Content uploaded by Roland Benabou
Author content
All content in this area was uploaded by Roland Benabou
Content may be subject to copyright.
SELF-CONFIDENCE AND PERSONAL MOTIVATION*
ROLAND BE
´NABOU AND JEAN TIROLE
We analyze the value placed by ratio nal agents on self-condence, and the
strategies employed in its pursuit. Condence i n one’s abiliti es generally en-
hances m otivation, ma king it a valuable asset for individuals with imperfe ct
willpower. This d emand for se lf-serving beliefs (wh ich can als o arise from hedoni c
or signaling motives) must be we ighed against the risks of overcondence. On the
supply s ide, we develop a model of s elf-deception through endogenous memory
that reconciles the motivated and rational features of human cognition. The
resulting i ntrapersonal ga me of strategic communication typically leads to mul-
tiple eq uilibria. Wh ile “positive thinking” can im prove welfare, it can also be
self-defea ting (and nonetheless pursued).
Believe what is in the line of your needs, for on ly by such belief is the need
fullled . . . Have f aith that you can successfully ma ke it, and your feet a re
nerved to its accom plishment [William James, Principles of Psychology].
I have done this, says my memory. I cannot have done that, says my pride,
remaining inexorable. Finally—memory yields [Friedrich Nietzsche, Beyond
Good and Evil].
I had during many years followed the Golden Rule, namely, that whenever a
published fa ct, a new observation or thought came across me, which wa s
opposed to my gen eral results, to make a memorandum of it without fail and
at once; for I ha d found by experie nce that such (contrary and thus unwel-
come) facts and thoughts were far more apt to escape from memory than
favorable o nes [Charles Darwin in The Life of Ch arles Darwin, by Francis
Darwin].
INTRODUCTION
The maintenance and enhancement of self-esteem has al-
ways been identied as a fundamental human impulse. Philoso-
phers, writers, educators, and of course psychologists all have
emphasized the crucial role played by self-image in motivation,
affect, and social interactions. The aim of this paper is to bring
* This paper was pre viously title d “Self-Condence: Intrapersonal Strategie s
[1999].” Fo r helpful comments and discussion we are grateful to Dilip Abreu,
Olivier Blanchard, Isabelle Brocas, Edward Glaeser, Daniel Gilbert, Ian Jewitt,
David Laibson, George Loewenstein, Andrew Postle waite, Marek Pycia, Matth ew
Rabin, Julio Rotemberg, and three anonymous referees. We also thank confer-
ence and sem inar participan ts at the Uni versity of Chicago, Columbia Uni-
versity, Cornell University, the Massachusetts Institute of Technology, the
National Bure au of Economic Research, Northwestern University, New York
University, the Oxford Young Eco nomists’ Conference at Oxford University, the
University of Pennsylvania, Princeton Univers ity, Stanford University, and Yale
University. Be´na bou gratefully acknowledges nancial suppo rt from the National
Science Foundation (SES-0096431).
©2002 by the President and Fel lows of Harvard College and the Massachuse tts Institute of
Technology.
The Quarterly Journal of Economics, August 2002
871
these concerns into the realm of economic analysis, and show that
this has important implications for how agents process informa-
tion and make decisions. Conversely, the tools of economic mod-
eling can help shed light on a number of apparently irrational
behaviors documented by psychologists.
Indeed, both the demand and the supply sides of self-con-
dence appear at odds with economists’ view of human behavior
and cognition. Why should people prefer rosy views of themselves
to accurate ones, or want to impart such beliefs to their children?
From car accidents, failed dot.com rms, and day trading to the
space shuttle disaster and lost wars, the costs of overcondence
are plain for all to see. Even granting that some “positive illu-
sions” could be desirable, is it even possible for a rational, Bayes-
ian individual to deceive himself into holding them? Finally, the
welfare consequences of so-called self-serving beliefs are far from
clear: while “thinking positive” is often viewed as a good thing,
self-deception is not, even though the former is only a particular
form of the latter.
To analyze these issues, we develop a simple formal frame-
work that unies a number of themes from the psychology liter-
ature, and brings to light some of their economic implications. We
rst consider the demand side of self-condence, and identify in
Section I three main reasons why people may prefer optimistic
self-views to accurate ones: a consumption value, a signaling
value, and a motivation value. First, people may just derive
utility from thinking well of themselves, and conversely nd a
poor self-image painful. Second, believing—rightly or wrongly—
that one possesses certain qualities may make it easier to con-
vince others of it. Finally, condence in his abilities and efcacy
can help the individual undertake more ambitious goals and
persist in the face of adversity. While we shall mostly focus on
this last explanation, all three should be seen as complementary,
and for many purposes work equally well with the supply side of
our model (self-deception).
The main reason why we emphasize the motivation theory is
its substantially broader explanatory power. Indeed, it yields an
endogenous value of self-condence that responds to the situa-
tions and incentives which the individual faces, in a way that can
account for both “can-do” optimism and “defensive” pessimism. It
also readily extends to economic and social interactions (altruistic
or not), explaining why people generally prefer self-condent
872 QUARTERLY JOURNAL OF ECONOMICS
partners to self-doubting ones, and invest both time and effort in
supporting the latter’s morale.
The rst premise of the motivation theory is that people have
imperfect knowledge of their own abilities, or more generally of
the eventual costs and payoffs of their actions.
1
The second one is
that ability and effort interact in determining performance; in
most instances they are complements, so that a higher self-con-
dence enhances the motivation to act. As demonstrated by the
opening quote from James [1890], this complementarity has long
been familiar in psychology.
2
It is also consistent with the stan-
dard observation that morale plays a key role in difcult endeav-
ors; conversely, when people expect to fail, they fail quite effec-
tively, and failure leads to failure more readily for individuals
characterized with low self-esteem [Salancik 1977].
The fact that higher self-condence enhances the individual’s
motivation gives anyone with a vested interest in his performance
an incentive to build up and maintain his self-esteem. First, the
manipulator could be another person (parent, teacher, spouse,
friend, colleague, manager) who is eager to see him “get his act
together,” or otherwise apply himself to the task at hand. Such
interpersonal strategies are studied in Be´nabou and Tirole [2001].
Second, for an individual suffering from time inconsistency (e.g.,
hyperbolic discounting), the current self has a vested interest in
the self-condence of future selves, as it helps counter their
natural tendency to quit too easily. It is in this context, which
builds on Carrillo and Mariotti [2000], that we shall investigate a
variety of intrapersonal strategies of self-esteem maintenance.
We shall thus see how and when people may choose to remain
ignorant about their own abilities, and why they sometimes de-
liberately impair their own performance or choose overambitious
tasks in which they are sure to fail (self-handicapping).
Section II thus turns to the supply side of the self-condence
1. The psychology literature generally views introspection as quite inaccu-
rate [Ni sbett and Wilson 1977], and s tresses that lea rning abo ut oneself is an
ongoing proce ss. Furthermore, the self is constantly changing (e.g., Rhodenwalt
[1986]): personal characteristics evolve with age, the goals purs ued shift over
one’s career and life cycle (often as the result of in teractions with others), and the
personal o r economic environment in which these objectives are rewarded is
typically variab le.
2. Thus, Gilbert a nd Coope r [1985] note that “the classic attributional model
of the causes of behavior . . . [is described by] the well-kno wn conceptual equation:
(E3A)6TD 5B, in which effort times ability, pl us or minus task difculty
equals the behavioral outcome.” Additional references are given in Section I. Note,
however, that there are also instances where ability and effort are substitutes. As
discussed b elow, we s hall consi der this case as well.
873SELF-CONFIDENCE AND PERSONAL MOTI VATION
problem, and the “reality constraints” that limit the extent to
which people can engage in wishful thinking. In our model we
maintain the standard assumption of individuals as rational
(Bayesian) information processors. While almost universal in eco-
nomics, this view is more controversial in psychology. On one
hand, a lot of the classical literature has emphasized rationality
and information-seeking in the process of self-identication,
documenting the ways in which people update their beliefs accord-
ing to broad Bayesian principles.
3
On the other hand, the more
recent cognitive literature abundantly documents the less ratio-
nal (or at least, subjectively motivated) side of human inference.
For instance, a substantial body of evidence suggests that
people tend to recall their successes more than their failures, and
have self-servingly biased recollections and interpretations of
their past performances.
4
Similarly, they tend to overestimate
their abilities and other desirable traits, as well as the extent to
which they have control over outcomes. They also rate their own
probabilities as above average for favorable future life events,
and below average for unfavorable ones; the more controllable
these events through their future actions, the more so.
5
We shall capture this class of self-deception phenomena with
a simple game-theoretic model of endogenous memory, or aware-
ness-management, which represents one of the main contribu-
tions of this paper. Drawing on evidence about the mechanics and
limitations of memory, it shows how to reconcile the motivated
(“hot”) and rational (“cold”) features of human cognition, and
could be used in any setting where a demand for motivated beliefs
arises. The basic idea is that the individual can, within limits and
3. Thus, attribution theory [He ider 1958] emphasizes the distinction between
temporary (situation al) and enduring (dispos itional) characte ristics. In econo mics
parlance, the individua l lters out noise in o rder to extract info rmation fro m past
events. In the social comparison process [Festinger 1954] individuals assess their
ability by comparing their performance with that of people facing similar condi-
tions (famili al, cultural, education al, etc.). In other words, they use “relative
performance evaluation,” or “benchmarking,” for self-evaluation. A good perfor-
mance by others in one’s reference group is thus generally detrimental to self-
esteem, a nd conversely some comfort is derived wh en others expe rience adversity
(Schadenfreude). Relatively sophisticated updating also applies to the interpre-
tation of praise and criticism : a person take s into account not onl y what others s ay
(or do), but also their possible intentions.
4. Why they would want to do so in a social conte xt is obvious. The interesting
question is why they may bias their own inference process.
5. See, e.g., Taylor and Brown [1988], Weinstein [1980], Alloy and Abra-
hamson [1979], and the many other references given in Section II. For recent
overviews of the general ph enomenon of self-deception, see Gilbert and Cooper
[1985] and e specially Baume ister [1998] o n the ps ychologi cal evidence, a nd Elster
[1999] and Mele [1999] for the philosophical debates and implications.
874 QUARTERLY JOURNAL OF ECONOMICS
possibly at a cost, affect the probability of remembering a given
piece of data. At the same time, we maintain rational inference,
so people realize (at least to some extent) that they have a selec-
tive memory or attention.
The resulting structure is that of a game of strategic commu-
nication between the individual’s temporal selves. In deciding
whether to try to repress bad news, the individual weighs the
benets from preserving his effort motivation against the risk of
becoming overcondent. Later on, however, he appropriately dis-
counts the reliability of rosy recollections and rationalizations.
The implications of this game of asymmetric information are
quite different from those of ex ante decisions about information
acquisition (e.g., self-handicapping or selective search). In particu-
lar, multiple intrapersonal equilibria (“self-traps”) may arise,
ranging from systematic denial to complete self-honesty. More
generally, we characterize the set of perfect Bayesian equilibria
and its dependence on the individual’s degree of time inconsis-
tency and repression costs (“demand and supply” parameters).
The model also has interesting implications for the distribu-
tion of optimism and pessimism across agents, which we examine
in Section III. We show that when the costs of repression are low
enough, most people typically believe themselves to be more able
than they actually are, as well as more able than both the average
and the median of the population. A minority will have either
realistically low assessments, or actually severely underestimate
themselves. We also highlight the key role played by Bayesian-
like introspection (understanding, at least partially, one’s own
incentives for self-esteem maintenance) in the model’s results,
and why incorporating this essential human trait is more fruitful
than modeling agents as naively taking all recollections and
self-justications at face value.
Section IV examines the welfare impact of equilibrium self-
deception. Is a more active self-esteem maintenance strategy,
when chosen, always benecial? How can people be “in denial” if
it does not serve their best interests? We show that, in addition to
the trade-off mentioned earlier between the condence-mainte-
nance motive and the risks of overcondence, ex ante welfare
reects a third effect, namely the spoiling of good news by self-
doubt. Intuitively, when adverse signals about his ability are
systematically repressed, the individual can never be sure that
only positive ones were received, even when this is actually true.
We characterize the conditions under which always “looking at
875SELF-CONFIDENCE AND PERSONAL MOTI VATION
the bright side” pays off on average or, conversely, when it would
be better to always “be honest with yourself,” as Charles Darwin
apparently concluded.
In Section V we turn to the case where ability and effort are
substitutes rather than complements. This typically occurs when
the payoff for success is of a “pass-fail” nature, or characterized by
some other form of satiation. Since a high perceived ability may
now increase the temptation to exert low effort (“coasting”), this
case allows us to account for what psychologists refer to as “de-
fensive pessimism:” the fact that people sometimes minimize,
rather than aggrandize, their previous accomplishments and ex-
pectations of future success. Another variant of the model consid-
ered in this section involves replacing the motivation value of
self-condence with a purely affective one. Section VI concludes
the paper. All proofs are gathered in the Appendix.
This paper is related to several strands of the new literature
that tries to better link economics and psychology. A hedonic
concern for self-image, in the form of preferences over beliefs, was
rst explored in Akerlof and Dickens’ [1982] well-known model of
dissonance reduction, and more recently in Rabin [1995], Wein-
berg [1999], and Ko¨szegi [1999]. In emphasizing an endogenous
value of self-condence and retaining the constraint of Bayesian
rationality, our paper is most closely related to the work of Car-
rillo and Mariotti [2000], who rst showed how information ma-
nipulation may serve as a commitment device for time-inconsis-
tent individuals (see also Brocas and Carrillo [1999]). The central
role played by memory also relates our model to Mullainathan
[2002] and Laibson [2001], although one of its main features is to
make recall endogenous.
I. THE DEMAND FOR SELF-CONFIDENCE
In most societies, self-condence is widely regarded as a
valuable individual asset. Going back at least to William James,
an important strand in psychology has advocated “believing in
oneself ” as a key to person al succ ess. Today, an eno rmous “self-
help” industry ourishes, a sizable part of which purports to help
people improve their self-esteem, shed “learned helplessness” and
reap the benets of “learned optimism.”
6
American schools place
6. These last two terms are borrowed from Seligman [1975, 19 90].
876 QUARTERLY JOURNAL OF ECONOMICS
such a strong emphasis on imbuing children with self-condence
(“doing a great job!”) that they are often criticized for giving it
preeminence over the transmission of actual knowledge. Hence
the general question: why is a positive view of oneself, as opposed
to a fully accurate one, seen as such a good thing to have?
Consumption value. A rst reason may be that thinking of
oneself favorably just makes a person happier: self-image is
then simply another argument in the utility function. Indeed,
psychologists emphasize the affective benets of self-esteem as
well as the functional ones on which we shall focus. One may also
hypothesize that such preferences over beliefs could have been
selected for through evolution: the overcondence that typically
results may propel individuals to undertake activities (explora-
tion, foraging, combat) which are more risky than warranted by
their private material returns, but confer important external
benets on the species. In subsection V.B we shall explain how a
hedonic self-image motive can readily be incorporated into our
general framework.
Signaling value. A second explanation may be that believing
oneself to be of high ability or morality makes it easier to convince
others (rightly or wrongly) that one does have such qualities.
Indeed, it is often said that to lie most convincingly a person must
believe his own lies. While the idea that people are “transparent”
and have trouble misrepresenting their private information may
seem unusual in economics, one could easily obtain an instru-
mental value of self-condence from a signaling game where
those who truly believe in their own abilities face lower costs of
representing themselves favorably to others.
Motivation value. The explanation that we emphasize most
is that self-condence is valuable because it improves the indi-
vidual’s motivation to undertake projects and persevere in the
pursuit of his goals, in spite of the setbacks and temptations that
periodically test his willpower. Morale is universally recognized
as key to winning a medal, performing on stage, getting into
college, writing a great book, doing innovative research, setting
up a rm, losing weight, nding a mate, and so forth. The link
between self-condence and motivation is also pervasive in the
psychology literature, from early writers like James [1890] to
contemporary ones like Bandura [1977], according to whom “be-
877SELF-CONFIDENCE AND PERSONAL MOTI VATION
liefs of personal efcacy constitute the key factor of human
agency” (see also, e.g., Deci [1975] or Seligman [1990]). The mo-
tivation theory also readily extends to economic (nonaltruistic)
interactions, explaining why people typically prefer self-condent
coworkers, managers, employees, teammates, soldiers, etc., to
self-doubting ones; and why they spend substantial time and
effort supporting the morale of those with whom they end up
being matched.
7
I.A. The Motivation Problem
Had I been less deniti vely determined to start working, I might have made
an effort to begin right away. But because my resolve was absolute and,
within twenty-four hours, in the empty frames of the next day where every-
thing t so well since I was not yet there, my go od reso lutions wo uld easily
be accomplished, it was better n ot to choose an evening where I was ill
disposed for a beginning to which, alas! the foll owing days would turn out to
be no more propitious [Marcel Proust, Remembrance of Things Past].
Consider a risk-neutral individual with a relevant horizon of
three periods: t50, 1,2. At date 0 he selects an action that
potentially affects both his ow payoff u
0
and his date 1 informa-
tion structure.
8
At date 1 he decides whether to undertake a task
or project (exert effort, which has disutility cost c.0) or not
(exert no effort). With some probability u, which denes his abil-
ity, the project will succeed and yield a benet Vat date 2; failure
generates no benet. The individual’s beliefs over u(dening his
self-condence or self-esteem) are described by distribution func-
tions F(u) at date 0 and F
1
(u) at date 1. In the intervening period
new information may be received, or previous signals forgotten;
we shall focus here on the rst, more standard case, and turn to
memory in Section II. Note that with risk neutrality the mean
u
#
1
[ *
0
1
udF
1
(u) will be a sufcient statistic for F
1
; for brevity,
we shall also refer to it as the agent’s date 1 self-condence.
Finally, we assume that the individual ’s preferences exhibit
time-inconsistency, due to quasi-hyperbolic discounting. There is
indeed considerable experimental and everyday evidence that
intertemporal choices exhibit a “salience of the present,” in the
7. Note that this last observation cannot readily be accounted for by the
“signaling” theory of self-condence either.
8. The simplest date 0 action is thus the choice of the amount of information
that will be available at date 1 (e.g., soliciting feedback, taking a test, keeping or
destroying records). Alternatively, this information may be derived from the
outcome of some activity pursued for its own sake at date 0 (le arning by doing,
drinking a l ot of wine).
878 QUARTERLY JOURNAL OF ECONOMICS
sense that discount rates are much lower at short horizons than
at more distant ones.
9
Denoting u
t
and E
t
[z] the ow payoffs and
expectations at t50, 1,2, the intertemporal utility perceived by
the individual as of date 1 is
(1) u11bdE1@u2#52c1bdu
#1V
when he undertakes the activity, and 0 when he does not. By
contrast, intertemporal utility conditional on the same informa-
tion set at date 1, but evaluated from the point of view of date 0,
is
(2) u01bE0@du11d2u2uu
#1#5u01bd@2c1du
#1V#
if the activity is undertaken at date 1, and u
0
otherwise.
10
Whereas dis a standard discount factor, breects the momentary
salience of the present. When b , 1, the individual at date 0 (“Self
0”) is concerned about his date 1 (“Self 1’s”) excessive preference
for the present, or lack of willpower, which leads to the under-
provision of effort (procrastination). Indeed, Self 1 only exerts
effort in the events where u
#
1
.c/bdV, whereas, from the point of
view of Self 0, it should be undertaken whenever u
#
1
.c/dV. Note
that while we focus here on the case where the individual’s
intrinsic ability uis unknown, it could equally be the expected
payoff in case of success V, the “survival” probability d, or the
task’s difculty, measured by the cost of effort c. All that matters
for our theory is that the individual be uncertain of the long-term
return to effort udV/cwhich he faces.
I.B. Condence Maintenance versus Overcondence
In an important paper Carrillo and Mariotti [2000] showed
that, in the presence of time inconsistency (TI), Blackwell gar-
blings of information may increase the current self’s payoff. This
result can be usefully applied, and further developed, in our
context.
Suppose that, at date 0, our individual can choose between
just two information structures for date 1. In the ner one, Self 1
learns his ability uexactly. In the coarser one, he learns nothing
9. See Ainslie [1992, 20 01] for the evidence, and Strotz [ 1956], Phelps and
Pollack [1968], Loewenstein and Prelec [1992], Laibson [1997, 2001], and
O’Donoghue and Rabin [1999] for formal model s and economic implications.
10. Note th at the equality in (2) makes use of the identity E
0
[uuE
1
[u]5 u
#
1
]5
u
#
1
, which holds when—and o nly when—there is no informati on loss betwee n dates
0 and 1.
879SELF-CONFIDENCE AND PERSONAL MOTI VATION
that Self 0 did not know: F
1
(u)5F(u), and hence u
#
1
5 *
0
1
u
dF
1
(u)[ u
#
F
. Let us rst assume that, in the absence of infor-
mation, Self 1 will undertake the task: u
#
F
.c/bdV. The value
attached by Self 0 to Self 1’s learning the value of uis therefore bd
times
(3) F;
E
c/bdV
1
~duV2c!dF ~u ! 2~du
#FV2c!5F2F,
where
(4) F;
E
0
c/dV
~c2duV!dF~u ! ,
(5) F;
E
c/dV
c/bdV
~duV2c!dF~u ! .
F
stands for the gain from being informed, which arises from the
fact that better information reduces the risk of overcondence on
the part of Self 1. Overcondence occurs when the individual’s
ability is below c/dVbut he is unaware of it, and thus inappro-
priately undertakes or perseveres in the project.
F
stands for the
loss from being informed, which may depress the individual’s
self-condence: if he learns that uis in some intermediate range,
c/dV,u,c/bd V, he will procrastinate at date 1 even though, ex
ante, it was optimal to work. Information is thus detrimental to
the extent that it creates a risk that the individual will fall into
this time-inconsistency (TI) region. If this condence maintenance
motive is strong enough (
F
.
F
), the individual will prefer to
remain uninformed:
F
,0. More generally, note that
F
is lower,
the lower is b. By contrast, in the absence of time inconsistency
(b 5 1) we have
F
50, and thus
F
$0: in classical decision
theory, information is always valuable.
The overcondence effect calls for more information, con-
dence maintenance for less. This trade-off has been noted by
empirical researchers. For instance, Leary and Downs [1995]
summarize the literature by noting that a) “persons with high
self-esteem perform better after an initial failure and are more
likely to persevere in the face of obstacles;” b) “high self-esteem is
not always functional in promoting task achievement. People
with high self-esteem may demonstrate nonproductive persis-
880 QUARTERLY JOURNAL OF ECONOMICS
tence at insoluble tasks, thereby undermining their effectiveness.
They may also take excessive and unrealistic risks when their
self-esteem is threatened.”
To understand the last statement, let us turn to the case
where u
#
F
,c/bdV. Since Self 1 now always exerts (weakly) less
effort than Self 0 would like him to, information can only help the
individual restore his decient motivation. Indeed,
(6) F5
E
c/bdV
1
~duV2c!dF~u ! .0.
Moreover,
F
is now higher, the lower is b. In such situations the
individual will avidly seek feedback on his ability, and his choices
of tasks and social interactions will have the nature of “gambles
for resurrection” of his self-esteem.
Putting together the different cases, we see that the value of
information is not monotonic with respect to initial self-con-
dence. Indeed, for someone with condence so low that u
#
F
,
c/bd V,
F
is always positive and increasing with respect to (sto-
chastic) increases in u.
11
For an individual with F(c/dV)50 but
F(c/bd V),1,
F
is always negative. Finally, for a person so
self-assured that F(c/bd V)50, m otivation is not a concern (as if
bwere equal to 1), but neither is overcondence:
F
5
F
5
F
5
0. Therefore, there must exist some intermediate range where
F
rst declines and becomes negative, then increases back toward
zero.
I.C. What Types of People Are Most Eager to Maintain Their
Self-Condence?
Let us now consider two individuals with different degrees of
initial self-condence, and ask which one is least receptive to
information. We denote their prior distributions over abilities as
F(u) and G(u), with densities f(u),g(u) and means u
#
F
,u
#
G
. To
make condence maintenance meaningful, let u
#
F
. u
#
G
.c/bdV.
For comparing levels of self-condence, however, just looking at
expected abilities turns out not to be sufcient.
DEFINITION 1. An individual with distribution Fover ability uhas
higher self-condence than another one with distribution Gif
the likelihood ratio f(u)/ g(u) is increasing in u.
11. Rewrite (6) as
F
5 *
0
1
1
{u$c/bdV}
(duV2c)d F(u) , where 1
{
z
}
denotes th e
indicator function, and note that the integrand is increasing in u.
881SELF-CONFIDENCE AND PERSONAL MOTI VATION
Abstracting for the moment from any cost attached to learn-
ing or not learning the true ability, it is easy to see from (3) that
F
$0 if and only if
(7)
E
0
c/bdVF~u !
F~c/bd V!du$
S
12b
b
D S
c
dV
D
.
The monotone likelihood ratio property (MLRP) implies that
F(u)/ F(c/bd V)#G(u)/G(c/bdV) for all u#c/bd V. Therefore, the
left-hand side of (7) is smaller under Fthan under G, meaning
that the person with the more positive self-assessment will accept
information about his ability for a smaller set of parameters.
Intuitively, he has more to lose from information, and is therefore
more insecure.
PROPOSITION 1. If an individual prefers not to receive information
in order to preserve his self-condence, so will anyone with
higher initial self-condence: if
G
,0 for some distribution
Gover u, then
F
,0 for any distribution Fsuch that the
likelihood ratio f/gis increasing.
I.D. Self-Handicapping
A well-documented and puzzling phenomenon is that people
sometimes create obstacles to their own performance.
12
In experi-
ments, subjects with fragile self-condence opt to take perfor-
mance-impairing drugs before an intelligence test. In real life,
people withhold effort, prepare themselves inadequately, or drink
alcohol before undertaking a task. They also set themselves over-
ambitious goals, where they are almost sure to fail. Test anxiety
and “choking” under pressure are yet other common examples.
Psychologists have long suggested that self-handicapping is often
a self-esteem maintenance strategy (instinctive or deliberate),
directed both at oneself and at others.
13
12. See, e.g., Berglas and Jo nes [1978], Arkin and Baumgardner [1985],
Fingarette [1985], or Gilovich [1991].
13. See, e.g., Berglas and Baumeister [1993]. Of course, self-handicapping
involves both intrapersonal (self-esteem mainte nance) and intrapersonal (self-
presentatio n) motives; our model captures only the former. As Baume ister [199 8]
notes, “by self-handicapping, one can forestall the drawing of unattering attri-
butions about oneself. Self-handicapping makes failure meaningless, and so if
people think you are intelligent the upcoming test cannot change this impression.”
In particular, people apparently self- handicap more in public situations [Kolditz
and Arkin 1982]. They then reap a double dividend, as this provides an excuse for
poor performance both to themselves and to others.
882 QUARTERLY JOURNAL OF ECONOMICS
To examine this question, consider an individual with prior
beliefs F(u), faced at date zero with a choice between an efcient
action that (for simplicity) will fully reveal his ability, and an
inefcient, “self-handicapping” one that entails an expected cost
h
0
(F)$0, but is totally uninformative about u. Assuming that
u
#
F
.c/bd Vas before, equation (7) immediately generalizes to
show that he will self-handicap if and only if 2bd
F
$h
0
(F), or
(8)
S
12b
b
D
c2dV
F
E
0
c/bdVF~u !
F~c/bd V!du
G
$h0~F!
bd F~c/bdV!.
Note rst that multiplying (8) by F(c/bdV) yields a decreasing
function of bon the left-hand side. Therefore, people who are
more concerned about sustaining motivation (more time-inconsis-
tent) are more likely to self-handicap, and will choose to do so
when the short-run costs of doing so are not too large. Next, let us
compare individuals with different prior beliefs about them-
selves. As before, those who are initially more self-condent have
more to lose from learning about their ability: by the MLRP, the
left-hand side of (8) is larger than if Fwere replaced with G.
However, the more self-condent are also less likely to receive bad
news, and this reduces the return on “investing in ignorance:” the
MLRP implies that F(c/bdV)#G(c/bdV), which tends to make
the right-hand side of (8) also larger under Fthan under G. Thus,
in general one cannot conclude whether people with higher or
lower self-condence are more likely to self-handicap.
14
This am-
biguity is fundamentally linked to the nonmonotonicity noted
earlier for the value of information: while the MLRP ensures that
the sign of
F
varies monotonically with initial beliefs, the abso-
lute amount does not. When self-handicapping costs are relatively
small, however—which is often the case in experiments—the
“more to lose” effect identied in Proposition 1 will prevail.
It is interesting in this respect to note that psychologists
have also not reached a rm conclusion on whether high or low
self-condence people are the most defensive of their egos, al-
though there does seem to be somewhat more evidence in favor of
the rst hypothesis. Thus, Greenier, Kernis, and Wasschul [1995]
contrast “humanistically oriented theories, . . . according to
14. A third consideration is whether the expected cost of self-handicapping
rises or falls with initial self-condence: h
0
(F):h
0
(G). In Be´nabou and Tirole
[2001] we provide examples of tasks that correspond to ea ch case.
883SELF-CONFIDENCE AND PERSONAL MOTI VATION
which high self-esteem individuals’ feelings of self-worth are built
on solid foundations that do not require continual validation,” with
experimental research showing that “high self-esteem individuals
are the more likely to display self-serving attributions, self-hand-
icap to enhance the potentially positive implications of good per-
formance, set inappropriately risky goals when ego-threatened,
and actively create less fortunate others with whom they can
compare favorably.”
II. THE PSYCHOLOGICAL IMMUNE SYSTEM
Just as there wa s in his study a ch est of drawers whi ch he managed never to
look at, whi ch he went out of his way no t to encounter when wa lking in or out,
because in one drawer was held the chrysanthemum which she had given
him on the rst evenin g, . . . so there was in him a place wh ich he never let
his mind approach, imposing on it, if necessary, the de tour of a long reason-
ing . . . : it was there that lived the memories of happier days [Marcel Proust,
Remembrance of Things Past].
We now turn to the supply side of the self-esteem problem.
Given that a positive self-assessment may be desirable (whether
for motivational, signaling, or hedonic reasons), what are the
means through which it can be achieved, or at least pursued?
Wired-in optimism. A rst hypothesis could be that evolution
has selected for a particular cognitive bias in humans, causing
them to systematically and involuntarily underweigh adverse
signals about themselves, and overweigh positive ones. This ex-
planation is rather problematic: the extent of overcondence or
overoptimism varies both over time and across tasks, and a great
many people actually suffer from undercondence (the extreme
case being depression). Furthermore, individuals often “work”
quite hard at defending their self-image when it is threatened,
going though elaborate schemes of denial, self-justication, fur-
niture-avoidance, and the like.
Blissful ignorance. When self-condence is valuable capital,
it may be preferable to remain uninformed than to put it at risk
by exposing oneself to new information. In particular, as seen in
the previous section, ex ante strategic ignorance [Carrillo and
Mariotti 2000] or even self-handicapping may help a time-incon-
sistent individual safeguard his motivation. In a context of he-
donic beliefs, workers in a hazardous job may not want to know
about the exact risks involved [Akerlof and Dickens 1982].
884 QUARTERLY JOURNAL OF ECONOMICS
Self-deception. Most often, the relevant issue is not whether
to seek or avoid information ex ante (before knowing what it will
turn out to say), but how to deal with the good and especially the
bad news concerning one’s performances and abilities that life
inevitably brings. This is where the mechanisms of defensive
denial, repression, self-serving attributions, and the like, so
prominently emphasized in psychology, come into play. We shall
capture this class of phenomena with a simple game-theoretic
model of endogenously selective memory.
II.A. Managing Awareness: The Role of Memory
Psychologists, and before them writers and philosophers,
have long documented people’s universal tendency to deny, ex-
plain away, and selectively forget ego-threatening information.
Freudian repression is the most obvious example, but various
other forms of motivated cognition and self-deception gure promi-
nently in contemporary psychology. Thus, a lot of research has
conrmed that people tend to recall their successes more than
their failures (e.g., Korner [1950] and Mischel, Ebbesen, and
Zeiss [1976]), have self-servingly biased recollections of their past
performances [Crary 1966] and readily nd “evidence” in their
personal histories that they possess characteristics which they
view (sometimes as the result of experimental manipulation) as
correlated with success in professional or personal life [Kunda
and Sanitioso 1989; Murray and Holmes 1993]. Similarly, they
often engage in “beneffactance,” viewing themselves as instru-
mental for good but not bad outcomes [Zuckerman 1979]. When
they commit a bad deed, they reframe the facts to try and con-
vince themselves that it was not so bad (“he deserved it,” “the
damage was limited”), or attribute the responsibility to others
[Snyder 1985].
At the same time, the impossibility of simply choosing the
beliefs we like has always stood in the way of a fully consistent
theory of self-deception. Sartre [1953] argued that the individual
must simultaneously know and not know the same information.
Gur and Sackeim [1979] dened self-deception as a situation in
which a) the individual holds two contradictory beliefs; b) he is
not aware of holding one of the beliefs; c) this lack of awareness is
motivated.
Our intertemporal model allows us to unbundle the “self that
knows” from the “self that doesn’t know,” and thereby reconcile
the motivation (“hot”) and cognition (“cold”) aspects of self-decep-
885SELF-CONFIDENCE AND PERSONAL MOTI VATION
tion within a standard information-theoretic framework. The ba-
sic idea is that the individual can, within limits, affect the prob-
ability of remembering a given piece of data. Under time
inconsistency, there is an incentive to try to recall signals that
help sustain long-term goals, and forget those that undermine
them. This is the motivation part.
15
On the other hand, we main-
tain the rational inference postulate, so people realize (at least to
some extent) that they have a selective memory or attention. This
is the cognition part.
ASSUMPTION 1 (MEM ORY OR AWARENESS MANAGEMENT). The individ-
ual can, at a cost, increase or decrease the probability of
remembering an event or its interpretation.
Formally, let l[[0,1] denote the probability that given
information received at date 0 will be recalled or accessed at date
1. We dene the natural rate of recall l
N
[[0,1] as that which
maximizes the date 0 ow payoff u
0
. Increasing or decreasing l
thus involves a “memory cost” M(l), i.e., a reduction in u
0
, with
M(l
N
)50, M9(l)#0 for l , l
N
and M9(l)$0 for l . l
N
.
16
Whether it refers to the subconscious or points to the differ-
ential accessibility and decay of memories stored in specialized
parts of the brain, virtually all modern psychology recognizes that
only part of the individual’s accumulated stock of information is
readily available for conscious, purposive processing and deci-
sion-making. Furthermore, the encoding and retrieval process is
subject to systematic inuences, both internal and external: a)
15. Alternative ly, it could arise fro m an affective or signaling val ue of
self-este em.
16. By den ition, “forgetting” means that an actual loss of information (a
coarsening of the informati onal partition) occurs. Thus, if at date 1 the individual
does not remem ber a date 0 signal s, he also does not recall any other piece of
informatio n that is pe rfectly co rrelated with s, such as the costs M(l) i ncurred in
the process of forgetting. T his comple te forgetting is o nly a simplied represen-
tation of a richer attribution (signal e xtraction) problem, however. Let dmeasure
the level of an actio n that affects recall but can also be undertaken for its own
sake: amount of wine consumed, time spent with friendly rather than critical
people, attention paid to the details of competing in formation, effort in making,
safekeepin g, or disposing of physical records, spatial or mental detours around
certain potential cues, etc. Choosing dfollowing an event sleads to a recall
probabili ty l 5 L(d) and has a direct utility u
0
(d,e), where eis a rando m taste
shock. Later on, the age nt may recall the action d* (s,e)[arg max
d
{u
0
(d,e)1
bdE
0
[u
1
1 du
2
ud,s]} that he took, and possibly the associated consequences
u
0
(d*(s,e),e) , bu t not the parti cular reali zation of ethat occurred (e.g., how much
did I really want to drink wine, or sit next to that person at dinner?). Recalling
d*(s,e) is thus generally insufcient to fully reconstruct s, o r to separate out
within rea lized utility the cost M(L(d* (s,e) ,e)) [max
d
u
0
(d,e)2u
0
(d*(s,e),e)
that was incurred purely for memory manipulation. This problem remains when
the functions u
0
,Lor M, or the distribution of e, depend on s.
886 QUARTERLY JOURNAL OF ECONOMICS
information that is rehearsed often is better remembered (indeed,
that is why we cram for an exam); conversely, if one is preoccu-
pied or distracted when an event unfolds, one has greater dif-
culty remembering the details; b) direct behavioral experience
makes the information more accessible in memory, because later
on recall is more likely to be activated by situational cues.
17
Such mechanisms seem to be at work in experiments where
subjects who are asked to behave in a self-deprecating manner
later report lower self-esteem than earlier, while persons who are
asked to display self-enhancing behavior report higher self-es-
teem [Jones et al. 1981]. This may be due to the fact that they
were led to rehearse unfavorable or favorable information about
themselves, thus increasing the probability of remembering it
later on. Similarly, receiving positive feedback seems to trigger a
cue-based “warm glow” effect, which automatically makes acces-
sible to the individual other instances of himself in positive situa-
tions [Greenwald 1980].
These frictions in the mechanics of memory give the individ-
ual some discretion about what data he is more likely to con-
sciously recall later on—thereby opening the door to m otivated
cognition. Thus, a person who wishes to remember good news and
forget the bad can linger over praise or positive feedback, re-
hearse it periodically, and choose to be more frequently in envi-
ronments or with people who will remind him of his past suc-
cesses.
18
Conversely, he can eschew situations that remind him of
bad news, tear up the picture of a former lover, or, like the
narrator in Proust’s novel, avoid passing by a chest of drawers
which contains cues to painful memories. He can work unusually
hard to “forget” (really, not think about) a failed relationship or
family problem, or even use drugs and alcohol.
The individual can also nd ways to discount self-threaten-
ing news in the rst place. A common such strategy is to seek out
information that derogates the informativeness of the initial data
[Frey 1981; Gilovich 1991]. After being criticized in a seminar or
17. For evidence and dis cussions see, e.g., Schacter [1996] and Fazio and
Zanna [1981]. Mullaina than [2002] and Laibson [2001] provide models of cue-
dependent co nsumption.
18. Thus, “we can expect [an autho r in a meeting] to spend more time
considering the comments of the lone dissenter who praised the project (and
conrmed his self-conception) than of the colleagues who disliked it, thus merci-
fully softening the ca valcade of criticism s” [Gilbert and Cooper 1985]. For a
discussion of self-presentation strategies and the ir link with self-enhancement,
see Rhodewalt [1986].
887SELF-CONFIDENCE AND PERSONAL MOTI VATION
referee report, a researcher will look hard for reasons why the
commenter has poor taste, a limited understanding of the issues,
a vested interest in a competing theory or body of empirical
evidence, and so forth. Interpersonal strategies can also be called
upon; thus, a verbal ght with one’s spouse or someone who
criticizes one’s work may (consciously or not) serve the purpose of
creating a distraction that will impair accurate recollection of the
details of the criticism (of course, it has costs as well . . . ).
As these examples make clear, it is important to note that we
need not literally assume that the individual can directly and
mechanically suppress memories. Our model is equally consistent
with a Freudian view where memories get buried in the uncon-
scious (with some probability of reappearance), and with more
recent cognitive psychology which holds that memory itself can-
not be controlled, but emphasizes the different ways in which
awareness can be affected: the choice of attention when the infor-
mation accrues, the search for or avoidance of cues, the process of
selective rehearsal afterwards, and again the choice of attention
at the time the information is (voluntarily or accidentally) re-
trieved.
19
We shall therefore use the terms “memory” and “aware-
ness of past information” interchangeably.
ASSUMPTION 2 (METACOGNITION ). While the individual can manipu-
late his conscious self-knowledge, he is aware that incentives
exist that result in selective memory.
As illustrated by the opening quotations from Nietzsche and
Darwin, if a person has a systematic tendency to forget, distort, or
repress certain types of information, he will likely become (or be
made) aware of it, and not blindly take at face value the fact that
most of what comes to his mind when thinking about his past
performances and the feedback he received is good news. Instead,
using (some) rational inference, he will realize that what he may
have forgotten are not random events.
20
Formally, this introspec-
tion or skepticism with respect to the reliability of one’s own
self-knowledge is represented by Bayes’ rule, which implies that
a person cannot consistently fool himself in the same direction.
19. One could even adhere to a minimalist version of the model where the
individual can only improve his rate of recall (through rehea rsal, record-maki ng,
etc.), but never lower it ( M(l)5 ` for all l , l
N
). All that matters is the potential
for a differential rate of recall or a wareness in respo nse to desirable or undesira ble
informatio n.
20. As Gilbert and Cooper [1995] note, “we are all insightful naive psycholo-
gists, well aware of human tendencies to be self-serving.”
888 QUARTERLY JOURNAL OF ECONOMICS
Less sophisticated inference processes lead to similar results, so
long as they are not excessively naive (see subsection III.A).
II.B. The Game of Self-Deception
Let the agent receive, at date 0, a signal sabout his ability u.
To make things simple, let stake only two values: with proba-
bility 1 2qthe agent receives bad news, s 5 L; and with
probability qhe receives no news at all, s 5 A. In other words, “no
news is good news.” Let
(9) uL;E@uus5L#,E@uus5A#;uH.
Since sis informative about the return to date 1 effort, the
agent’s Self 1 would benet from having this signal. If it is
ego-threatening, however, Self 0 may have an interest in sup-
pressing it. The recollection at date 1 of the news swill be
denoted sˆ[{A,L}. We assume that memories can be lost but not
manufactured ex nihilo, so s 5 Aalways leads to s
ˆ5A. A signal
s 5 L, on the other hand, may be forgotten due to natural
memory decay or voluntary repression. Let ldenote the proba-
bility that bad news will be remembered accurately:
(10) l;Pr@sˆ5Lus5L#.
As explained earlier, the agent can increase or decrease this
probability with respect to its “natural” value l
N
#1; choosing a
recall probability linvolves a “memory cost” M(l). We shall now
analyze the equilibrium in several stages.
1. Inference problem of Self 1. Faced with a memory sˆ[
{L,A}, Self 1 must rst assess its credibility. Given that memories
cannot be invented, unfavorable ones are always credible. When
Self 1 does not recall any adverse signals, on the other hand, he
must ask himself whether there was indeed no bad news at date
0, or whether it may have been lost or censored. If Self 1 thinks
that bad news are recalled with probability l*, he uses Bayes’
rule to compute the reliability of a “no recollection” message as
(11) r*;Pr@s 5Ausˆ5A;l*#5
q
q1~12q!~12l*!.
His degree of self-condence is then
(12) u ~r*!;r*uH1~12r*!uL.
889SELF-CONFIDENCE AND PERSONAL MOTI VATION
2. Decisions and payoffs. We normalize the payoff in case of
success to V51, and assume that the cost of date 1 effort is
drawn from an interval [cI,c#], with probability distribution F(c)
and density w(c).0. We assume that c#. bdu
H
. bd u
L
.cI,
which means that at date 1 there is always a positive probability
of no effort, and a positive probability of effort.
21
Given a signal sat date 0 and a memory s
ˆ at date 1, Selves
0 and 1, respectively, assess the productivity of date 1 effort as
E[uus] and E[uus
ˆ ]. S elf 1 only wor ks when the r ealization o f the
effort cost is c, bdE[uus
ˆ ], so Self 0’s payoff is
(13) bd
E
0
bdE@uusˆ#
~dE@uus# 2c!dF~c!.
To build intuition, suppose for a moment that Self 0 could freely
and costlessly manipulate Self 1’s expectation, E[uus
ˆ ]. W hat be-
liefs would he choose for a naive Self 1? Maximizing (13), we nd
that Self 0 would like to set E[uus
ˆ ] equal to E[uus]/b. This makes
clear how time consistency gives Self 0 an incentive to boost or
maintain Self 1’s self-condence; the problem, of course, is that
Self 1 is not so easily fooled. These two effects are consistent with
the common view in psychology that a moderate amount of “posi-
tive illusion” about oneself is optimal, but that many people nd
it quite difcult to strike this desirable balance.
3. Costs and benets of selective memory or attention. Focus-
ing on the “bad news” case, denote as U
C
(u
L
ur*) the expected
utility of Self 0 (gross of memory-management costs) when the
adverse information is successfully forgotten, and as U
T
(u
L
) the
corresponding value when it is accurately recalled. The subscripts
Cand Tstand for “censored” and “truth,” respectively. Hiding
from Self 1 the signal s 5 Lraises his self-condence from u
L
to
u(r*), leading him to exert effort in the additional states of the
world where bd u
L
,c, bdu (r*). As with ex ante ignorance, this
has both costs and benets; thus if r* is high enough that
bu (r* ) . u
L
, the net gain or loss from self-deception is
21. In Section I we to ok the distribu tion of uto be continuous , and cwas xed.
In this section chas a continu ous distributi on, and ucan take o nly two value s. The
two formulations are actually isom orphic (even if the latter happens to be more
convenient h ere): all that really matters is the distribution of d uV/c.
890 QUARTERLY JOURNAL OF ECONOMICS
(14) UC~uLur*!2UT~u L!
5bd
S
E
bduL
duL
~duL2c!dF ~c!2
E
duL
bdu ~r*!
~c2duL!dF~c!
D
.
The rst integral is decreasing in b, becoming zero at b 5 1: it
represents the gain from condence-building, which alleviates
Self 1’s motivation problem. The second integral is increasing in
b: it reects the loss from overcondence, which causes Self 1 to
attempt the task in states of the world where even Self 0 would
prefer that he abstain. Note that these effects are now endoge-
nous. Thus, the overcondence cost in (14) is larger, the more
reliable Self 1 considers the memory process to be, i.e., the larger
r*. Conversely, if r* is so low as to have bu (r*) , u
L
, the
overcondence effect disappears entirely, but the condence-
building effect is now limited by u(r*).
4. Strategic memory or awareness management. Faced with
a signal s 5 Lthat is hurtful to his self-esteem , Self 0 chooses the
recall probability lso as to solve
(15) max
l
$lUT~uL!1~12l!UC~uLur*!2M~l!%.
Given the convexity of M(l), the optimum is uniquely determined
(given r*) by the rst-order condition, which involves comparing
the marginal benet from self-deception, U
C
(u
L
ur*) 2U
T
(u
L
),
with the marginal cost M9(l). Finally, the Bayesian rationality of
Self 1 means that he is aware of Self 0’s choosing the recall strategy
lopportunistically according to (15), and uses this optimal lin
his assessment of the reliability of memories (or lack thereof).
DEFINITION 2. A Perfect Bayesian Equilibrium (PBE) of the mem-
ory game is a pair (l*,r*) [[0,1] 3[q,1] that solves (11)
and (15), meaning that
i) The recall strategy of Self 0 is optimal, given Self 1’s
assessment of the reliability of memories.
ii) Self 1 assesses the reliability of memories using Bayes’
rule and Self 0’s recall strategy.
We shall be interested in two main issues.
1. Nature and multiplicity of equilibria. What modes of self-
esteem management are sustainable (from “systematic denial” to
891SELF-CONFIDENCE AND PERSONAL MOTI VATION
“complete self-acceptance”), depending on a person’s characteris-
tics such as his time-discounting prole or cost of memory ma-
nipulation? Can the same person, or otherwise similar people, be
“trapped” in different modes of cognition and behavior?
2. Welfare analysis. Is a more active self-esteem mainte-
nance strategy always benecial, or can it end up being self-
defeating? Would a person rather be free to manage his self-
condence and awareness as he sees t, or prefer a priori to nd
mechanisms (friends, mates, environments, occupations, etc.)
that ensure that he will always be confronted with the truth
about himself, no matter how unpleasant it turns out to be?
Because PBE’s are related to the solutions r*[[q, 1] to the
nonlinear equation obtained by substituting (11) into the rst-
order condition for (15), namely
(16) c~r,b! ;bd
E
bduL
bd~ruH1~12r!uL!
~duL2c!dF~c!
1M9
S
12q/r
12q
D
50,
we shall use a sequence of simpler cases to demonstrate the main
points that emerge from our model. Note, for further reference,
that c(r,b) represents Self 0’s (net) marginal incentive to forget.
II.C. Costless Memory or Awareness Management
Repression is autom atic forgetting [Henry Laughlin, The Ego and Its De-
fenses 1979].
We rst solve the model in the case where the manipulation
of memory is costless, M[0. While it does not capture the
psychological costs of repression (as opposed to the informational
ones), this case already yields several key insights, and is very
tractable.
22
PROPOSITION 2. When M[0, there exist b
#
and b
#in (0,1), b
#
, b
#,
with the following properties. For low degrees of time incon-
sistency, b . b
#, the unique equilibrium involves minimum
repression (l*51); for high degrees, b , b
#
, it involves
22. While we assume here that lcan be freely varied betwee n 0 and 1, the
results woul d be identical if it were con strained to lie in some interval [l
#,l
#]. With
l
#.0 one can never forget (or avoid undesired cues) for sure.
892 QUARTERLY JOURNAL OF ECONOMICS
maximum repression (l*50). For intermediate degrees of
time inconsistency, b[[b
#
,b
#], there are three equilibria,
including a partially repressive one: l*[{0,L(b),1}, where
L(b) decreases from 1 to 0 as brises from b
#
to b
#.
These results are illustrated on Figure I. The intuition is
simple, and apparent from (14). When bis high enough, overcon-
dence is the dominant concern; therefore, adverse signals are
systematically transmitted. Conversely, for low values of bthe
condence-building motive dominates, so ego-threatening signals
are systematically forgotten. For intermediate values both effects
are relevant, allowing multiple equilibria, including one where
memory is partially selective. What makes all three equilibria
self-fullling is precisely the introspection or “metacognition” of
the Bayesian individual, who understands that his self-knowl-
edge is subject to opportunistic distortions. The more censoring
by Self 0, the more Self 1 discounts the “no bad news to report”
recollection, and therefore the lower the risk that he will be
overcondent. As a result, the greater is Self 0’s incentive to
censor. Conversely, if Self 0 faithfully encodes all news into mem-
ory, Self 1 is more likely to be overcondent when he cannot recall
any bad signals, and this incites Self 0 to be truthful.
Note that in the censoring equilibrium (l*50), none of Self
0’s information is ever transmitted to Self 1: r*5q. In the
language of communication games, this is a “babbling equilib-
rium.” The mechanism at work here is nonetheless very different
from the ex ante suppression of information considered earlier
when analyzing self-handicapping, or in Carrillo and Mariotti
[2000]. Self 0 does not want to suppress good news, only bad
FIGURE I
893SELF-CONFIDENCE AND PERSONAL MOTI VATION
news; but in doing the latter, he cannot help but also do the
former. As we shall see later on, this may end up doing him more
harm than good, whereas the usual “strategic ignorance” is only
chosen when it improves ex ante welfare.
The last observation to be drawn from Figure I is that, as b
rises from 0 to 1, there is necessarily (i.e., for any equilibrium
selection) at least one point where l* has an upward discontinu-
ity. Small differences in the psychic or material costs of memory
management, repression, etc., can thus imply large changes in
the selectivity of memory, hence in the variability of self-con-
dence, and ultimately in performance.
II.D. Costly Memory or Awareness Management
To break down the renewed assaults of m y memory, my imagination effec-
tively labored in the opposite direction [Marcel Proust, R emembrance of
Things Past].
In this subsection we use specic functional forms to study
the problem set up in subsection II.B. The memory cost function
is
(17) M~l! 5a~12ln l! 1b~12ln~12l!!,
with a.0 and b$0. It is minimized at the “natural” recall rate
l
N
5a/( a1b), and precludes complete repression. When b.0,
perfect recall is also prohibitively costly, and Mis U-shaped. As to
the distribution of effort costs, we take it to be uniform, w(c)51/c#
on [0 ,c#], with c#. b du
H
.
With these assumptions the incentive to forget, namely
U
C
(u
L
ur*) 2U
T
(u
L
) in (16), is proportional to a third degree
polynomial in r, with either one or three roots in [q,1 ] (see the
Appendix). Therefore, for any (a,b) there are again either one or
three equilibria. One can go further, and obtain explicit compara-
tive statics results, by focusing on the simpler case where recall is
costless but repression is costly. The following proposition is illus-
trated in Figures IIa to IIb.
PROPOSITION 3. Let b50. A higher degree of time inconsistency or
a lower cost of repression increases the scope for memory
manipulation, generating partially repressive equilibria and
possibly even making perfect recall unsustainable. Formally,
1) For any given bthere exist thresholds aIand a#with 0 #aI#
a#, and continuous functions l
1
(a),l
2
(a), respectively, increas-
ing and decreasing in a, such that (i) for a[(0,aI), the unique
894 QUARTERLY JOURNAL OF ECONOMICS
equilibrium corresponds to l*5 l
1
(a); (ii) for a[(aI,a#), there
are three equilibria: l*[{l
1
(a),l
2
(a),1}; (iii) for a[(a#,1`),
the unique equilibrium corresponds to l*51.
2) There exist critical values b
1
, b
2
, b
3
such that (i) for b$
b
3
,a#50; (ii) for b[[b
2
,b
3
), aI50,a#, as in Figure IIa; (iii)
for b[(b
1
,b
2
), 0 ,aI,a#, as in Figure IIb; (iv) for b#b
1
, 0 ,
aI5a#, as in Figure IIc.
The most representative case is that of Figure IIb, where
each of the three ranges [0,aI], [aI,a#], and [ a#,1 `), correspond-
ing, respectively, to high repression, multiplicity, and truthful-
ness, is nonempty. The effects of aare intuitive; we just note that
FIGURE IIa
Case b
2
, b , b
3
FIGURE IIb
Case b
1
, b , b
2
895SELF-CONFIDENCE AND PERSONAL MOTI VATION
small changes in awareness costs can induce large changes in
self-esteem and behavior.
23
Interestingly, a lower willpower b, by
shifting the equilibrium set toward lower l’s, tends to make the
individual incur higher repression costs.
III. BELIEFS AND MAKE-BELIEFS
24
As discussed earlier, surveys, experiments, and daily obser-
vation consistently suggest that most people overestimate their
past achievements, abilities, and other desirable traits, both in
absolute terms and relative to others (e.g., Weinstein [1980] and
Taylor and Brown [1988]). Well-educated, reective individuals
seem to be no exception: as Gilovich [1991] relates, “a survey of
college professors found that 94% thought they were better than
their average colleague.” These ndings are often put forward as
evidence of pervasive irrationality in human inference.
25
It turns
23. It is also interesting to note that the specication with uniformly distrib-
uted costs is formally equivalent to one where cis xed (say, c[1) but e ffort is
a continuous decision, with net discounted payoff bdue2e
2
/ 2 for Self 1 and
bd(due2e
2
/ 2) for Self 0. Thus, in our model, discon tinuities in behavior are n ot
predicated on an indivisibility.
24. The terminology of “beliefs and make-beliefs” is borrowed from Ainslie
[2001].
25. Without denying the validity of thi s kind of evidence, we would like to
emphasize th at it should be interpreted with caution. Answers to surveys or
experimenta l questionnaires may reect self-presentation motives (for the bene-
ts of the interviewer), or selective memory rehearsal strategies (for the individ-
ual’s own benet, as predicted by our model). Second, for every person who is
“overcondent” about how great they a re (profe ssionall y, intellectua lly, socially,
FIGURE IIc
Case b , b
1
896 QUARTERLY JOURNAL OF ECONOMICS
out, however, that rational self-deception by Bayesian agents can
help account for most people holding biased, self-serving beliefs,
which in turn have aggregate effects.
III.A. Optimistic and Pessimistic Biases
Continuing to work with our awareness-management model,
let us compare the cross-sectional distributions of true and self-
perceived abilities.
26
In a large population, a proportion 1 2qof
individuals are of low ability u
L
, having received a negative
signal, s 5 L. The remaining q, having received s 5 A, have high
ability u
H
. Average ability is qu
H
1(1 2q)u
L
5 u (q); we assume
that q,1/ 2 , so that m edian ability is u
L
. Consider now the
distribution of self-evaluations. Suppose for simplicity that, when
faced with ego-threatening information (s 5 L), everyone uses
the same censoring probability l*[(0,1).
27
As before, let r*
denote the corresponding reliability of memory, given by (11).
Thus, when individuals make decisions at date 1,
— a fraction (1 2q)(1 2 l *) overestimate their ability by
u(r*) 2 u
L
5r*(u
H
2 u
L
);
— a fraction q underestimate it by u
H
2 u (r*) 5(1 2
r*)(u
H
2 u
L
).
If the costs of repression or forgetting are low enough (e.g., a
small ain Figure IIc), one can easily have ( 1 2q)(1 2 l*) .
1
2
,
and even (1 2q)(1 2 l *) ’
,1. Thus, most people believe
themselves to be more able than they actually are, more able than
average, and more able than the majority of individuals.
28
Adding
those who had truly received good news (s 5 A), the fraction of
the population who think they are better than average is even
larger, namely 1 2 l*(1 2q). The remaining minority think,
maritally), another o ne may be found who i s undercondent, depressed, paralyzed
by guilt and self-do ubt, but unlikely to acknowledge this to anyone except his
closest condant, counselor, or therapist. These could even b e the sam e people at
different poi nts in time.
26. The interesting distinction is betwe en self-perceived ability E[uusˆ] and
objectively assessed abi lity E[uus], rather than between E[uus] and the individ-
ual’s true uwhich it may m easure only i mperfectly. To simplify the expositi on, we
shall there fore assume in this section that sis perfectly informative about u; i.e.,
u 5 E[uus][{u
L
,u
H
}. Alternatively, one could just read “objectively assess ed
ability” wherever “ability” appears.
27. Either this is the uniq ue equilibriu m, as in Figure IIc, or else we focus on
a symmetric situation for s implicity.
28. Note that these statements (like most experimental data) are about the
agent’s perception of his rank in the distribution of true abilities—not in the
distribution of self-assessments which, as a Bayesian, he realizes are generally
overoptimis tic.
897SELF-CONFIDENCE AND PERSONAL MOTI VATION
correctly, that they are worse than average; as a result they have
low motivation and are unlikely to undertake challenging tasks.
They t the experimental ndings of depressed people as “sadder
but wiser” realists, compared with their nondepressed counter-
parts who are much more likely to exhibit self-serving delusions
[Alloy and Abrahamson 1979].
As seen above, Bayes’ law does not constrain the skewness in
the distribution of biases:
29
it only requires that the average bias
across the (1 2q)(1 2 l *) optimists and the qpessimists be zero:
indeed, (1 2q)(1 2 l *)r*2q(1 2r*) 50 by (11). In other
words, Bayesian rationality only imposes a trade-off between the
relative proportions of overcondent versus undercondent
agents in the population, and their respective degrees of over- or
undercondence. Note, however, that a zero average bias in no
way precludes self-esteem maintenance strategies from having
aggregate economic effects. Clearly, in our model they do affect
aggregate effort, output, and welfare, as none of these is a linear
function of perceived ability.
III.B. To Bayes or Not to Bayes?
Having shown that even rational agents can deceive them-
selves most of the time (albeit not all the time), we nonetheless
recognize that it may be more realistic to view people as imperfect
Bayesians who do not fully internalize the fact that their recol-
lections may be self-serving. At the other extreme, taking beliefs
as completely naive would be even more implausible. As argued
in subsection II.A (also recall Nietzsche and Darwin), if a person
consistently destroys, represses, or manages not to think about
negative news, he will likely become aware that he has this
systematic tendency, and realize that the absence of adverse
evidence or recollections should not be taken at face value. This
introspection is the fundamental trait of the human mind which
the Bayesian assumption captures in our model. Without it, self-
delusion would be very easy and, when practiced, always optimal
(ex ante). With even some of this metacognition, self-deception
becomes a much more subtle and complex endeavor.
In Be´nabou and Tirole [1999] we relax Bayesian rationality
and allow the agent at date 1 to remain unaware not just of what
29. This was rst pointed out by Carrillo and Mariotti [200 0] for strate gic
ignorance, and is a feature that our model also shares with those of Brocas and
Carrillo [1999] and Ko¨szegi [1999].
898 QUARTERLY JOURNAL OF ECONOMICS
he may have forgotten, but also of the fact that he forgets. Self 1’s
assessment of the reliability of a recollection s
ˆ5Ais thus
modied to
(18) rp~l! ;Pr@s 5Ausˆ5A;l# 5
q
q1p~12q!~12l! ,
where lis the actual recall strategy and p[[0,1] parameterizes
cognitive sophistication, ranging from complete naivete (r
0
(l)[
1) to full rationality (r
1
(l)[r*(l)). As long as pis above a
critical threshold, meaning that the individual’s self-conception is
not too unresponsive to his actual pattern of behavior, all the
Bayesian model’s results on multiplicity and welfare rankings of
personal equilibria go through. Thus, for the purpose of under-
standing self-deception and overoptimism , the explanatory power
gained by departing from rational inference is rather limited
(perception biases need no longer sum to zero), whereas much can
be lost if the departure is too drastic: without sufcient introspec-
tion, one cannot account for “self-traps” or self-doubt.
IV. WELFARE ANALYSIS OF SELF-DECEPTION
The art of being wi se is the art of knowing what to overlook [William James,
Principles of Psychology 1890].
There is nothing worse than self-deception—when the deceiver i s at hom e
and always with you [Plato, quoted by Mele 1997].
Is a person ultimately better off in an equilibrium with a
strategy of active self-esteem maintenance and “positive think-
ing” (l*,1), or when he always faces the truth? Like Plato and
William James, psychologists are divided between these two con-
icting views of self-deception. On one side are those who endorse
and actively promote the self-efcacy/self-esteem movement (e.g.,
Bandura [1977] and Seligman [1990]), pointing to studies which
tend to show that a moderate dose of “positive illusions” has
signicant affective and functional benets.
30
On the other side
are skeptics and outright critics (e.g., Baumeister [1998] and
Swann [1996]), who see instead a lack of convincing evidence, and
30. There is of course a huge industry based on that premise, with countl ess
web sites devoted to “self-esteem,” and hundreds of books with titl es such as How
to Raise Your Self-Esteem, 31 Days to High Self-Esteem, How to Change Your Life
So You Have Joy, Bliss & Abundance, 365 Ways to Build Your Child’s Self-Esteem,
501 Ways to Boost Your Child’s Self-Esteem, 611 Ways to Bo ost Your Self Esteem:
Accept Your Love Handles an d Everything A bout Yourself, ABC I Like Me, etc.
899SELF-CONFIDENCE AND PERSONAL MOTI VATION
point to the dangers of overcondence as well as the loss of
standards that results when negative feedback is systematically
withheld in the name of self-esteem preservation. Our model will
provide insights into the reasons for this ambiguity.
Consider an equilibrium with recall probability l*#1 and
associated credibility r* (via (11)). With probability 1 2q, Self 0
receives bad news, which he then forgets with probability 1 2 l*;
the resulting expected payoff is l*U
T
(u
L
)1(1 2 l*)U
C
(u
L
ur*) 2
M(l*). With probability qthe news is good, which means that no
adverse signal is received. The problem is that the credibility of a
“no bad news” memory in the eyes of Self 1 may be quite low, so
that he will not exert much effort even when it is actually optimal
to do so. Indeed, the payoff to Self 0 following genuinely “good
news” is only
(19) UT~uHur*!5bd
E
0
bdu ~r*!
~duH2c!dF~c!,
which is clearly less than U
T
(u
H
u1) whenever cost realizations
between u(r* ) and u(1) have positive probability. In that case
there is a loss from self-distrust or self-doubt, compared with a
situation where Self 0 always truthfully records all events into
memory. Like a ruler whose entourage dares not bring him bad
news, or a child whose parents praise him indiscriminately, an
individual with some understanding of the self-serving tendency
in his attention or memory can never be sure that he really “did
great,” even in instances where this was actually true.
Averaging over good and bad news, the agent’s ex ante wel-
fare in equilibrium equals
(20) W~l *,r*!;qUT~uHur*!
1~12q!@l*UT~uL!1~12l*!UC~uLur*!2M~l *!# .
Let us now assume that truth (perfect recall) is also an equilib-
rium strategy, with cost M(1 ); as we shall see, a very similar
analysis applies if l*51 is achieved by using some a priori
commitment mechanism (chosen before sis observed). Denoting
the difference in welfare with this benchmark case as
DW(l*,r*) [W(l* ,r*) 2W(1, 1), we have
DW~l*, r*!5~12q!@12l*!~UC~uLur*!2UT~uL!!
2M~l *!1M~1#2q@UT~uHu1!2UT~uHur*!#,
900 QUARTERLY JOURNAL OF ECONOMICS
or, equivalently,
(21) DW~l*, r*!5~12q!
S
~12l*!
E
bduL
bdu ~r*!
~duL2c!dF~c!
2M~l*!1M~1!
D
2q
E
bdu ~r*!
bduH
~duH2c!dF~c!.
The rst expression describes the net gain from forgetting bad
news; the second one the loss from disbelieving good news. While
the individual is better motivated or even overmotivated follow-
ing a negative signal about his ability, he may actually be under-
motivated following a good signal. A few general results can be
immediately observed.
First, if memory manipulation is costless (M[0), then
a partial recall (mixed strategy) equilibrium can never be
better than perfect recall. Indeed, in such an equilibrium the
gain from hiding bad news is zero ( U
C
(u
L
ur*) 5U
T
(u
L
)) be-
cause the self-enhancement and overcondence effects just cancel
out. The cost from self-distrust, on the other hand, is always
present.
When repression is costly, this reasoning no longer applies,
as the term in large brackets in (21) becomes M(1) 2M(l*) 2
(1 2 l*) M9(l* ) .0, by the convexity of M. Similarly, when
systematic denial (l*50) is an equilibrium, it generates a posi-
tive “surplus” in the event of bad news: U
C
(u
L
uq)2U
T
(u
L
).
M9(0 ) $0. How does this gain compare with the loss from
self-distrust in the good-news state? As seen from (21), the key
intuition involves the likelihood of cost realizations sufciently
high to discourage effort in the absence of adverse recollections
(s
ˆ5A). When such events are infrequent, the self-distrust effect
is small or even absent, and on average, self-deception pays off.
When they are relatively common, the reverse is true.
PROPOSITION 4. Let M[0. If the cost density w(c) decreases fast
enough,
2]ln w~c!
]ln c.22b
12bfor all c[@0,c##,
then ex ante welfare is higher if all bad news is censored from
memory than if it is always recalled. If the inequality is
901SELF-CONFIDENCE AND PERSONAL MOTI VATION
reversed, so is the welfare ranking. For a given cost distri-
bution w, self-deception is thus more likely to be benecial for
a less time-consistent individual.
Note that even the second result was far from obvious a
priori, since both the gain and the loss in (21) decrease with b: in
equilibrium, memory manipulation tends to alleviate procrasti-
nation when s 5 L, but worsen it when s 5 A.
31
To summarize,
we have shown that
1) When the tasks one faces are very difcult and one’s
willpower is not that strong, a strategy of active self-
esteem maintenance, “looking on the bright side,” avoid-
ing “negative” thoughts and people, etc., as advocated in
numerous “self-help” books, can indeed pay off.
2) When the typical task is likely to be only moderately
challenging, and time inconsistency is relatively mild, one
can only lose by playing such games with oneself, and it
would be better to always “be honest with yourself ” and
“accept who you are.”
It is important to note that in the second case, the individual
may still play such denial games, even though self-honesty would
be better. First, he could be trapped in an inferior equilibrium.
Second, m otivated cognition may be the only equilibrium, yet still
result in lower welfare than if the individual could commit to
never try to fool himself.
32
A couple of examples will help make
these results more concrete.
a) With a uniform density on [0,c#](c#. bdu
H
), self-deception
is always harmful compared with truth-telling. This ap-
plies whether both l*50 and l*51 are in the equilib-
rium set, or only one of them. (See Proposition 2 and
Figure I.) It also applies, a fortiori, when repression is
costly.
b) Conversely, self-deception is always benecial when
w(c)5 gc
2n
on [cI,1`), with 0 ,cI, bdu
L
,gchosen so
31. The intuition is relatively simple, however. The net loss across states
from a “hear no evil-see no evil” strategy l*50, namely 2DW(q,0), is simply the
ex ante value of information (always recallin g the true s, rathe r than having only
the uninformed prior u(0) 5q) . Only when time inconsistency is stro ng enough
can this value can be negative.
32. This case is also interesting because it involves two degrees of lack of
commitment: it is because the agent cannot commit to working at date 1 that his
inability to commit not to tamper with memory at date 0 becomes an issue, which
may end up hurting him more than if he had sim ply resigne d hims elf to the
original time-consistency problem.
902 QUARTERLY JOURNAL OF ECONOMICS
that the density sums to one, and n.(2 2 b)/ (1 2 b ). In
this case it can also be shown that l*50 is the only
equilibrium if M[0 .
c) Finally, we provide in the Appendix a simple example
where both l*50 and l*51 coexist as equilibria, and
where either one— depen ding o n parameter values—may
lead to higher ex ante welfare.
We have thus far interpreted the “always face the truth”
strategy as an equilibrium, sustainable alongside with l*. Alter-
natively, it could result from some initial commitment of the type
discussed earlier (chosen before sis observed), which amounts to
making oneself face steeper costs of self-deception (increasing
M(l) for l , l
N
).
33
This reinterpretation requires minor modi-
cations to (21), but the main conclusions remain unaltered.
34
The potential multiplicity of equilibria in our model raises
the issue of coordination among the individual’s temporal selves.
Observe that Self 1 always values information about the produc-
tivity of his own efforts, and therefore always ranks equilibria (or
commitment outcomes) in order of decreasing l’s. When Self 0
also prefers the l*51 solution, it is plausible (we are agnostic on
this point) that the individual will nd ways to coordinate on this
Pareto superior outcome. When some repression (any l*,1) is ex
ante valuable, however, there is no longer any clearly natural
selection rule. In either case, our main welfare conclusions re-
main unchanged even if one assumes that Self 0 always manages
to select his preferred equilibrium. First, for some range of bor a,
l*51 ceases to be an equilibrium even though its still maxi-
mizes ex ante welfare. Thus, once again, the individual is trapped
in a harmful pattern of systematic denial. Conversely, for rela-
tively high values of athe only equilibrium may be l*51 (more
generally, a high l*), even though the individual would, ex ante,
be better off it he could manage to repress bad news more easily.
Interestingly, our multiplicity and welfare results provide a
role for parents, friends, therapists, and other benevolent outside
parties to help an individual escape the “self-traps” [Swann 1996]
33. For instance, an indivi dual with b , b
#in Proposition 2 may be worse off
when memory ma nagement i s free (M50) tha n when it is prohibitively costly
(M(l)5 1` for a ll l Þ 1). For instance, specication (a) above shows that such
is always the case when the cost distribution w(c) is uniform.
34. Term M(1) in (20)–(21) is s imply replaced by b
21
d
2t
M
#/(1 2q)1m(1),
where M
#is the up-front cost of the co mmitment mechanism, 2t , 0 is the period
when the comm itment was made, and m(1) $0 is the cost of perfect recall faced
at t50 as a result of this decision (whereas m(l)5 1` for all l , 1).
903SELF-CONFIDENCE AND PERSONAL MOTI VATION
in which he might be stuck: depressive state of low self-esteem,
chronic blindness to his own failings, etc. They can make him
aware that a better personal equilibrium is feasible, and teach
him how to coordinate on it by following certain simple cognitive
rules. They may also offer a form of informational commitment,
serving as the repositories of facts and feelings which the indi-
vidual realizes that he has an incentive to forget (“let’s talk about
that incident with your mother again”). More generally, they
allow him to alter the “awareness/repression” technology M(l)
(and hence the set of feasible equilibria), whether through their
own feedback and questioning, or by teaching him certain cue-
management techniques. Indeed, much of modern cognitive ther-
apy aims at changing people’s self-image through selective recol-
lection and rehearsal of events, self-serving attributions about
success and adversity, or conversely helping them “see through”
harmful self-delusions.
V. VARIANTS AND EXTE NSIONS
V.A. Defensive Pessimism
While people are most often concerned with enhancing and
protecting their self-esteem, there are also many instances where
they seek to minimize their achievements, or convince themselves
that the task at hand will be difcult rather than easy. A student
preparing for exams may thus discount his previous good grades
as attributable to luck or lack of difculty. A young researcher
may understate the value of his prior achievements, compared
with what will be required to obtain tenure. A dieting person who
lost a moderate amount of weight may decide that he “looks fatter
than ever,” no matter what others or the scale may say.
Such behavior, termed “defensive pessimism” by psycholo-
gists, can be captured with a very simple variant of our model.
The above are situations where the underlying motive for infor-
mation-manipulation is still the same, namely to alleviate the
shirking incentives of future selves; the only difference is that
ability is now a substitute rather than a complement to effort in
generating future payoffs. This gives the agent an incentive to
discount, ignore, and otherwise repress signals of high ability, as
these would increase the temptation to “coast” or “slack off.”
Substitutability may arise directly in the perform ance “pro-
duction function” which, instead of the multiplicative form
904 QUARTERLY JOURNAL OF ECONOMICS
p(e,u)5 uethat we assumed, could be of the form p(e,u) with
p
eu
,0. More interestingly, it will typically occur when the
reward for performance is of a “pass-fail” nature: graduating from
school, making a sale, being hired or red (tenure, partnership),
proposing marriage, etc. To see this, let performance remain
multiplicative in ability and effort: p(u,e,e)5 eue, where eis a
random shock with cumulative distribution H(e). The payoff V,
however, is now conditional on performance exceeding a cutoff
level p#. Self 1’s utility function is thus
(22) bdVPr@eue$p##2ce 5bdV~12H~p#/ue!! 2ce.
It is easily veried that if the density h5H9is such that
xh9(x)/ h(x). 21 on the relevant range of x[ p#/ue, the optimal
effort is decreasing in u. Note that these results yield testable
predictions: by comparing subjects’ condence-maintenance be-
havior across experiments (or careers) where payoffs are comple-
ments and substitutes, one should be able to distinguish between
the motivation-based theory of self-condence and the hedonic or
signaling alternatives.
An even simpler form of defensive pessimism arises in situa-
tions where the action subject to time inconsistency is such that
the benets precede the costs. One can think of the trade-off
between the immediate pleasure of smoking, drinking, spending
freely, etc., and the long-term, large but uncertain costs of such
behaviors. Suppose, for instance, that at date 1 the decision is to
consume or not consume. The rst option yields utility b, but with
probability ventails a cost Cat date 2; the second option yields
zero at both dates. Clearly, if we dene “effort” e[12xas
abstinence from consumption, c5bas its (opportunity) cost and
vCas its expected long-term payoff, we see that this problem ts
exactly with our model. Thus, to counteract his tendency toward
short-term gratication, the agent will try to maintain beliefs
that vand Care high.
35
Note that these variables are comple-
ments to e512xin his utility function. If we had framed the
problem in terms of uncertainty over the probability of being
immune to the health risks (say) from tobacco or alcohol, this
probability 1 2 v would be a substitute with e, so the agent would
like to understate it to his future selves. Whether costs precede or
35. This is the kind of example to which Carrillo and M ariotti [2000] apply
their mode l of strategic ignorance, poi nting to studies that suggest that most
people actually overestimate the health risks from smoking. More generally, the
role o f the timing of costs a nd benets i s emphasized in Brocas an d Carrillo [2000 ].
905SELF-CONFIDENCE AND PERSONAL MOTI VATION
follow benets thus simply amounts to a relabeling of variables.
The only thing that matters for the direction in which the agent
would like to manipulate his beliefs concerning a variable is its
cross-derivative with the decision variable that is being set inef-
ciently low due to time inconsistency.
V.B. Self-Esteem as a Consumption Good
We have until now emphasized the value of self-condence
for personal motivation. This approach provides an explanation of
both why and how much people care about their self-image: its
value arises endogenously from fundamental preferences, tech-
nological constraints, and the structure of incentives. As ex-
plained earlier, the motivation theory also readily extends to
social interactions.
This functional view of self-esteem, while pervasive in psy-
chology, is by no means the only one (e.g., Baumeister [1998]). As
discussed earlier, a common and complementary view involves
purely affective concerns: people just like to think of themselves
as good, able, generous, attractive, and conversely nd it painful
to contemplate their failures and shortcomings. Formally, self-
image is simply posited to be an argument of the utility function.
This potentially allows people to care about a broader set of
self-attributes than a purely motivation-based theory: they may,
for instance, want to perceive themselves as honest and compas-
sionate individuals, goo d citi zens, faithful spouses— or, on the
contrary, pride themselves on being ruthless businessmen, ultra-
rational economists, irresistible seducers, etc. There is somewhat
of an embarrassment of riches here, with few constraints on what
arguments should enter the utility function, and with what sign.
Let us therefore focus, as before, on the trait of “general
ability,” which presumably everyone views as a good. This is also
the type of attribute from which agents are assumed to derive
utility in Weinberg [1999] and Ko¨szegi [1999], as well as in some
interpretations which Akerlof and Dickens [1982] offered for their
model of dissonance reduction. The trade-off between the costs
and benets of information can then be modeled by positing
preferences of the form,
(23) E@max $u
ˆV2c,0%1u~u
ˆ!#,
where u
ˆdenotes the individual’s self-perceived ability (expected
906 QUARTERLY JOURNAL OF ECONOMICS
probability of success) at the time of the effort decision.
36
The rst
term always generates a demand for accurate information, to
improve decision-making. Suppose for now that the hedonic valua-
tion u(u
ˆ) is increasing and concave; these properties, respectively,
imply a positive demand for self-esteem, and risk aversion with
respect to self-relevant signals. The individual may then, once
again, avoid free information or engage in self-handicapping.
37
Similarly, all our results based on memory management on the
supply side carry over to this case. Thus, “positive thinking” and
similar self-deception strategies may be pursued even though
they are ultimately detrimental (recall the quotation from Plato),
while conversely personal rules not to tamper with the encoding
and recall of information, such as Darwin’s, can be valuable. The
basic insight is, again, one of externalities across information
states: having only good news is not such a great boost to self-
esteem once the agent realizes that he would have had reasons to
censor any bad news that might have been received.
Unfortunately, psychology provides little guidance on what
the appropriate shape of the hedonic preference function should
be (by contrast, there is ample evidence of people’s general bias
toward short-term rewards, tendency to procrastinate, etc.). It is
thus equally likely that u(u
ˆ) is convex, at least over some range;
in such cases the individual will be an avid information-seeker,
choosing tasks that are excessively hard or risky but very infor-
mative, as a way of gambling for (self-) resurrection. Even mono-
tonicity may not be taken for granted, since psychologists have
documented both optimism and defensive pessimism. The latter,
whether originating from motivation concerns or hedonic ones
(lowering one’s expectations of performance because surprise
36. A more general formula tion, encompass ing both affecti ve and instrumen-
tal concerns, would be E
0
[max {E
1
[uV2c],0} 1J(F
1
)], where F
1
(u) denotes the
agent’s date 0 and date 1 subjective probability distribution over his true ability.
The functional J[represents e ither an e xogenous hedonic utility, or an endoge-
nous value function capturing the instrumental value of beliefs for self-motivation
or s elf-presenta tion (signali ng) purposes. In our model J(F
1
) is easily computed,
and related to b.
37. In a different context, Rabin [19 95] makes beliefs about the negative
externalitie s of one’s actions (on other people, animals, or the environment) a n
argument of the utility function, and assumes concavity. This provides an expla-
nation for why people may prefer not to know of the potential harm caused by
their consumption choices. Caplin and Leahy [2001] study a general class of
preferences where initial perceptions of future lotterie s enter into the intertem-
poral util ity function. Depending on whethe r the dependence is concave or convex,
a person will choose to avoid information that woul d make the future lottery more
risky or, on the contrary, seek out information and situations that increase the
stakes.
907SELF-CONFIDENCE AND PERSONAL MOTI VATION
sharpens both the sweetness of success and the bitterness of
defeat), requires that u(u
ˆ) be sometimes decreasing.
VI. CONCLUSION
Building on a number of themes from cognitive and social
psychology, we proposed in this paper a general economic model
of why people value their self-image, and of how they seek to
enhance or preserve it through a variety of seemingly irrational
behaviors—from handicapping their own performance to practic-
ing self-deception through selective memory or awareness
management.
This general framework can be enriched in many directions.
On the motivation side, we noted earlier that anyone with a
vested interest in an individual’s success (or failure) has incen-
tives to manipulate the latter’s self-perception. Thus, in princi-
pal-agent relationships or bargaining situations, the manage-
ment of self-condence will matter even when everyone is fully
time-consistent. These issues are explored in Be´nabou and Tirole
[2001], where we examine the provision of incentives by informed
principals (parents, teachers, managers) in educational and
workplace environments. Because offering rewards for perfor-
mance may signal low trust in the abilities of the agent (child,
student, worker) or in his suitability to the task, such extrinsic
motivators may have only a limited impact on his current perfor-
mance, and undermine his intrinsic motivation for similar tasks
in the future—as stressed by psychologists.
Another interesting direction is to further explore the rich set
of behavioral implications that arise from the interaction of im-
perfect willpower and imperfect memory. Thus, in Be´nabou and
Tirole [2002] we develop a model of self-reputation over one’s
willpower that can account for the “personal rules” (diets, exer-
cise regimens, resolutions, moral or religious precepts, etc.)
through which people attempt to achieve self-discipline. The sus-
tainability of rule-based behavior is shown to depend on the
effectiveness of the individual’s self-monitoring (recalling past
lapses and their proper interpretation), which may be subject to
opportunistic distortions of memory or inference of the type stud-
ied in the present paper. The model also helps explain why people
may sometimes adopt excessively “legalistic” rules that result in
compulsive behavior such as miserliness, workaholism, or
anorexia.
908 QUARTERLY JOURNAL OF ECONOMICS
APPENDIX
Proof of Proposition 2. For all rand bin [0,1], let us dene
(A.1) x~r,b! ;
E
bduL
bdu ~r!
~duL2c!dF ~c!,
which, up to a factor of bd, measures the incentive to forget bad
news, U
C
(u
L
ur*) 2U
T
(u
L
).
LEMMA 1. For all r[[0 ,1], there exists a unique B(r)[[0,1]
such that x(r,B(r)) 50 and
i) x(r,b).0 for all b , B(r), while x(r,b),0 for all b .
B(r);
ii) B(r). u
L
/u(r), and B(r) is strictly decreasing in r.
Proof. For any given r, it is clear from (A.1) that x(r,b).0
for b[[0,u
L
/u(r)], while x(r,1) ,0. Moreover, for all b .
u
L
/u(r), we have
(A.2) ]x~r,b!
]b 5d2u ~ r!@uL2bu ~r! #w~ bd u ~r!!
2d2uL@uL2buL#w ~b duL!,0.
This establishes the existence and uniqueness of the root B(r)[
[u
L
/u(r), 1] . Moreover,
(A.3) ]x~r,b!
]r5bd2~uH2uL!@uL2bu ~r!#w~b du ~r!!,
so ]x (r,B(r))/]r,0 since B(r)u(r). u
L
. Therefore, by the
implicit function theorem, B9(r),0 for all r.i
To conclude the proof of Proposition 2, consider the following
cases.
a) For b$B(q) we have, for all r[[q,1], b . B(r) and
therefore x(r,b),0. Memorizing bad news is thus the
optimal strategy, which establishes claim (i) of the
Proposition.
b) For b#B(1) we have, for all r[[q,1], b , B(r) and
therefore x(r,b).0. Forgetting bad news is thus the
optimal strategy, which establishes claim (iii).
c) For b[(B(1),B(q)) there exists by the lemma a unique
inverse function R(b)[B
21
(b), such that x(R(b),b)50.
Moreover, Ris decreasing and for any r[(q,1), x(r,b)
909SELF-CONFIDENCE AND PERSONAL MOTI VATION
has the sign of R(b)2r. Therefore, the only equilibrium
with r,R(b) is r5q(or l 5 0), and the only equilibrium
with r.R(b) is r51 (or l 5 1). Finally, r5R(b) is also
an equilibrium, which corresponds to L(b)5(1 2
q/R(b))(1 2q). Dening b
#
[B(1) and b
#[B(q) con-
cludes the proof.
Proof of Proposition 3. We shall solve for equilibria in terms
of the reliability of memory, r*; the recall strategy l* is then
obtained by inverting (11). With the assumptions of the proposi-
tion, the incentive to forget, given by (16), equals
(A.4) c~ r,b! 5r~D u !
S
b2d3
c#
D S
~12b!uL2br
2~D u !
D
r
2ar
S
12q
r2q
D
1br
S
12q
q~12r!
D
,
where Du [ u
H
2 u
L
. Dening, for all b[[0,1],
(A.5) R~b! ;
S
12b
b
D
2uL
Du ,
(A.6) V~b! ;~D u !2
S
b3d3
2c#
D S
q
12q
D
.
It is clear that c(r,b)$0 if and only if
(A.7) P~r,b! ;V ~b !~R~b! 2r!~r2q!$aq 2b
S
r2q
12r
D
.
Multiplying by (1 2r) shows that the sign of c(r,b) is that of a
third-degree polynomial in r, with lim
r
N
q
c(r,b)5 2` since a.
0, and lim
r
N
1
c(r,b)5 1` when b.0. Thus, for a given bthere
are either one or three solutions to c(r,b)50 in [q,1], i.e., one or
three equilibria.
Let us now specialize (A.7) to the case where remembering is
costless but forgetting or repressing is costly, b50. Solving
c(r,b)50 then reduces to looking for the intersections of the
quadratic polynomial P(r,b) with the horizontal line aq. We shall
denote a#[q
21
max {max
r
[
[q,1 ]
P(r,b), 0} and aI[q
21
max
{P(1,b),0} #a#. There are several cases to consider.
1) For R(b),q, or equivalently b . R
21
(q)[ b
3
, it is clear
that P(r,b),0 on [q,1]; therefore, the only equilibrium
is r51. Moreover, aI5a#50.
910 QUARTERLY JOURNAL OF ECONOMICS
2) For q,R(b)[ b
3
the polynomial P(r,b) is positive on
r[[q,R(b)], implying a#.0, and negative outside.
a) If a.a#, then P(r,b),0 on [q,R(b)], so the only
equilibrium is again r51.
b) If a#a#, the equation P(r,b)5aq has two roots r
1
(a)
and r
2
(a), both in the interval [q,R(b)], with r
1
(a)#
r
2
(a), r
1
decreasing and r
2
increasing. One can associ-
ate with these two functions, l
1
(a) and l
2
(a), by invert-
ing (11). Let us now distinguish the following subcases.
i) For q,R(b),1, or equivalently b
2
[R
21
(1 ) ,
b , R
21
(q)5 b
3
, both r
1
(a) and r
2
(a) are in
(q,R(b)) and represent equilibria. On [q,r
1
(a)) and
(r
2
(a),1] we have P(r,b),aq , hence c(r,b),0.
This means that the third (and only other) equilib-
rium is r51. Furthermore, aI50.
ii) For 1 ,R(b),22q, or equivalently b
1
[
R
21
(2 2q), b , b
2
5R
21
(1), the polynomial
P(r,b) reaches its maximum at ( q1R(b)) / 2 ,1.
Thus, P(r,b) is positive and hill-shaped on [q,1],
and aI5P(1,b).0 . This implies that for aI,a,
a#we have q,r
1
(a),r
2
(a),1, while for a,aIwe
have q,r
1
(a),1,r
2
(a). In the rst case the
equilibria are r[{r
1
(a),r
2
(a),1}, as in case (i)
above. In the latter situation the only equilibrium is
r5r
1
(a).
iii) For 2 2q,R(b), or equivalently b , b
1
[
R
21
(2 2q), the polynomial P(r,b) is strictly in-
creasing on [q,1], so the only equilibrium is r5r
1
(a)
whenever a,aI5P( 1,b)5a#. It is r51 whenever
a$aI.
Proof of Proposition 4. Setting l*50, r*5qand M[0 in
(21) yields
DW~0,q!5~12q!
E
bduL
bdu ~q!
~duL2c!dF~c!
2q
E
bd u ~q!
bduH
~duH2c!dF~c!
911SELF-CONFIDENCE AND PERSONAL MOTI VATION
5q
E
0
bd u ~q!
~duH2c!dF~c!1~12q!
E
0
bd u ~q!
3 ~duL2c!dF~c!2q
E
0
bduH
~duH2c!dF~c!
2 ~12q!
E
0
bduL
~duL2c!dF~c!
5
E
0
bd u ~q!
@d~quH1~12q!uL!2c#dF~c!
2q
E
0
bduH
~duH2c!dF~c!
2 ~12q!
E
0
bduL
~duL2c!dF~c!.
Dening the function G(Z,b)[ *
0
Z
(Z2 bc)dF(c) , we can then
write
(A.8) DW~0,q!5b21@G ~bd ~quH1~12q!uL!,b!
2qG~bd uH,b! 2~12q!G~bdu L,b!#.
Clearly, DW(0,q).0 when Gis concave in Z, and DW(0 ,q),0
when it is convex. Indeed, b dDW(0,q) is (minus) the ex ante value
of information, i.e., of always knowing the true E[uus] rather than
have only the uninformed prior or posterior u(q). The proposition
immediately follows from the fact that ]
2
G(Z,b)/ ]Z
2
5(2 2
b)w(Z)1(1 2 b)Zw9(Z).
Welfare Rankings of Multiple Equilibria. We construct here
a simple example where l*50 and l*51 coexist as equilibria,
and where either one can lead to higher ex ante welfare.
First, let u
L
, u
H
and q[(0 ,1), so that u
L
, u (q)5qu
H
1
(1 2q)u
L
, u
H
. For b , 1 but not too small we have bu
L
, u
L
,
bu (q), u (q), bu
H
, u
H
. Next, let the date 1 cost take two
values: c[{cI,c#}, with cI/d[(bu
L
,u
L
), c#/d[(u(q), bu
H
) and p [
912 QUARTERLY JOURNAL OF ECONOMICS
Pr[c5cI][(0,1). The l*50 strategy is then always an
equilibrium, since c(q,b)/bd 5 p(du
L
2cI).0. As to l*51, it
is also an equilibrium whenever c(1 ,b)/ bd 5 p(d u
L
2cI)2(1 2
p)(c#2 d u
L
),0, or
(A.9) p
12p,
c#2duL
duL2cI;r.
With this condition both l*50 and l*51 are equilibria (with a
mixed-strategy one in between, which is always dominated by l*
51 since M[0), and l*50 yields higher welfare when
DW~0,b! 5~12q!p~duL2cI!2q~12p!~duH2c#!.0,
or
(A.10) p
12p.
S
q
12q
D S
duH2c#
duL2cI
D
;r9.
Since u(q), dc#, it is easily veried that r9 , r. Thus, for p/(1 2
p)[(r9,r), the l 5 0 equilibrium is ex ante superior to the one
with l 5 1. For p/(1 2 p), r9 the reverse is true.
PRINCETON UNIVERSITY, NATIONAL BUREAU OF ECON OMIC RESEARCH,AND CENTRE
FOR ECONOMIC POLICY RESEARCH
INSTITUT D’ECONOMIE INDUSTRIELLE, GREMAQ/CNRS, CERAS/CNRS, ECOLE DES
HAUTES ETUDES EN SCIENCES SOCIALES,AND MASSACHUSETTS INSTITUTE OF TECHNOLOGY
REFERENCES
Ainslie, G., Picoeconomics: The Strategic Interaction of Successive Motivational
States within the Person (Studies in Rationality and Social Change) (Cam-
bridge, UK: Cambridge University Press, 199 2).
——, Breakdown of Will (Cambridge, UK: Cambridge University Press, 2001).
Akerlof, G., and W. Dickens, “The Economic Consequences of Cognitive Disso -
nance,” American Econ omic Review, LXXII (1982 ), 307–319.
Alloy, L. T., and L. Abrahamson, “Judgement of Contingency in Depressed and
Nondepresse d Students: Sadder but Wiser?” Journal of Experimental Psy-
chology: General, CVIII (19 79), 441– 485.
Arkin, R. M., and A. H. Baumgardner, “Self-Handi capping,” in Attribution: Basic
Issues and Applications, J. Harvey and G. Weary, eds. (New York: Academic
Press, 1985).
Bandura, A., Self Efcacy: The Exercise of Control (New York: W. H. F reeman
Company, 197 7).
Baumeiste r, R., “The Self,” in The Handbook of Social Psychology, D. Gilbert, S.
Fiske, and G. Lindzey, eds. (Boston, MA: McGraw-Hill, 1998).
Be´ nabou, R., and J. Tirole, “Self-Condence: Intrapersonal Strategies,” I DEI
mimeo, J une 199 9.
Be´ nabou, R., and J. Tirole, “Intrinsic and Extrinsic Motivation,” Princeton Uni-
versity, mimeo, December 2001.
Be´ nabou, R., and J. Tirole, “Willpower and Personal Rules,” CEPR Discussion
Paper No. 3143, January 2002.
Berglas, S., and E. Jones, “Drug Choice as a Self-Handicapping Strategy in
913SELF-CONFIDENCE AND PERSONAL MOTI VATION
Response to Noncontingent Success,” Journal of Personality and Social Psy-
chology, XXX VI (1978), 405– 417.
Brocas, I., and J. Carrillo, “Entry Mistakes, Entrepreneuri al Boldness and Opti-
mism,” ULB-ECARE mimeo, June 1999.
Brocas, I., and J. Carrillo, “The Value of Information When Preferences Are
Dynamically I nconsisten t,” European Economic Review, XLIV (2000),
1104 –1115.
Caplin, A., and J. Leahy, “Psycho logical Expected Utility Theory and Anticipatory
Feelings,” Quarterly Journal of Economics, CXVI (2001), 55–79.
Carrillo, J., and T . Mariotti, “Strategic Ignorance as a Self-Disciplining Device,”
Review of Economic Studies, LXVI (20 00), 529 –544 .
Crary, W. G., “Reactions to Inco ngruent Self-Experiments ,” Journal of Consulting
Psychology, XX X (1966 ), 246 –252.
Darwin, F., The Life and Letters of Charles Darwin, Edited by his son Francis
Darwin (New York: D. Appleton and Co., 1898).
Deci, E., Intrinsic Motivation (New York: Plenum Press, 1975).
Elster, J., “Mo tivated Belief Formation,” C olumbia Universi ty, mimeo, J une 1999.
Fazio, R., and M. Za nna, “Direct Evidence and Atti tude-Behavior Consistency,” in
L. Berkowitz, ed., Advances in Experimental Social Psychology, Vol. 14 (New
York: Academic Press, 1981 ).
Festinger, L., “A Theory of Social Comparison Processes,” Human Relations, VII
(1954), 117–140.
Fingarette, H., “Alco holism and Self-Decep tion,” in Self-Deception and Self-Un-
derstanding, M. Martin, ed. (Lawrence, KS: University Press of Kansas, 1985).
Freud, S., A General Introduction to Psychoanalysis (New York: Garden City
Publishing C o., 1938).
Frey, D., “The Effect of Negative Feedback about Oneself and Cost of Information
on Preference for Info rmation abo ut the So urce of this Feedback,” Journal o f
Experimental So cial Psychology, XVII (1 981), 42–50 .
Gilbert, D., and J. Cooper, “Social Psychological Strategies of Se lf-Deception,” i n
Self-Deception and Self-Understanding, M. Martin, ed. (Lawrence, KS: Uni-
versity Press of Kansas, 1985).
Gilovich, T., How We Know Wh at Isn’t So (New York: Free Press, 199 1).
Greenier, K., M. Kernis, and S. Wasschul, “Not All High (or Low) Self-Este em
People Are the Same: Theory and Research on the Stability of Self-Esteem,”
in Efcacy, Agency and Self-Esteem, M. Kern is, ed. (New York: Plenum Press,
1995).
Greenwald, A., “The Totalitarian Ego: Fabrication and Revision of Personal His-
tory,” American Psychology, XXXV (1980), 603– 613.
Gur, R., and H . Sackeim, “Self-Deception: A Concept in Search of a Phen omenon,”
Journal of Personality and Social Psychology, XXXVII (1 979), 147–169.
Heider, F., The Psychology of Interpersonal Relations (N ew York: Wiley, 1958).
James, W ., The Principles o f Psychology (Cleveland, OH: World Pub lishing, 1890).
Jones, E., F. Rhodewalt, S. Berglas, and J. Skelton, “Effects o f Strategic Self-
Presentation on Subsequent Self-Esteem,” Jo urnal of Personality and Social
Psychology, XL (1981), 407– 421.
Kolditz, T., and R. Arkin, “An Impressio n-Managem ent Interpretati on of the
Self-Handica pping Strategy,” Journal of Person ality and Social Psychology,
XLIII (1982), 49 2–502.
Korner, I., Experimental Investigation of Some Aspects of the Problem of Repres-
sion: Repressive Fo rgetting, Contributions to Education, No. 970 (New York:
Bureau of Publications, Teachers’ Col lege, Columbia Uni versity, 1 950).
Ko¨ szegi, B., “Self-Image and Economic Behavior,” MIT, mimeo, October 1999.
Kunda, Z., and R. Sanitio so, “Motivated Changes in the Self-Concept,” Journal of
Personality and Social Psy chology, LXI (1989), 88 4– 897.
Laibson, D., “Golde n Eggs and Hyperbolic Discounting,” Qu arterly Journal of
Economics, CXII (1997), 443– 47 8.
——, “A Cue-Theory of Cons umption,” Quarterly Journal of Economics, CXVI
(2001), 81–11 9.
Laughlin, H . P., Th e Ego and Its Defenses, The Na tional Psychi atric Endowment
Fund, eds., second edition (New York: Jason Aaronson, Inc., 1979).
Leary, M., and D. Down, “Interpersonal Functions of the Self-Esteem Motive: The
914 QUARTERLY JOURNAL OF ECONOMICS
Self-Esteem Sys tem as Sociometer,” in Efcacy, Agency and Self-Esteem, M.
Kernis, ed. (New York: Plenum Press, 1995).
Loewenste in, G., and D. Prelec, “Ano malies in Intertemporal Choice: Evidence
and Interpretati on,” Quarterly Journal of Economics, CVII (1992), 573–597.
Mele, A., “Real Self-Deception,” Behavioral and Brain Sciences, XX (1999), 91–136.
Mischel,W., E. B. Ebbesen,and A. R. Zeiss,“Determinantsof Selective Memoryabout
the Self,” Journal of Consulting and Clinical Psychology, XLIV (1976), 92–103.
Mullaina than, S., “A Memo ry-Based Model of Bounded Rationality,” Quarterly
Journal of Econo mics, CXVII (2002), 735–774.
Murray, S. L ., and J. G. Holmes, “Seeing Virtues in Faults: Ne gativity and the
Transformatio n of Interpersonal Narratives in Clos e Relationship s,” Journal
of Personality and Social Psychology, XX (1993), 650 – 6 63.
Nisbett, R. , and T. Wi lson, “Telling More Than We Can Know: Verbal Reports on
Mental Proces ses,” Psychological Review, LXXXIV (1977), 231–259.
O’Donoghue, T., and M. Rabin, “Doing it Now or Later,” American Economic
Review, LXXXIX (1999), 103–124.
Phelps, E., and R. Pollack, “On Second-Best Natio nal Savings and Game-Equilib-
rium Growth,” Review of Economic Studies, XXXV (1968), 185–199.
Rabin, M ., “Moral Preferen ces, Moral Rules, a nd Belief Mani pulation,” University
of California, mimeo, April 1995.
Rhodewalt, F. T., “Self-Prese ntation and the Phenomenal Self: On the Stability
and Ma lleabili ty of Self-Conceptions,” in Public Self and Private Self, R.
Baumeiste r, ed. (New York: Springer Verlag, 1986).
Salancik, G., “Commi tment and the Control of Organizational Beh avior and
Belief,” in New Directions in Organizational Behavior, B. Staw and G. Salan-
cik, eds. (Chicago: St. Clair Press, 1977).
Sartre, J. P., The Existential Psychoanalysis (H. E. Barnes, trans.) (New York:
Philosoph ical Library, 19 53).
Schacter, D., Searching f or Memory (New York: Basic Books, 1996).
Seligman, E., L earned Optimism: How to Change Your Mind and Your Life (New
York: Simo n and Schuster, 1990).
Snyder, C., “Collaborative Companions: The Relationship of Self-Deception and
Excuse Making,” in Self-Deception and Self-Understanding, M. Martin , ed.
(Lawrence, KS: U niversity Press of Kansas, 19 85).
Strotz, R., “Myopia an d Inconsistency in Dynamic Utility Maximization,” Review
of Economic Studi es, XXII (1956), 165–180.
Swann, W. B., J r., Self Traps: The Elusive Quest for Higher Self-Esteem (New
York: W. H. Freeman and Co mpany, 1996).
Taylor, S. E., and J. D. Brown, “Illusion and Well-Being: A Social Psychological
Perspective on Mental Health,” Psychological Bulletin, CIII (19 88), 193–210.
Weinberg, B., “A Mod el of Overconde nce,” Ohio State Un iversity, mime o, August
1999.
Weinstei n, N., “Unrealistic Optim ism about Future Life Events,” Journal of
Personality and Psychology, XXXIX (19 80), 806 – 820.
Zuckerman, M ., “Attribution of Success and Failu re Revisi ted, or the Motiva tional
Bias Is Ali ve and Well in Attribution Theory,” Journal of Personality, XLVII
(1979), 245–287.
915SELF-CONFIDENCE AND PERSONAL MOTI VATION
