
employees invest in their own firm’s stock depends strongly on the past
performance of the 124 beliefs anD exPeCtations CH•4 stock. In companies in the
lowest 20% of performance in the past ten years, 10.4% of employee savings were
allocated to the same firm’s stock, compared to 39.7% for f irms in the top 20%.
Overinference in stock holdings can cause predictability in returns, since investors
will tend to overinvest in stocks with high past returns, mak ing them overpriced
and reducing their later returns, as demonstrated by De Bondt and Thaler (1985). 3
Synthesis At this point it might appear that the contradictory effects of the
‘gambler’s fallacy’ and the hot hand are difficult to reconcile with each other.
Indeed, there appear to be some cases of situations where different people can
experience both effects. The lot tery is one of these; most players exhibit the
‘gambler’s fallacy’ in selecting numbers, like avoiding recent winners, but one
recent Chinese study finds the opposite, where players chase winners, even bearing
a cost to do so (Yuan, Sun and Siu, 2014). A study by Shleifer and Vishny (1998)
illustrates this phenomenon in a different environment, again demonstrating that
the law of small numbers can lead to both effects, causing both underreaction and
overreaction to market signals. In the short term investors follow the ‘gambler’s
fallacy’, believing that a series of identical sig nals, like the stock price rising, will be
followed by a fall (a ‘mean-reverting’ regime). Thus they do not invest in the stock
(underreact), causing it to be underpriced, and returns will continue to be high over
a short period of time, demonstrating positive correlation or momentum. However,
after a longer sequence, the investors overinfer, and expect a ‘trending’ regime,
whereby the stock is now expected to continue to rise. This ‘hot hand’ effect causes
overreaction, as investors now overinvest, making the stock overpriced, and
reducing returns, this time demonstrating negative correlation of returns in the long
term. There are other applications of the law of small numbers that help to solve the
apparent contradictions between the ‘gambler’s fallacy’ effect and the ‘hot hand’
effect. One of these again relates to the purchase of lottery tickets. As we have
seen, people often avoid betting on numbers in a lottery if they have recently won,
demonstrating a ‘gambler’s fallacy’ effect’. Yet there is evidence that people also
have an increased probability of buying their tickets from stores that sold winning
tickets the previous week; winning stores experience a 12–38% relative sales
increase in the week following the sale of a large-prize winning ticket. Guryan and
Kearney (2008) have investigated this ‘hot hand’ effect and propose an explanation
for the paradoxical combination of the two effects in lottery bet ting. They suggest
that: A belief in the hot hand comes not from the signals produced by the data-
generating process – as the representativeness explanation would require – but
rather from the characteristics of the data-generating process itself, namely
whether the data-generating process is perceived as having an animate or an inten
tional element. Research in psychology provides some support for this hypothesis
(Ayton and Fischer, 2004; Caruso, Waytz and Epley, 2010). The result of this is that
the hot hand may actually be real in certain situations. For example, Raab, Gula and
Gigerenzer (2012) found that in volleyball the hot hand existed for half of the
players in their study. Furthermore, this had policy implications for the other team
players in terms of allocation; it was beneficial to pass the ball to players perceived
to have the hot hand, and this increased their team’s winning chances. There have
also been two recent studies of professional sports that have confirmed the ‘hot
hand’ effect. In an analysis of results in tennis tournaments, Jetter and Walker
(2015) find that an additional win in the most recent ten matches raises the likeli
hood of winning by 3.2–3.4 percentage points. In a study of professional golfers
Rosenqvist 125 PT• II FOUNDATIONS and Skans (2015) find that those ‘making the
cut’ in top tournaments had a substantially better performance record afterwards
than players just failing to make the cut. In terms of a lottery situation, Guryan and
Kearney suggest that the selection of the balls does seem to involve a random
process, without any intentional element, and there fore the law of small numbers
would cause people to exhibit a ‘gambler’s fallacy’ effect as they expect a small
sample to resemble the underlying population; thus winning numbers are not
expected to occur again in the near future. However, with stores there could be a
human element in how winning stores are selected, leading to a ‘hot hand’. How
this human element operates in this case is open to speculation: it might be that
the store is chosen deliberately by the person buying the winning ticket; or the
location of the winning ticket could be attributable to a corrupt lottery
commissioner, bearing in mind that the winning store owner receives 1% of the
prize and thus has an incentive for bribery. Guryan and Kearney also note that the
‘lucky store’ effect is larger in areas with more high-school drop-outs, more people
living in poverty and more elderly. They suggest that this may be caused by
cognitive biases. Conformity bias This phenomenon occurs when people tend to
‘follow the crowd’ in deciding how to behave, rather than think independently. Thus
their judgment of probabilities depends on the behavior of others in a non-Bayesian
manner. The first research relating to this was by Asch (1951), and he published a
number of other studies in the 1950s elaborat ing on this. In the initial study groups
of eight students were shown two cards, the first showing a single line, and the
second showing three lines, where one matched the line on the first card in length,
while the other two lines were clearly shorter or longer. This is illustrated in Figure
4.1. The students were then asked in turn to state aloud which line on the second
card matched the line on the first card. Seven of the students, however, were
‘confederates’ of the investigator and briefed to give an identical answer, which
would sometimes be incorrect. The other non-confederate student was always
asked to give his answer last. In control groups, with no confederate pressure, the
error rate was very low, less than 1%. Under confederate pressure the average
error rate rose to 37%, with 5% of subjects always conforming, and 75% making at
least one error on the 12 critical trials. After various criticisms of the methodology
Asch refined this experiment, but the conformity bias remained. Other evidence
supporting this bias was provided by Figure 4.1 Asch experiment A B C Source:
Asch, S.E. 1955, p.3. Reproduced with permission. Copyright © 1955 Scientific
American, a Division of Nature America, Inc. All Rights reserved. 126 beliefs anD
exPeCtations CH•4 the notorious Milgram and Stanford Prison experiments
(Milgram, 1963; Zimbardo, 1972), where subjects were willing to impose pain on
others, under the auspices of obedi ence to authority. Since this early work there
has been much other evidence, both in experiments and in the field, where people
have displayed a herd mentality in following the behavior of others. The
phenomenon of conformity bias also has policy implications, for individu als, firms
and governments, and we will give an example here relating to an experiment
described in the previous chapter involving hotels desiring to reduce towel usage
(Baca Motes et al., 2013). As well as finding that wearing a lapel pin established
consistency and increased commitment among guests, it was also found that a
notice in rooms stating that the majority of guests reused towels had the same
effect of increasing the number of towels hung, even if the statement was actually
untrue. In a rational analysis of conformity and how to determine beliefs or
probabilities according to the beliefs of others, rational Bayesian updaters should
imitate others to some extent, but should also realize that this will result in a
correlation of beliefs. This should in turn cause them to ‘anti-imitate’ the behavior of
others. Eyster and Rabin (2014) show that any learning rule in which people
regularly imitate more than one per son without anti-imitating others will lead to a
positive probability of people converging to confident and wrong long-run beliefs.
Although at times it appears irrational, there are two sound bases for conformity
bias. The first lies in evolutionary psychology, which suggests that following the
crowd is an adaptive behavior since it aids survival in many situations, like fleeing
from a fire, a phenomenon frequently seen in various animal species. In terms of
human behav ior, a very important example relates to investor behavior in financial
markets. Herding behavior is frequently observed here, both in booms and in
crashes. Conformity bias is a major cause of this, and there are two reasons. First,
we may simply assume that others know something that we don’t and therefore it is
a good idea to follow them; and sec ond, since most large-scale investments are
carried out by professional fund managers who are responsible to their clients for
making a good return, it is better to be wrong with the crowd than to risk being
right but in the minority. The safer strategy for keep ing clients happy and keeping
one’s job is to follow the crowd. Another similar example concerns research and
academia. In all sciences there tend to be dominant models which are currently
favored by the mainstream, sometimes referred to as the current paradigm (Kuhn,
1970). Since the status of scientists depends on publications, it is safer to keep
within the broad confines of the existing paradigm when trying to achieve
publication. As time goes on, this envelope becomes stretched as existing models
struggle to explain new anomalies. At some point one or a few bold researchers will
introduce a new model, sig nificantly different from the previous one, and a new
paradigm may become established as further researchers find support for the new
model. In the early twentieth century this happened in physics, where ‘classical
physics’ was overturned by quantum mechanics and relativity theory. The second
basis for herding is that it means we do not have to use cognitive effort. Instead of
fully evaluating a list of options we use the simple heuristic of doing what others do,
like going to the same restaurant, buying the same sunshades or watching the
same movies. It should be noted that this behavior can have two possible
interpretations: (1) it implies a belief that the crowd has good reasons for its
behavior, and that if we follow it there is a greater probability of maximizing utility
or at least increasing it in terms of making a better decision; and (2) people get to
like things that other people like, thus receiving an ‘other-conferred’ utility.
Therefore conformity bias is relevant both for val ues, preferences and choices as
well as beliefs and expectations. Further discussion of this issue is given in Chapter
10 in the context of social preferences and neuroscientific evidence. 127 PT• II
FOUNDATIONS Roundness Some research has shown that in certain circumstances
consumers may have a preference for either odd or round numbers. Round numbers
may be preferred for simplicity, but in lottery situations odd numbers may be
perceived as more likely to occur, since they may be seen as more concrete and
realistic (Dehaene and Mehler, 1992; Sevilla and Bagchi, 2014). 4.3 Self-evaluation
bias This factor is sometimes described in terms of overconfidence, but we shall see
that, while overconfidence is an important aspect, there are other aspects, including
its opposite, underconfidence. In addition there is self-serving bias, which while it
often involves over confidence, can relate to other aspects of belief. We shall see
that visceral factors are relevant here also. Thus we use the term self-evaluation
bias as a general all-embracing term that includes all aspects of beliefs where some
kind of evaluation of the role of the self relative to a situation is involved.
Overconfidence It has been claimed that ‘No problem in judgment and decision
making is more preva lent and more potentially catastrophic than overconfidence’
(Plous, 1993). It is useful to distinguish between three different kinds of
overconfidence, since this helps to explain apparent inconsistencies in empirical
findings, and we will follow the classification proposed by Moore and Healy (2008).
This involves the concepts of overestimation, over placement and overprecision. 1
Overestimation This relates to overestimation of one’s actual ability, performance,
level of control or chance of success. Empirical evidence suggests that this is a
widespread phenomenon extending to many situations. People overestimate their
abilities to perform various tasks, overestimate how quickly they can finish a project
(which seems to be happen ing with this book!) and overestimate their faculty for
future self-control (examined in Chapter 8). Studies have documented this
overestimation of abilities in a wide vari ety of different professional fields: team
managers choosing players in the NFL pick (Thaler, 2016); political experts (Tetlock,
1999 and 2005); financial consultants (Hoelzl and Rustichini, 2005); investors
(Kadous et al., 2014); and CEOs (Malmendier and Tate, 2005, 2008). People can
also become overconfident in their abilities after viewing the performance of a
skillful master, for example in sports (Scopelliti, Botti and Donato, 2013) and they
can be unrealistically optimistic about their future pros pects (Buehler, Griffin and
Ross, 1994; MacDonald and Ross, 1999). Some researchers have attempted to
explain the results of studies indicating the existence of overconfidence in other
ways, in particular in terms of it being a statisti cal artifact caused by regression to
the mean (see Kahneman, 2012), but the majority suggest that it is caused by
underlying psychological factors. We shall also see that in some cases people
underestimate their abilities and are overly pessimistic. This will explained in the
section relating to underconfidence. The existence of both overcon f idence and
underconfidence represents something of a puzzle. The reason for the puzzle is that
it is obviously beneficial for individuals to be able to gauge their abilities accurately
from the biological viewpoint of survival and reproduction. This issue is discussed
further later in the section. 128 beliefs anD exPeCtations CH•4 2 Overplacement
This aspect of overconfidence is sometimes referred to as the ‘better-than-average’
(BTA) effect, since it relates to estimating one’s abilities relative to others rather
than in absolute terms. Again, this phenomenon has been documented by a large
number of studies in different areas of behavior. Respondents typically rate them
selves in the top 50% regarding driving ability (Svenson, 1981); ethics (Baumhart,
1968); managerial prowess (Larwood and Whittaker, 1977); productivity (Cross,
1997); health (Weinstein, 1980); skill in solving puzzles (Camerer and Lovallo,
1999); social skills (Swann and Gill, 1997); and physical abilities (Dunning,
Meyerowitz and Holzberg, 1989). A somewhat different example of overplacement
has been observed where people overestimate the amounts others are willing to
pay for things compared with oneself, with such varied items as chocolate truffles,
books, teddy bears, smoked salmon, sporting equipment, iPhones, artwork, gift
certificates, gambles, a trip to the moon, and a magic pill that confers the ability to
speak French (Frederick, 2012). As with overestimation, some researchers have
tried to explain these findings in terms of regression to the mean, a lack of financial
incentives, or people not caring about their reporting, but in general the findings
appear to be quite robust, present ing the same puzzle as mentioned above. 3
Overprecision This refers to excessive certainty regarding the accuracy of one’s
beliefs. Studies fre quently ask their participants questions with numerical answers
(e.g. ‘How long is the Nile River?’) and then have participants estimate confidence
intervals for their answers. Results show that these confidence intervals are too
narrow, suggesting that people are too sure they know the correct answer. For
example, Alpert and Raiffa (1982) found that a group of MBA students who were
asked for 98% confidence inter vals stated intervals that only contained the correct
answer 57% of the time instead of the expected 98%. Similar results have been
found in experimental studies by Klay man and colleagues (1999) and Soll and
Klayman (2004), and have been duplicated in the field in the case of trading by
individual investors (Odean, 1999; Barber and Odean, 2001). In this last case
investors overestimated the precision of their informa tion about individual
companies, with the result that they traded too much. Barber and Odean further
found that men were more overconfident in this respect than women. The
phenomenon of overprecision has also been found in the field of voters’ political
beliefs. One study indicated that a number of voter characteristics were asso ciated
with this kind of bias, in particular ideological extremeness, voter turnout, and
partisan identification (Ortoleva and Snowberg, 2015). Kahneman (2011) reports
that experts in various professional fields are particularly prone to overprecision.
One example relates to CFOs of large organizations forecasting returns on the
Standard and Poor index; when asked to give 80% confidence intervals for their
forecasts these turned out to be only a quarter of the correct size, which was
between −10% and +30%. Another example concerns doctors, where autopsy
results showed that in cases where they had been ‘completely certain’ of their
antemortem diagnosis they were actually wrong 40% of the time. Once again we
have a puzzle, in the sense that on the face of things it would seem that experts
who are proved wrong should not prosper. This issue is addressed in the synthesis
at the end of the section. Underconfidence Empirical studies have sometimes found
conflicting results, in that sometimes people underestimate their abilities, control,
and also underplace their performance relative to others (Kirchler and Maciejovsky,
2002; Burson, Larrick and Klayman, 2005; Fu et al., 129 PT• II FOUNDATIONS 2005).
Some studies have reported overconfidence when the tasks were easy (like driving),
or success likely, and underconfidence when tasks were difficult (like playing the
piano), or success unlikely. This phenomenon is referred to as the ‘hard-easy’ effect
(Lichtenstein and Fischoff, 1977). There has also been conflict in that other studies
have reported underconfidence with easy tasks and where success is likely. Moore
and Healy (2008) suggest that this conflict is caused by the confound between
overconfidence and overplacement. They have proposed a theory that can explain
these empirical anomalies and resolve the apparent conflicts. This is described as
follows: People often have imperfect information about their own performances,
abilities, or chance of success. However, they often have even worse information
about others. As a result, people’s estimates of themselves are regressive, and their
estimates of others are even more regressive. Consequently, when performance is
high, people will underestimate their own performances, underestimate others even
more so, and thus believe that they are better than others. When perfor mance is
low, people will overestimate themselves, overestimate others even more so, and
thus believe that they are worse than others. (p. 503) Thus according to this theory
it is possible, and indeed likely that people will combine overestimation with
underplacement and vice versa. Moore and Healy conducted an experiment
involving students performing trivia quizzes, and the results supported their theory.
There is one other factor that has been suggested as playing a role as far as
excessive optimism and pessimism is concerned. This is the role of arousal,
sometimes referred to as visceral influences. Thus this phenomenon is sometimes
referred to as the affect heuristic, as mentioned earlier. Vosgerau (2010) has
proposed that people judge the likelihood of desirable and undesirable events to be
higher than similar neutral events because they misattribute the arousal caused by
those events to their greater perceived likelihood. Thus we may overestimate the
likelihood of a terrorist attack or getting cancer; similarly we may overestimate the
likelihood of our country winning the World Cup in soccer. Vosgerau finds evidence
of this misattribution phenomenon in four studies. The misattribution effect above
may also explain another curious aspect of behavior: people’s reluctance to
exchange lottery tickets (Risen and Gilovich, 2007). Miller and Taylor (1995) have
pointed out that precisely because undesirable outcomes that result from actions
taken are more painful than identical outcomes that result from actions foregone,
instances in which one has been punished for acting are likely to be overrep
resented in the memory. The aversion caused by this anticipated regret from
switching would then be mistaken for the increased probability of the event
occurring. Thus we may be reluctant to exchange lottery tickets; similarly, we may
be disinclined to switch lines at the supermarket checkout when our line appears to
be going slowly and the line next to us is speeding along. There may be another
aspect to this behavior that we will discuss later, in connection with ‘tempting fate’.
Self-serving bias This term refers to any cognitive process that is distorted by the
need to maintain and enhance self-esteem, and has been used to describe a
number of belief biases that are different in nature. For example, it has been used
to refer to the asymmetry whereby people ascribe their successes to their own
ability or skill, but ascribe failures to situ ational factors, the actions of other people,
or bad luck (Zuckerman, 1979). People also tend to overestimate their contribution
to joint or team projects (Ross and Sicoly, 1979). These are aspects of
overconfidence, and conform to the findings of much social cognitive 130 beliefs
anD exPeCtations CH•4 research, which suggests that people shape their beliefs
and judgments of the social world to maintain sacrosanct beliefs of the self as a
capable, lovable, and moral individual (for a recent survey, see Dunning, 2007). This
kind of self-serving bias is consistent with the observation by Glaeser (2004):
Mistaken beliefs will be more common when errors increase the current flow of
utility. Thus, if people enjoy anticipating a rosy future, they should believe stories
that make them overly optimistic and in particular, they should happily accept sto
ries about a life after death. (p. 4) A further aspect of this general phenomenon is
that self-serving bias relates not just to individuals’ evaluations of themselves, but
also to groups with which they are affiliated. Observe any team game with partisan
spectators; the different fans will interpret the play, and in particular aspects
involving foul play or penalties, quite differently. A study of investor behavior also
demonstrates how self-serving bias relates to how beliefs relate to the beliefs of
others (Egan, Merkle and Weber, 2014). They find that investors are posi tively
influenced by the optimistic beliefs of other investors in terms of investing in stocks,
which is an example of conformity bias, but they also find two other biases which
are self serving: (1) investors tend to believe that their own opinion is relatively
more common among the population (false consensus); and (2) others who hold
divergent beliefs are considered to be biased (bias blind spot). There is some
evidence for both psychological and neurological mechanisms related to this
overoptimism. It has been reported that depressed subjects make more accurate
assessments, and so are more realistic than normal subjects; this phenomenon has
been labeled depressive realism (Abramson, Metalsky and Alloy, 1979). It has also
been sug gested that the phenomenon of Pavlovian withdrawal associated with
predictions of negative outcomes is an important route to the overoptimism of
normal subjects, and that one of the underlying neural malfunctions associated with
depression is associated with a weakening of this withdrawal, thereby leading to
more accurate, but more pessimistic, evaluations (Huys and Dayan, 2008). This
means that when normal people contemplate the future, any thought leading
towards a negative outcome will engender a Pavlovian withdrawal response, which
may lead to the thought being terminated. There are simi larities here with
Damasio’s somatic market hypothesis. It has been suggested that this withdrawal is
mediated by the neurotransmitter 5-HT, which opposes dopamine (Daw et al.,
2002), and that depressives have low effective 5-HT levels (Graeff et al., 1996),
resulting in their withdrawal mechanism being impaired. There is also evidence of
another type of self-serving bias, relating to a non-conscious desire for autonomy.
People generally consider themselves better decision-makers than others and
devalue the thoughts and feelings others use to make decisions (Pronin and Kugler,
2007; Pronin, 2008). As a result they may resist communications from external
sources that are designed to be persuasive, and endorse their original beliefs even
more strongly (Brehm, 1966; Stein, Ackerman and Bargh, 2012). The Remain
campaign dis covered this effect to their cost in the UK Brexit referendum in 2016. It
should be noted that the effect here operates in the opposite direction to conformity
bias described earlier. This issue is discussed later in Chapter 10, in the context of
nudging strategies. Confirmation bias Confirmation bias is a major type of self-
serving bias, and refers to the tendency to search for, process, interpret and recall
information in a way that confirms one’s prior beliefs, while also having the
tendency to discount information that is inconsistent with one’s prior beliefs
(DellaVigna, 2009). Thus it has some factors in common with the desire for auton
omy described above. For example, in financial markets, as traders receive
additional 131 PT• II FOUNDATIONS private information, in the short term they
interpret the information that confirms their existing beliefs as being more
informative, rejecting non-confirming information, and this causes them to become
more overconfident and trade excessively. As we have already seen, this
phenomenon can also be caused by the representation heuristic, since recent
information is often regarded as the norm. The main implication of this is that it
leads to momentum, meaning that there is positive correlation of returns in the
short term, so that a stock that goes up in value in one day or over a few days may
well continue to go up in the next short time period. In the long term prior beliefs
are adjusted in line with the additional information and valuation returns to
fundamentals. These effects may oper ate in the opposite direction to the effects of
the law of small numbers, but if they are strong enough they may contribute to
bubbles in asset markets. Confirmation bias may also explain why people like to
invest in their own company’s stock, if they are overcon f ident about own company
performance. Furthermore, it may help to explain why they prefer to invest in
national companies rather than foreign companies for a similar reason. Another
type of confirmation bias exists in the sciences and academia. Once a researcher
has developed a certain theory, and particularly when it has become published and
related to their reputation, the originator becomes emotionally and professionally
attached to that theory. This can cause the person to carry out biased research, by
cherry picking data and ignoring inconvenient conflicting evidence, as we have seen
in Chapter 2 with the diet-heart hypothesis. It can also lead to distorted, imbalanced
and misleading conclusions. Again this occurred with the Minnesota Coronary
Experiment described in Chapter 2. The results were not published until 1989, 16
years after the end of the experiment, and the conclusions suggested that the diet-
heart hypothesis may have been confirmed if the experiment had lasted longer, and
that ‘there was a favorable trend’ for younger patients. This is an extremely
charitable view of the results, considering that the average duration for subjects on
the experiment was over a year, and that the results showed that there was an
inverse rather than the hypothesized positive relation ship between blood-
cholesterol levels and coronary deaths, particularly for older patients. A similar
distorted conclusion was found in the follow-up to the long-standing Framing ham
study (Dawber, Moore and Mann, 1957, 2015), often quoted as being a major
supporter of the DHH. In the follow-up study it was found that, for subjects over 50,
those with falling cholesterol levels experienced an increase in both CHD and
overall mortality, a similar finding to the MCE. The authors suggested that after age
50 years there was a possible confounding effect because people with low
cholesterol may be suffering from diseases predisposing to death. There is a
substantial literature in psychology relating to confirmation bias, with recent
surveys conducted by Lord and Taylor (2009) and Mercier and Sperber (2011). It
has also been modeled by Rabin and Schrag (1999), who propose that people use
modi f ied Bayesian probabilities involving non-zero probabilities of misperceiving
signals. Thus they may misperceive a signal as confirming hypothesis A when
hypothesis B is correct. Cognitive dissonance These examples of confirmation bias
demonstrate the existence of an important psycho logical phenomenon related to
self-serving bias known as cognitive dissonance (Festinger, 1957). This theory
proposes that people suffer mental discomfort when they simultane ously hold
conflicting attitudes, beliefs or ideas, and that they try to reconcile these by
changing them to achieve consistency. As we have seen in the previous chapter,
the desire for consistency, in both beliefs and behavior, is a basic psychological
phenomenon. As a result people are prone to self-deception; people may
‘confabulate’ their intentions, mean ing that they invent them after they have taken
some action, The situation is demonstrated by Aesop’s fable of the fox and the sour
grapes. The fox wanted the grapes, but when she 132 beliefs anD exPeCtations
CH•4 found she couldn’t reach them she decided that they were probably sour, so
she revised her original intention and believed that she never really wanted the
grapes in the first place. The phenomenon of cognitive dissonance has been
claimed to have played an important role in recent political campaigning, both in
the Brexit referendum in the UK and in the Trump campaign in the USA. This
political aspect is discussed further in Case 4.3. There is also evidence of cognitive
dissonance in the financial markets. A study by Chang and colleagues (2016)
indicates that it is an important factor underlying the dis position effect, discussed
in more detail in Chapter 6. Investors may be more willing to sell stocks that have
risen in value since purchase than stocks that have fallen because to do the latter
may be seen as an admission that they may have made an error of judg ment. The
evidence suggesting this comes from the fact that when investors delegate the
decisions to money fund managers the disposition effect disappears and is replaced
by an opposite effect. Furthermore, increasing the salience of the delegation was
found to increase the extent of the reverse-disposition effect. Thus cognitive
dissonance theory can explain contrary aspects of investor behavior, where
investors in individual stocks show a disposition effect but managers investing in
mutual funds exhibit the reverse effect. Perhaps the most extreme cases of mass
cognitive dissonance have occurred with the millennium religious movements over
the centuries, usually predicting a cataclysmic end of the world. However, when
such prophecies fail, ‘rather than disbanding in disillusion, the cultists often become
strengthened in their convictions’ (Cialdini, 1984). Cialdini then identifies several
such cults where this has happened: Montanists of second-century Tur key,
Anabaptists in sixteenth-century Holland, Sabbataists of seventeenth-century Izmir,
and Millerites in nineteenth-century America. A more recent case of the same
phenom enon has been described by Festinger, Riecken and Schacter (1956), who
were able to observe the behavior of a similar cult in detail by posing as insiders.
They were able to witness the astonishing conclusion of the prophesy failure: ‘the
little group, sitting alone all night long, had spread so much light that God had
saved the world from destruction.’ One cannot wish for a clearer example of
cognitive dissonance. The tendency of people to strengthen their views after the
appearance of evidence refuting them has been documented on a number of
occasions, causing the phenomenon to be labeled the ‘backfire effect’ by Nyhan
and Reifler (2010). These authors found a similar effect in an experimental study
examining the political beliefs of liberals and con servatives towards the Iraq War in
2003. Once again this phenomenon presents a puzzle, since it would not appear at
first sight that self-deception would serve a useful purpose in terms of biological
fitness. Visceral fit The phenomenon of cognitive dissonance involves visceral
factors, since we tend to be emotionally attached to our beliefs. Recent research
has also pointed to a similar phe nomenon, in that when there is a fit or match
between our current visceral state and the visceral state associated with an
outcome we are judging, we tend to increase our estimate of the likelihood of this
outcome occurring (Risen and Critcher, 2011). For example, if we are in a visceral
state of being warm, this tends to increase our belief in the reality of global
warming. Of course, if this experiment is performed naturally, as when we ask peo
ple about their beliefs in the probability of global warming on a hot day, then the
resulting bias could be explained by the law of small numbers. In this case people
would be using the current warm temperature as a diagnostic device for estimating
the probability of warm temperature in the future. However, Risen and Critcher
found that their subjects expressed a stronger belief in global warming even when
the experiment was performed in a warm room. They therefore eliminated the
explanation that temperature was being used as a diagnostic, and instead propose
a simulational fluency explanation. This means that 133 PT• II FOUNDATIONS
people construct mental images of hot outdoor scenes more clearly when they are
in a hot room than when they are in a normal room, suggesting that, when warm,
participants had a more fluent or clear representation of heat-relevant stimuli. The
above research concentrates mainly on the effect of heat as a visceral factor.
However, there may be wider implications of the concept of visceral fit to other
visceral states. For example, a possible change in government policy, like higher
taxes, may make us angry; if we are currently in an angry state, would this make us
believe such a change in policy is more likely? Further research is needed in this
area to clarify the effects of visceral fit in different situations. Projection bias
Another kind of bias where people have systematically incorrect beliefs is that they
expect their future preferences to be too close to their present ones. This kind of
misestimation of probabilities can also be regarded as a kind of self-evaluation bias,
in that we tend to make incorrect judgments of our future selves. For example, it is
generally not a good idea to go to the supermarket when we are hungry – we tend
to buy all kinds of junk that we don’t normally eat or want to eat, and not only is our
bill higher than normal, but we end up with stuff we don’t consume or don’t want to
consume. This happens because at the time of shopping we incorrectly anticipate
that our future hunger will be as great as it is now. The term ‘projection bias’ was
introduced by Loewenstein, O’Donoghue and Rabin (2003) to describe this
phenomenon. They proposed a simple model as follows: assume that utility u is a
function of consumption c and of state variable s (which incorporates tastes or
preferences), so that: u = u(c, s) The current state is s′ and the (unknown) future
state is s. Then, when predicting the future utility û(c,s), and person with projection
bias expects utility: û(c, s) = (1 – α)u(c, s) + αu(c, s′) (4.2) whereas the person
without projection bias (who has complete knowledge about the future state s) has
expected utility û(c, s) = u(c, s). The parameter α (which must be between 0 and 1)
measures the extent of projection bias, so that if α = 0 there is no projection bias,
and if α = 0 there is full projection bias. Read and van Leeuwen (1998) confirmed
this effect in a study of office workers. These workers were asked to select a healthy
snack or an unhealthy snack to be delivered a week later (in the late afternoon).
One group of workers was asked the question at a time when they may have been
hungry, in the late afternoon, and 78% chose an unhealthy snack. The other group
was asked the same question after lunch, when they were probably satiated, and
only 42% chose the unhealthy snack. Evidence of projection bias has also been
provided in the field; an example is a study by Conlin, O’Donoghue, and Vogelsang
(2007). They examined the effect of weather at the time of purchase on the return
of cold-weather apparel items that had been ordered. The NM predicts that there
should be no relationship here, or a negative one if colder weather at time of
purchase is correlated with colder weather later, making people less likely to return
the item. The projection bias hypothesis predicts the opposite effect, with people
overestimating their later use and being more likely to return the item. The authors
of study did indeed find the opposite effect, estimating that a reduction in the order-
date temperature of 30°F (17°C) increases the average return rate of a cold-
weather item by nearly 4%. In this case the model (4.2) above estimates the value
of α to be about 0.5, indicating that consumers predict future tastes roughly half-
way between present tastes and actual future tastes. 134 beliefs anD exPeCtations
CH•4 Further evidence of projection bias in the field, again related to weather,
relates to car purchases. Busse and colleagues (2015) examined 40 million vehicle
transactions, f inding that the choice to purchase a convertible or a four-wheel-drive
is highly depend ent on the weather at the time of purchase in a way that is
inconsistent with neoclassical utility theory. An associated kind of bias is hindsight
bias, which could be considered to be a ret rospective projection bias. This means
that events seem more predictable in retrospect than in prospect, as in ‘we knew it
all along’. There is again evidence for this phenomenon both from experiments and
in the field. For example, a study by Biais and Weber (2009) conducted an
experiment with 85 investment bankers in London and Frankfurt and found not only
evidence of hindsight bias among some subjects, but also that the biased agents
have lower performance. Synthesis During the discussions in this section we have
seen that there are various puzzles as far as explaining the existence of various
types of overconfidence. These puzzles are all related to the concept of biological
fitness and adaptation. None of the types of overconfidence described appear to
improve biological fitness, and therefore one might expect such disad vantageous
psychological traits to have been weeded out by Darwinian natural selection. For
example, as argued by Anderson and colleagues (2012), recognizing one’s
limitations would help people set more realistic goals (Ehrlinger and Dunning,
2003), avoid contests one will lose (Camerer and Lovallo, 1999), and select
strategies that facilitate success (Neale and Bazerman, 1985). It is therefore useful
at this point to try to synthesize the various types of overconfi dence by discussing
some of the psychological explanations that have been offered. Three different
types of causal factor have been identified (Kahneman, 2011; Anderson et al.,
2012), all of which may endow overconfidence with some advantages; these are
emotional factors, cognitive factors, and social factors. The emotional factors relate
to self-esteem and self-aggrandizement. Overconfidence, by improving self-esteem,
may improve mental health, and reduce stress and the incidence of depression. It
may also encourage people to persist in tasks that they would otherwise give up on.
Cognitive factors include what Kahneman refers to as the WYSIATI syndrome,
meaning ‘what you see is all there is’. This is a blinkered perspective where we can
see our own efforts but ignore the efforts, or even existence of others, in an
extreme case of salience. Thus this phenomenon is also referred to as the
competition neglect bias, and it can affect both individuals and groups, for example
companies. Kahneman suggests that it can cause overplacement bias in particular,
since we tend to underrate the perfor mance of others. In this case this is not
because of self-aggrandizement, but because we do not perceive the behavior of
others. Another cognitive factor that is relevant here is that some people may be
too incompetent to appreciate their own incompetence (Kruger and Dunning, 1999).
Williams, Dunning and Kruger (2013, p. 2) report that ‘incompetent performers
display little insight into just how poorly they perform. On average, the bot tom 25%
of performers on some task, such as logical reasoning, grammar, or taking a
classroom exam, think their performance lies above the 60th percentile,
overestimating their raw score by as much as 50%.’ The authors go on to suggest
that the cognitive prob lem here may be due to the consistent application of
inappropriate algorithms for solving problems. Social factors are also important in
explaining various types of overconfidence, in par ticular overprecision, self-serving
bias and self-deception. Kahneman (2011) suggests that experts who make precise
predictions appear more knowledgeable and therefore have more market value; on
the other hand, experts who make much vaguer but more realistic 135 PT• II
FOUNDATIONS predictions are likely to appear clueless and provoke disparagement.
The explanation of self-deception is more complex. Evolutionary psychologist
Steven Pinker has speculated that self-deception has evolved as a form of
commitment (Pinker, 1997). The nature and purpose of commitment is discussed in
more detail in Chapter 8 in relation to intertem poral decision-making, but at this
stage we can simply say that Pinker’s theory involves the concept of an
evolutionary arms race in psychological terms. Our emotions are a form of
commitment, so for example people may be less inclined to harm us if they know it
will make us angry and retaliate. However, anger can be faked to have the same
effect. To be credible, commitments like the facility for anger have to be hard to
fake. Smiling is notoriously hard to fake, since voluntary or deliberate smiling
involves different muscles and parts of the brain (the cerebral cortex) compared to
involuntary or genuine smiling, controlled by the limbic system. Taking the arms
race one step further, in order to ‘fake’ emotions that are hard to fake, Pinker
proposes that the best solution is to genuinely feel emotions like anger, fear,
shame, guilt, sympathy and gratitude, that is, believe in false feelings and
intentions that one does not really have. Trivers (2013) takes this explanation
further, suggesting that genuine-seeming expressions of overconfidence are more
likely to convince others, and therefore improve the social status of those who are
better able to deceive themselves. This would apply in particular to supposed
experts. Since high social status has important implications in terms of control over
resources and thus biological f itness, self-deception may paradoxically be adaptive
after all. 4.4 Magical beliefs This title is a general term for certain irrational beliefs
that violate the assumptions of the NM, but do not fit it any of the above three
categories. They are often termed ‘supersti tions’ in folk psychology. Superstitions
are another example of situations where according to dual process models system 1
dominates our thinking and the resulting errors are not corrected by system 2. Two
main categories are important to discuss here. Tempting fate This phenomenon has
been touched on earlier in connection with arousal and misat tribution of
probability. We have seen for example that people can be reluctant to switch lines
at supermarket checkouts or exchange lottery tickets. There are widespread
instances or applications of this: if you don’t take your umbrella to work, it’s bound
to rain; if you don’t do your homework reading, the teacher is bound to pick on you
in class to answer questions on it. There are various factors involved here: we have
seen that the misattribution effect states that undesirable outcomes resulting from
actions taken are more painful than identical outcomes that result from actions
foregone, and the aversion caused by this anticipated regret from switching would
then be mistaken for the increased probability of the event occurring. We shall also
see in the next chapter that loss-aversion is an important factor governing decision-
making in risky or uncertain situations. It is interesting to note that this superstition
is a cultural universal. In some cultures people explicitly believe in fate or some
supernatural being or force which can act with discretion in the relevant
circumstances, but even in cultures where there is no explicit belief in the
intervention of some supernatural agent, the superstition exists at an intui tive level
that we should not tempt fate. A further point here is superstitions frequently
persist even when we know that they are errors (Risen, 2016). The implication is
that the processes of error detection and correction are disconnected. Thus system
2 can detect an error but still fail to correct it. Although further research is needed
in this area, it may be 136 beliefs anD exPeCtations CH•4 that cognitive dissonance
theory is relevant here, meaning that the correction of the error involves an
emotional loss or threat which an individual may prefer to avoid. It is also important
to realize that superstitions, however irrational they may appear to be, can have
significant physiological effects on their holders. A dramatic example of this
concerns the superstition in China and Japan concerning the number four. The rea
son for this is that in Mandarin, Cantonese and Japanese, the words ‘four’ and
‘death’ are pronounced nearly identically. As a result the number four evokes
discomfort and stress in many Chinese and Japanese people, and this number is
avoided in various circum stances, for example floor and room numbers and some
phone numbers. A large-scale study by Phillips and colleagues (2001) examined
over 47 million deaths in the USA from 1973 to 1998 and found that cardiac deaths
peak on the fourth of the month for Ameri cans of Chinese and Japanese descent,
but this pattern is not seen among whites. Contagion Disgust is a strong aversive
emotion or visceral factor. It has been described as ‘a revul sion at the prospect of
(oral) incorporation of an offensive substance’ (Rozin and Fallon, 1987). However,
we can note that disgust can also be prompted by touch or even proxim ity, not just
ingestion. Disgust causes certain unique responses as an emotion: a distinct facial
expression with closed nostrils, an attempt to get away from the disgusting object,
and a physiological response of nausea, as well as an emotional state of revulsion.
These responses are different from those related to anger or fear, although there
may be some similarities. It might be initially thought that disgust is not that
important in terms of having a frequent and significant effect on behavior. Evidence
suggests otherwise, for two reasons: 1 A large number of everyday objects can
cause disgust – A survey by Morales and Fit simons (2007) found that six of the top-
ten-selling non-food supermarket items elicit feelings of disgust, including trash
bags, cat litter and diapers. Many food or ingested items also elicit disgust, such as
cigarettes, mayonnaise, oils and lard. Therefore con sumers are likely to experience
some degree of disgust routinely on shopping trips. 2 The property of contagion
means that other products coming into contact with a disgusting object are
contaminated – this process is described by a phenomenon referred to by
anthropologists as ‘sympathetic magic’. This is not just a belief sys tem found in
primitive cultures; it exists in the same general forms in all cultures, although in
developed countries people are often reluctant to admit such beliefs for fear of
appearing foolish. One of the fundamental laws of sympathetic magic is the ‘law of
contagion’. This law states that objects or people can affect each other by merely
touching, that some or all of the properties of the disgusting object or per son or
transferred, and that this transfer is permanent. Thus the law is sometimes referred
to as ‘once in contact, always in contact’. The ramifications of the two factors
described above are widespread as far as consumer behavior is concerned, but
before discussing these it is appropriate to give some expla nation as to why the
‘law of contagion’ is a universal phenomenon, given its sometimes strange effects.
For example, Rozin, Millman and Nemeroff (1986) found that a drink touched briefly
by a sterilized cockroach became undesirable, as did a laundered shirt previously
worn by a disliked person, although subjects were often not able to verbalize or
acknowledge their contagion belief. Sometimes people are not even conscious of
their emotional disgust, but it is reflected in lower evaluations of products. We really
need to consider the evolutionary psychology involved here: a product contagion
heuristic really follows on from a general ‘contact causing’ inference. For example, if
people eat a lot of fat, they tend to get fat; eating a lot of garlic leads to a garlic
odor on the breath and 137 PT• II FOUNDATIONS the body. Furthermore, biologically
speaking, it would have been a useful adaptation in human history to avoid
situations where contamination was possible. Cockroaches can cause contamination
of food through microbial infection, as can contact with raw meat, dirt or feces. Thus
contamination has historically speaking been a significant and dangerous problem
in terms of human survival; we seem as a species to err on the cau tious side and
misapply the concept in situations where science now informs us it is not relevant. It
should be noted at this point that the phenomenon of contagion involves two
aspects from an economic viewpoint: it affects values, preferences and choices, in
par ticular through visceral factors; and it affects beliefs and the estimation of
probabilities, discussed in this chapter. Also, particularly important from the
viewpoint of behavioral economics, contagion and disgust can activate different
parts of our dual-sensory sys tems, meaning the fast-thinking and slow-thinking
systems in the terms of Kahneman (2011). This dual processing aspect has been
illustrated in a study by Schulze, Maertens and Wansink (2013), where consumers
were required to consider both sandwiches con taining (healthy) dogfood and fat-
free ice cream. The experiment involved manipulation of cognitive load, so that in
its absence consumers were able to implement slow-thinking cognitive processes in
evaluating the health benefits of dog food, which would otherwise evoke the fast-
thinking emotion of disgust. What are the consumer behavior ramifications of these
findings? Morales and Fit simons (2007) found that direct physical contact itself was
not necessary for a contagion effect to occur, merely a perception of contact was
sufficient. Thus raw meat, or drinks in transparent containers, was more likely to
cause a perception of contamination than if packaged in opaque containers. In the
supermarket situation contamination can occur in terms of either proximity of items
on the shelves or proximity in the shopping cart. Thus when lard is positioned on
shelves near other baking products, pans and utensils, as is commonly found, these
other products are likely to receive a lower evaluation from consumers. There are
obviously implications here in terms of managerial policy. Managers need to take
care in determining shelf location for products to minimize the effects of lower
evaluations. Even though they cannot control proximity in the shopping cart, they
can allow consumers to take avoidance measures. Opaque and substantial
packaging may be important for some products. Some supermarkets are now
providing facilities for double wrapping meat. Governments also need to be aware
that there can be consumer panics in response to food health scares, involving
overreactions caused by visceral factors; reassur ance may be necessary in these
circumstances. There is one final point that is worth raising here in terms of disgust.
We have con centrated so far on situations where the disgust is purely physical;
people can also feel moral disgust. An outstanding example can be seen in the BP
oil spill in the Gulf of Mexico. In this case the discussion may combine elements of
physical and moral disgust; people do not like to see birds coated in oil, but they
have also been revulsed by the seeming negligence and reactions of BP
management, and the ramifications of this have been huge. Not only has BP been
treated as a pariah by press and public internationally, but other oil companies have
suffered also, as has the US president. Evidence also exists of increased anti-British
sentiment in the USA as a result of the disaster. The plunge in the BP share price,
arguably not justified by economic fundamentals, has important implications for UK
pension funds (and ultimately pension investors) that have invested heavily in BP
stock. A more recent example of this phenomenon relates to an incident on a United
Airlines flight in April 2017, where a passenger was involuntarily dragged from the
plane by security. The already troubled airline has faced huge criticism in the press
and social media, and the affair has been badly handled by the CEO. It is too early
at this stage to see if the company’s stock price will suffer accordingly. 138 beliefs
anD exPeCtations CH•4 So far we have examined only the negative aspects of
contagion. As was indicated in the introduction to the chapter, contagion can have
positive aspects also. It can vastly inflate the values of objects, such as ‘Blackie’, a
guitar once owned and played by Eric Clapton, which sold for $959,500 in 2004.
According to Paul Bloom, a psychologist from a team at Yale University who looked
into this phenomenon a bit more closely, when inter viewed on his research by the
New York Times (Tierney, 2011, p. A16): Our results suggest that physical contact
with a celebrity boosts the value of an object, so people will pay extra for a guitar
that Eric Clapton played, or even held in his hands. This is the same kind of thinking
that makes people reluctant to wear the sweater of a murderer. Bloom and
colleagues find that people value highly the possessions of celebri ties even if they
despise them, since they expect the possessions of notorious celebrities, or even
pariahs like Saddam Hussein, to be valued by others (Newman, Diesendruck and
Bloom, 2011). Furthermore, the values of these possessions are significantly
reduced if they are washed or in some way sterilized. A similar psychology applies
even with replicas of objects owned by celebrities. In this case the phenomenon is
known as ‘imitative magic’, meaning that things that look alike are alike. Thus a
replica of ‘Blackie’, perfect down to the cigarette burns and belt buckle scratches
fetched $30,500 in auction in March 2011. Less perfect replicas sell for lower prices,
but are still valued highly. The replica fetish is important in the musical business,
extending not only to guitars and strings but also to amplifiers, microphones and
other instruments. Fernandez and Lastovicka (2011) have studied the phenomenon
of consumption fetishes and their connection to magical beliefs extensively in
modern mass consumer culture. Ultimately, they consider that the foundations for
these magical beliefs are to be found in evolutionary biology, as explained by John
Lastovicka in the same New York Times article: Beliefs about contagion, and
especially biological contagion, by our ancestors are one of the reasons why we are
here today. Those who did not stay away from those who died from the plague in
the Dark Ages also died of the plague; those who died of the plague in the Dark
Ages likely have few, if any, descendants today. So in our modern and scientific
world, these manners of magical thinking still persist. The subject of magical beliefs
and contagion is examined in more detail in Case 4.2. 4.5 Causes of irrationality We
have now surveyed a variety of situations where people exhibit a formation or
holding of beliefs that violates the kind of rationality proposed in standard
economics. It is there fore worthwhile at this stage to discuss the causes of the
underlying phenomena. Baumeister (2001) has identified five different causes of
irrational, or what he terms self-defeating, behavior. We can really equate self-
defeating behavior with behavior that is not in a person’s long-run self-interest. One
can of course question whether it is legitimate to equate irrational behavior with
self-defeating behavior, and this aspect will be discussed further in the final
subsection. However, what is important here is the usefulness of the categories that
he proposes in terms of analysis. These categories are: emotional distress,
threatened egotism, self-regulation failure, rejection and belonging ness, and
decision fatigue. The first of these categories involves a variety of factors, and it is
helpful to discuss memory and cognitive dissonance as separate categories. 139
PT• II FOUNDATIONS Emotional distress The general impact of emotions on
preferences and choices has been discussed in the previous section. We now need
to explain how and why these effects occur, recognizing that this remains a highly
controversial area in psychology. There has been a lot of research into the effects of
emotions on decision-making. The conventional attitude taken by economists, and
also by philosophers in the Kantian tradition, is that emotions tend to cloud good
judgment, resulting in ‘irrational’ decisions or self-defeating behavior. However, this
raises the issue mentioned in the previous sec tion in relation to evolutionary
psychology: how can emotions serve as an adaptive evolved psychological
mechanism? If they were a maladaptation, people with genes for emotional
behavior would not have passed them on to succeeding generations, and we would
now be living in a world full of unemotional people, like Doctor Spock from Star
Trek; this is clearly not the case. In the late 1980s the economist Robert Frank
proposed a theory that emotions served as a commitment mechanism, and thus
were a useful adaptation. Frank’s theory was supported by independent research by
Jack Hirshleifer. The neuroscientist Damasio (1994) also researched the role of the
emotions in decision-making, by examin ing patients with brain damage, again
concluding that emotions could be an aid as well as a hindrance. These theories are
discussed in more detail later in this section. At this stage we can summarize the
situation by saying that emotions can lead to either better or worse decisions,
depending on the circumstances. While the effects may be unpleasant and
destructive, the evolutionary advantages of such changes in behavior are obvious.
The need to satisfy or reduce basic drives is fundamental to survival and
reproduction. This also applies to the sensation of pain; its unpleasantness is a
signal that something is wrong with the biological system, and we should be doing
something to remedy the situation (get out of the heat/cold, rest an injured limb,
defend ourselves against the person attacking us). The theory of emotions as an
evolved psychological mechanism or mental adaptation was described by Frank
(1988) in the seminal work Passions within Reason. According to Frank our emotions
serve as commitment devices, meaning that they commit us to perform certain
actions at a later time if other people behave in certain ways. The nature of
commitment in general is considered in Chapter 8 in relation to intertemporal
decision-making, and it is also relevant in game theory. Frank’s insight was to see
the role of emotions in prompting us to perform actions that we would not carry out
if we were acting on purely ‘rational’ grounds. A simple example can illustrate the
situation. Imagine that we make an agreement with another person such that we
perform some work for them now in exchange for being paid afterward. Such
‘delayed exchange’ contracts have been extremely common in human history, on
both a formal and informal basis. The per son doing the work first is always subject
to a ‘holdup’ problem (unless the details are formalized in a written contract), in
that the other party can renege on the deal. Without any formal contract the
cheated party has no comeback, and a ‘rational’ person may sim ply write off the
loss, and put it down to experience. An ‘emotional’ person on the other hand would
be angry with the cheat and take steps to gain revenge, at risk and cost to him self,
which the ‘rational’ person would be unwilling to take. However, the knowledge that
an emotional person may react in this way might well be enough to prevent the
other party from cheating in the first place. This is an example of what is called a
‘reputation effect’; emotional people may gain a reputation for not standing for any
nonsense or backsliding in their dealings, thus encouraging others to be straight
with them. This example illustrates how our emotions can serve our long-run self-
interest, but Pinker (1997) has gone a step further, showing how our emotions can
backfire on us, refer ring to them as ‘doomsday devices’, after the movie Doctor
Strangelove. The problem with doomsday devices is that they cannot be disarmed,
even if they are activated by mistake, 140 beliefs anD exPeCtations CH•4 and will
explode regardless of the consequences. Thus they may lead to futile and self
destructive reactions; a well-known example is the successive rounds of retaliation
that occur with feuds between gangs or clans. It is possible that the reaction to
social rejection, discussed in more detail later, is of this type. Emotions are indeed a
two-edged sword. One aspect of emotional distress where research has indicated a
bad effect on deci sions is the role of anger on risk-taking. Leith and Baumeister
(1996) found that people who were upset were more inclined to take foolish risks,
like betting on long shots in a lottery. There were various possible explanations for
this; for example, people who were already upset had less to lose by taking a long
shot and more to gain, while people who were in a good or neutral mood had more
to lose by taking a long shot. However, Leith and Baumeister were able to eliminate
this explanation by further experimentation, requiring respondents to reflect on
their decisions for about a minute before choosing. Although the respondents were
still angry when they made their decision they now became more risk-averse. Thus
it seems that emotional upset does indeed cloud judgment of risk, and that when
upset people are forced to think about things they make better decisions. Fear is
another negative emotion that can result in adverse or self-defeating conse
quences rather than favorable ones. The traditional response is a fight-or-flight
reaction, which is generally a useful adaptation. However, as some governments
have found, trying to invoke excessive amounts of fear in a population may
backfire. This has been true of some anti-smoking campaigns. Sometimes, when
people are faced with information they really do not want to have to process in
terms of its unfavorable consequences, they are inclined to filter it out and ignore it.
The research findings here are somewhat conflicting. Some research has shown that
a message where fear is combined with disgust can be more compelling (Morales,
Wu and Fitzsimons, 2012), but as discussed earlier, disgust can also be aversive,
causing people to ignore the message. A less extreme but more chronic negative
emotion is anxiety, which can also cause self-defeating responses. In this case it
appears that it affects the ability of people to process new information about a
situation, resulting in a failure to implement Bayesian updating and leading to poor
decision-making (Browning et al., 2015). Memory As far as our emotional states
over time are concerned there are two important factors that need to be discussed,
in terms of both their causes and their implications: 1 People tend to revert to a
‘normal emotional state’ after any kind of emotional expe rience, whether it be
pleasant or unpleasant. 2 People tend to overestimate the length of time that it will
take to revert to this normal state. The first aspect of human nature has long been
known. According to Adam Smith in The Theory of Moral Sentiments of 1759: The
mind of every man, in a longer or shorter time, returns to its natural and usual state
of tranquillity. In prosperity, after a certain time, it falls back to that state; in
adversity, after a certain time, it rises up to it. (p. 172) This ‘certain time’ turns out
to be a shorter rather than a longer time in general. There is now a substantial body
of research showing that emotional reactions to life- changing events are
surprisingly short-lived (Suh, Diener and Fujita, 1996; Frederick and Loewenstein,
1999). When people win large amounts of money in a lottery, they do not remain
happy for very long (Brickman, Coates and Janoff-Bulman, 1978; Kaplan, 1978). In
the opposite direction, the majority of bereaved spouses reported themselves to be
doing well two years after the death (Lund et al., 1989; Wortman, Silver and
Kessler, 1993). 141 PT• II FOUNDATIONS Similarly, people who have suffered
serious injury confining them to a wheelchair have recovered equanimity within a
period of a year. The second aspect of human nature is less well known. However,
experiments have been performed that measure people’s forecasts of emotional
events and compare these with their actual duration, and there is evidence of a
consistent durability bias in both directions. Thus there is a tendency for people to
overestimate the duration of their reactions to both positive and negative emotional
events (Gilbert et al., 1998; Wilson et al., 2000). Several theories have been
proposed to explain both of the above factors. These are discussed further in the
final chapter in relation to happiness. Cognitive dissonance Self-deception is an
important category of irrational behavior, as we have seen. It can be the result of a
type of emotional distress discussed earlier. The most important psycho logical
theory that is relevant here is that of cognitive dissonance, originated by Festinger
(1957), and discussed in an earlier section. This theory states that people are
motivated to avoid having their attitudes and beliefs in a dissonant or conflicting
relationship, and they feel uncomfortable when dissonance occurs. This discomfort
can cause people to do many things that could be classed as irrational. Thus
cognitive dissonance generally involves people justifying their actions by changing
their beliefs. This is because it is often easier to change one’s beliefs than to change
actions that have already been taken. However, cognitive dissonance may also
involve situations where beliefs are held steadfastly in spite of contrary evidence.
This is the kind of situation where the fourth criterion for rationality is relevant.
These violations of rationality occur when it is harder to change an ingrained belief
system than to change one’s interpretation of empirical evidence. There may be a
variety of ways to explain away uncomfortable empirical findings, as many smokers
can attest. We can of course ask the question why such behavior, or the mental
processes leading to such behavior, can have evolved as an adaptive response. It
might initially seem that such processes would be maladaptive, obscuring the
realities of situations and leading to bad decisions. While self-deception may
certainly lead to bad decisions, as we will see in more detail in some case studies, it
may also be advantageous in other respects. In par ticular it may bolster confidence
and self-esteem, increasing one’s sense of well-being, and, if one can also deceive
others, it may have the effect of increasing one’s status in society. The evolutionary
psychologist Pinker (1997) has gone so far as to claim that self deception can be
adaptive because it makes it easier to deceive others. If we really believe that we
are the best person to do a certain job, in spite of our lack of ability, then we are
more likely to convince others and be offered the job. Threat to self-esteem There is
considerable evidence that concern with self-esteem can affect the quality of
decision-making. In particular there appears to be a relationship between low self-
esteem and self-defeating behavior such as self-handicapping, binge eating and
alcohol abuse. These consequences are discussed later in the section. However,
research also indicates that the relationship is not a straightforward one. People
with high, but misplaced, self esteem may also indulge in alcohol and drug abuse,
believing that they are strong enough to withstand the harmful physical effects and
the tendency to addiction. This can be referred to as the ‘peacock’s tail’ syndrome,
after the theory of the evolutionary biologist Zahavi (1975), that the seemingly
useless and wasteful peacock’s tail evolved as a sign of health for its owner, who
was strong enough to withstand the waste of resources. 142 beliefs anD
exPeCtations CH•4 Baumeister, Heatherton and Tice (1993) also found evidence of
a more complicated relationship between self-esteem and quality of decision-
making. In general they found that people with high self-esteem made better
decisions in risk-taking experiments, in terms of judging their own performance
better than people with low self-esteem, and gam bling in an appropriate manner.
However, when people with high self-esteem received a blow to their pride they
started to make bad decisions, worse even than those with low self-esteem, by
making large bets that were not justified by their own performance. They seemed to
be anxious to wipe out the loss of face involved. Failure of self-regulation Self-
regulation in the current context refers to the need for individuals to reflect on
advantages and disadvantages before making decisions rather than acting
impulsively. One aspect of this has already been described, in connection with
emotional distress. Another aspect of self-regulation involves the weighing of long-
run costs against short-run benefits of decisions. This aspect is often referred to as
intertemporal decision-making, and is discussed in Chapters 7 and 8. Self-regulation
in this situation involves the delay of gratification. The ability for self-regulation is
obviously a useful adaptation, enabling our ancestors to withstand temptations that
would have resulted in early death, and encour aging them to make long-run
investments in the health of themselves and their families. There may be different
reasons why self-regulation breaks down, as we have seen in various contexts. One
factor that can be repeated at this stage is that the capacity for self-regulation is an
exhaustible resource, much like physical strength (Muraven and Baumeister, 2000;
Muraven, Tice and Baumeister, 1998). When this capacity is reduced the
phenomenon is sometimes referred to as ego depletion, and is discussed further in
Chapter 8. For example, a study by Shiv and Fedorikhin (1999) showed that
cognitive load reduced self-control, as people who had to remember longer
numbers were more likely to eat chocolate cake than fruit salad. Other studies have
shown that the depletion effect can be eliminated by replenishing the self-
regulation resources, for example by consuming glucose (Gailliot et al., 2007; Vohs
et al., 2008), suggesting that brain blood glucose is the relevant resource. However,
there are some conflicting findings regarding failure of self-regulation and ego
depletion. Later studies have found that a placebo can serve equally well to
eliminate the depletion effect (Molden et al., 2012; Sanders et al., 2012).
Furthermore, the effect can also be offset by motivational factors, including positive
mood induction (Tice et al., 2007), explicit feedback on task performance (Wan and
Sternthal, 2008), motivational incentives (Muraven and Slessareva, 2003; Muraven,
Shmueli and Burkley, 2006), the belief that willpower is unlimited (Job, Dweck and
Walton, 2010), personal prayer (Friese and Wänke, 2014) and feelings of autonomy
support (Muraven, 2008). Another complication is that the similarity between the
capacity for self-regulation and physical strength is two-fold. First, as we have seen,
it is easily depleted in the short run, so that people cannot continue to resist
temptation indefinitely; also as they have to deal with more stress in one situation,
they tend to lose control in other situations, for example by smoking, drinking or
eating more. But there is also a second similarity: the capacity for self-regulation
appears to be something that can be increased in the long run, just as a muscle
adapts to physical exercise by becoming stronger in the long run. For example,
Muraven, Baumeister and Tice (1999) found that repeated exercises in self-control,
such as trying to improve posture, over a period of two weeks led to improve ments
in self-control in laboratory tasks relative to people who did not exercise. Similarly,
when subjects are assigned two consecutive tasks to perform which require similar
control processes, the previous use of self-control can enhance the subsequent self-
control per formance (Dewitte, Bruyneel and Geyskens, 2009). 143 PT• II
FOUNDATIONS Decision fatigue It seems that people not only tire when it comes to
self-control, they also tire of mak ing decisions in general. This may well be the
main reason that people are creatures of habit; having a routine avoids the need to
expend scarce resources by making choices. A good illustration of this phenomenon
is provided by the research of Vohs and col leagues (2008). They found that a group
of respondents who had to make a series of product choices had a reduced capacity
for self-regulation compared with a control group. The capacity for self-regulation
was measured by asking the respondents to drink as much as they could of an
unpleasant, bitter tasting beverage. This finding suggests that people tire of making
decisions, and when they do so it is possible that any further decisions that are
forced on them before they have had time to recover may result in a fall in quality.
Military psychologists have found a similar tendency with commanders in battle
(Dixon, 1976). Interpersonal rejection Humans have a strong innate desire to belong
to a social group that is virtually univer sal. The evolutionary advantages of this are
obvious, which is why this desire tends to be even greater and more fundamental
than the desire for self-esteem. However, if peo ple feel rejected socially, this
appears to be such a psychological blow that they cease to function effectively in a
number of ways. Experimental research indicates that they make poorer decisions,
making more unhealthy choices, gambling foolishly and also becoming more
aggressive and less cooperative. Even performance on intelligence tests is
adversely affected. The reasons for this general loss of effective function are not
clear at present; further research needs to be performed in this area, most likely of
a neuroscientific nature. It is likely that rejection causes a change in the body’s
output of hormones and neurotransmitters. Research has already established for
example that winning teams and their supporters both enjoy an increase in
testosterone output following victory, while los ers and their supporters suffer from
a drop in testosterone. Foundations in evolutionary neurobiology We now need to
say something in general about all the above causes of irrationality. Our starting
point is to take a reductionist approach. We must examine how the human organ
ism, and in particular the brain, evolved if we are to gain a real understanding of
behavior. The human organism did not evolve in order to be a rational decision-
making system, or to maximize utility, well-being or hedonic pleasure. The forces of
natural selection have caused us to be designed as a system that maximizes
biological fitness. Biological fitness relates not only to our own individual survival
and reproduction, but also in broader terms to the survival of our relatives who
share the same genes. Those of our ancestors who were most successful in
achieving biological fitness were most able to spread their genes, ensur ing the
survival of more people with the same genetic abilities. In order to achieve this end,
however, the brain and other body mechanisms must have a signaling system to
guide the brain to make the correct decisions. This is where pain and pleasure enter
the picture. Pain generally can tell us we have made a bad decision as far as
biological fitness is con cerned, whereas pleasure can tell us we have made a good
decision. Thus we can say that the individual is prompted to maximize hedonic
pleasure as a means to the ultimate end of maximizing biological fitness. It is
largely the indirect nature of this mechanism that leads to the objection described
earlier that inappropriate norms are used to judge rationality. Furthermore, it should
be recognized that this hedonic pleasure relates not only to conventional goods but
also to what we call moral sentiments. This point has been made 144 beliefs anD
exPeCtations CH•4 earlier, in Chapter 3, but it is important to realize that talk about
‘life being about more than happiness’ misunderstands this crucial insight. Although
morality has been heavily influenced by cultural factors over the last few thousand
years, it originally evolved for the same reason as our physical organs, to maximize
biological fitness. Ultimately this involves the same signaling system in terms of
pain and pleasure. If we feel the pain of guilt this may be a signal that we have
made a bad decision; others may punish us if they discover our actions. Likewise
the pleasure, in terms of ‘warm glow’ or pride, in perform ing a virtuous action or
‘doing our duty’ may signal a good decision; others may reward us. The essential
problem with this mechanism is that our hedonic system can be easily hijacked.
One main reason for this is that there is always a time lag between the optimal
design and the demands of the current environment. Just as the military are often
accused of preparing to fight the last war, our brains and physiological systems are
geared to deal ing with the demands of a past environment. Thus we have cravings
for salt and sugar, which in the past were vital nutrients necessary for survival, but
which now cause all kinds of health problems when consumed in excess. Our
endorphin receptors in the brain can be fooled into craving opiates as a source of
pleasure, getting us addicted to hard drugs. It can also be argued that our hedonic
system may be hijacked in the case of our moral sentiments as well; for example, a
bad or brutal environment may eliminate feelings of guilt for performing antisocial
actions. Another reason why our hedonic systems can be hijacked is that they use
heuristic devices to achieve their ends. Natural selection is a ‘blind watchmaker’, as
Dawkins (1986) has elegantly described it, and is a mechanistic rather than a
teleological process. This means that it has no ‘purpose’; it builds on what has
developed from the past, rather than by looking ahead to the future and setting a
goal. The implications of this mechanistic process are often misunderstood even by
educated and intelligent commentators, so it is worthwhile expanding on this
aspect. The maximization of biological fitness, or ‘self ish gene’ theory as it is
sometimes described in reductionist terms (Dawkins, 1976), is sometimes rejected
out of hand on the basis that it cannot explain why we use contracep tive devices
and other non-reproductive sexual practices. The response to this is that our brains
are not designed to further reproduction directly. This would involve fantastically
complicated neural machinery, which may well not adapt well to changes in the
envi ronment, and which would use great resources of precious energy. Instead, our
brains operate using basic heuristic processes, so that sexual activity in general
gives pleasure, regardless of whether it results in reproduction. The association of
sexual activity with pleasure is generally sufficient to promote reproduction, and
certainly has been through out evolutionary history until very recently. On this
foundation of neurobiology, Zak has argued that people are neither rational nor
irrational; rather, people are ‘rationally rational’ (Zak, 2011): The rational rationality
model predicts that people will invest scarce cognitive resources in solving a
decision problem only when the expected payoff is suffi ciently large. Otherwise,
human beings will expend the minimum resources needed to achieve a “good
enough” outcome. “Good enough” means that there is a wide range of acceptable
choices. Rational rationality is similar to Herbert Simon’s notion of satisficing
(Simon, 1991), but clearly identifies when people will satisfice and when they will
not. Rational rationality occurs because cognitive resources are constrained and the
brain evolved to conserve energy and deploy these resources only as needed. The
neuroscience behind rational rationality requires that any economic model identify
why individuals would expend scarce brain resources when making a decision
rather than rely on previously learned heuristics. (p. 55) Thus, as we have already
seen, it may make sense to formulate dual-process models of reasoning and
judgment, which involve the operation of different decision-making systems 145 PT•
II FOUNDATIONS in different situations (Epstein, 1994; Osherson, 1995; Evans and
Over, 1996; Sloman, 1996; Stanovich, 1999). The essence of such models is that in
certain situations people use analytical, logical, rule-based systems with a relatively
high computational burden, while in other situations people use various types of
heuristic procedures. The use of heuristics can be viewed as a shortcut; frequently it
results in an efficient use of personal resources, leading if not to optimization at
least to satisficing. However, like many shortcuts, the use of heuristics can also lead
to many bad decisions in situations where a more cognitive, analytical approach is
desirable. Thus heuristics are both a good and a bad method of decision-making,
depending on the circumstances. If we accept the abundant evidence of the
mechanistic nature of evolution and the way it builds structures without purpose,
the implication of the use of heuristic devices is that they provide simple rules for
appropriate action in a given situation, but they tend to be highly fallible. Many of
the anomalies that we have now observed with the stand ard model are a result of
this factor. Therefore, because of the way in which our brains and minds have
evolved, we may be bad at performing what may seem simple abstract tasks, using
inappropriate heuristic devices. However, these are tasks that have never been
required in our ancestral past. On the other hand, human beings are extremely
good at performing complex tasks that we take for granted, like visually following
an object, changing focus and perceptions of color, speed and distance as it moves,
and making the necessary biomechanical adjustments involved in catching a ball.
Even the most advanced artificial intelligence systems designed cannot rival this
performance. The moral appears to be that we are good at what we need to be
good at, or, more correctly, we are good at what we needed to be good at in our
evolutionary past. This kind of behavior may then not be so ‘irrational’ after all; this
raises the issue of appropriate norms, which is discussed further in the final
chapter. At the psychological level, the kind of ‘irrationality’ that we observe in
human belief systems may not really be irrational in evolutionary terms either, in
spite of initial appear ances. Hood (2010) has claimed that the human mind has
adapted to reason intuitively, in order to develop theories about how the world
works even when mechanisms cannot be seen or easily deduced. This adaptation
has had huge benefits in terms of the development of scientific theories related to
invisible forces like gravity and electromagnetism. How ever, according to Hood, it
also results in people being prone to making irrational errors, in particular relating
to superstition and religion. This is because in our evolutionary past it has been
more advantageous from a survival viewpoint to believe in a cause-and-effect
relationship that does not exist (for example, God punishing people with bad
weather) than not to believe in a cause-and-effect relationship that does exist (for
example, the growl behind the nearby bush being caused by a lurking predator).
Thus people tend to be overly fond of positing cause-and-effect relationships, even
when none exists. Hood claims that it is therefore unlikely that we will evolve a
rational mind, and that religion and superstition are here to stay.