ArticlePDF Available

Focalism: A Source of Durability Bias in Affective Forecasting

Authors:

Abstract

The durability bias, the tendency to overpredict the duration of affective reactions to future events, may be due in part to focalism, whereby people focus too much on the event in question and not enough on the consequences of other future events. If so, asking people to think about other future activities should reduce the durability bias. In Studies 1-3, college football fans were less likely to overpredict how long the outcome of a football game would influence their happiness if they first thought about how much time they would spend on other future activities. Studies 4 and 5 ruled out alternative explanations and found evidence for a distraction interpretation, that people who think about future events moderate their forecasts because they believe that these events will reduce thinking about the focal event. The authors discuss the implications of focalism for other literatures, such as the planning fallacy.
ATTITUDES AND SOCIAL COGNITION
Focalism: A Source of Durability Bias in Affective Forecasting
Timothy D. Wilson, Thalia Wheatley, and Jonathan M. Meyers
University of Virginia
Daniel T. Gilbert
Harvard University
Danny Axsom
Virginia Polytechnic Institute and State University
The durability bias, the tendency to overpredict the duration of affective reactions to future events, may
be due in part to focalism, whereby people focus too much on the event in question and not enough on
the consequences of other future events. If
so,
asking people to think about other future activities should
reduce the durability bias. In Studies 1-3, college football fans were less likely to overpredict how long
the outcome of a football game would influence their happiness if they first thought about how much time
they would spend on other future activities. Studies 4 and 5 ruled out alternative explanations and found
evidence for a distraction interpretation, that people who think about future events moderate their
forecasts because they believe that these events will reduce thinking about the focal event. The authors
discuss the implications of focalism for other literatures, such as the planning fallacy.
The pleasures and pains, joys and sufferings, which people actually
experience, often fall short of what they had anticipated ... In antic-
ipating a coming event we have it alone in mind, and make no
provision for other occurrences. (Tatarkiewicz, 1962/1976)
If a genie popped out of a lamp and offered you three wishes,
would you attain lasting happiness? Most of us think that, like
Aladdin, we would become happier people. Perfect health, true
love,
and untold riches would be ours for the asking, and who
would not enjoy blessings such as these? To obtain lasting happi-
ness,
however, people have to know what to wish for. In the
present studies, we tested the hypothesis that people often think
about the future in ways that reduce the accuracy of their affective
forecasts.
Undoubtedly, people know a great deal about what will make
them happy. Most of us recognize that it would be better to ask the
genie for good health, true love, and lots of money than for severe
Timothy D. Wilson, Thalia Wheatley, and Jonathan M. Meyers, Depart-
ment of Psychology, University of Virginia; Daniel T. Gilbert, Department
of Psychology, Harvard University; Danny Axsom, Department of Psy-
chology, Virginia Polytechnic Institute and State University.
We gratefully acknowledge the support of research Grant RO1-
MH5607S from the National Institute of Mental Health. We thank Alesha
Pelter, Jeff Smith, and Reggie Tyree for their help in conducting the
research.
Correspondence concerning this article should be addressed to Tim-
othy D. Wilson, Department of Psychology, Gilmer Hall, University of
Virginia, Charlottesville, Virginia 22903. Electronic mail may be sent
to tdw@virginia.edu.
arthritis, a dysfunctional marriage, and the minimum wage. How-
ever, predictions about the affective consequences of future events
may not always be correct. Miswanting is the case in which people
do not like or dislike an event as much as they thought they would
(Gilbert & Wilson, 2000; Mitchell, Thompson, Peterson, & Cronk,
1997).
Gilbert and Wilson (2000) identified a number of sources of
miswanting. Sometimes, for example, an affective forecast is
based on a faulty understanding of exactly what the event will
entail. When people think about winning a million dollars, they
probably imagine spacious mansions, round-the-world trips, and a
cavalier attitude toward their children's college tuition. They might
not anticipate the difficulty of maintaining relationships with en-
vious friends, the hundreds of annoying phone calls from needy
people seeking handouts, and the late-night worries about taxes
and investments. The events that we imagine occurring are often
quite different from the events that actually occur (Griffin & Ross,
1991).
Even if people know exactly what will happen, however, they
can still make inaccurate forecasts about the affective conse-
quences of that event. This is particularly true when people think
about the duration of their affective reactions. They may know
exactly what winning a million dollars entails and may accurately
predict that they will be ecstatic when Ed McMahon arrives at their
doorstep and hands them a check with lots of zeros. They might
overestimate, however, the duration of this ecstasy. Gilbert and
Wilson (2000) argued that people often overestimate the duration
of their emotional reactions to future events. This durability bias is
important, because people typically wish for and work toward
Journal of Personality and Social Psychology, 2000, Vol. 78, No. 5, 821-836
Copyright 2000 by the American Psychological Association. Inc. 0O22-3514/0Oy$5.OO DOt: I0.1037/W022-3514.78.5.82I
821
822
WILSON, WHEATLEY, MEYERS, GILBERT, AND AXSOM
events that they believe will cause lasting happiness, not just a
moment's pleasure. If they overestimate how long their pleasure
will last, they might be working toward the wrong things.
Gilbert, Pinel, Wilson, Blumberg, and Wheatley (1998) found
evidence for the durability bias in six studies that examined the
accuracy of people's affective forecasts. In one study, assistant
professors predicted that their tenure decision would have an
impact on their happiness for several years, whereas former assis-
tant professors who had achieved tenure were no happier than
former assistant professors who had not. In another study, voters in
a gubernatorial election predicted that they would be significantly
happier a month after the election if their candidate won than if
their candidate lost. In fact, the supporters of the winning and
losing candidates were just as happy a month after the election as
they were before the election. Wilson, Meyers, and Gilbert (1999)
replicated this result in a study of the 1996 presidential election.
Democrats predicted that they would be substantially happier the
week after the election if President Clinton were victorious; in fact,
they were no happier following the election than they had been
before. Republicans predicted that they would be substantially less
happy if President Clinton were victorious; in fact, they were only
slightly less happy than they were before. The durability bias has
proved to be a robust phenomena, obtained in diverse samples of
people who made predictions about both short-term and long-term
events (Gilbert et al., 1998; Wilson et al., 1999).
One cause of the durability bias is immune neglect, which is the
failure to take into account how much one's psychological immune
system will ameliorate reactions to negative events. When some-
thing bad happens, people work hard to reconstrue the event in
ways that make it less painful. Because the psychological immune
system operates largely outside of awareness, people do not take it
into account when forecasting their future emotional reactions.
They overpredict the duration of their reactions to future negative
events because they do not appreciate the extent to which they will
transform the events psychologically in a way that blunts their
impact.
Gilbert et al. (1998) noted that immune neglect is not the only
cause of the durability bias. Indeed, because the psychological
immune system works to ameliorate negative affect but not posi-
tive affect (Taylor, 1991), immune neglect explains only mispre-
dictions about the duration of reactions to negative events. Con-
sistent with this prediction, Gilbert et al. (1998) found a stronger
durability bias in reaction to negative than to positive events. There
was,
however, a positive durability bias in some of their studies. In
subsequent research, we have found significant positive durability
biases, such as the Wilson et al. (1999) study in which Democrats
overpredicted how happy they would be after President Clinton's
1996 reelection. Clearly, an additional mechanism is needed to
explain these findings.
We suggest that there is also a problem of focalism, whereby
people focus too much on the occurrence in question (termed the
focal event) and fail to consider the consequences of other events
that are likely to occur.
1
People think about the focal event in a
vacuum without reminding themselves that their lives will not
occur in a vacuum but will be filled with many other events. As
noted by Tatarkiewicz (1962/1976, p. Ill) in the opening quote,
people "make no provision for other occurrences" when predicting
their happiness following a positive or negative event.
Other occurrences can mitigate the effects of a focal event in a
number of
ways,
such as by reducing how much people think about
the event, causing people to reframe the event, and by triggering
affective reactions that compete with or nullify the consequences
of the event. We will focus on the first of these possibilities,
namely that by failing to consider the occurrence of other future
events, people overestimate how much the focal event will occupy
their thoughts and influence their happiness. When imagining a
positive tenure decision, for example, assistant professors might
not think about other events that will compete for their attention,
such as the upcoming deadline for the chapter they have yet to
write, the dinner party they are hosting in a week, and the fact that
their car needs a new battery. By failing to consider other occur-
rences such as these, they will overestimate how much they will
think about their tenure decision.
Given that some aspects of people's lives are predictable—for
example, professors know what their teaching schedule will be the
following semester and that they will have to endure several boring
committee meetings—people are capable of taking some nonfocal
events into account when predicting their future happiness. Be-
cause many aspects of the future are unpredictable, it would be
unfair to chastise people for not taking into account events that
they cannot know will occur. How could people anticipate that
their car battery will die the week after their tenure decision? Our
point is that whatever happens after the event will compete for
people's attention, regardless of whether these events are unpre-
dictable (the demise of a car battery) or predictable (boring com-
mittee meetings, playing with our children, reading a good book,
or puttering around the vegetable garden). Research on subjective
well-being suggests that people's attention turns quickly to their
current concerns, reducing the impact of past events on their
happiness (Frederick & Loewenstein, 1999; Sun, Diener, & Fujita,
1996).
People do not have to be clairvoyant to appreciate this fact
when making affective forecasts.
The focalism hypothesis is related to other well-known in-
stances in which people give disproportionate weight to accessible
information (Higgins, 1996; Schwarz, 1990). For example, when
people explain why a given hypothesis might be true, they focus
too much on reasons supporting the hypothesis and too tittle on
reasons for alternative hypotheses. Similarly, when asked to imag-
ine a specific behavior, such as giving blood, people focus too
much on ways in which the behavior could occur and too little on
ways in which the behavior might not occur (for reviews see
Anderson, Krull, & Weiner, 1996; Koehler, 1991). Even when
asked to think about ways in which an event might not have
occurred, people tend to focus on a limited range of alternatives
that are easy to bring to mind, at the expense of alternatives that are
more difficult to imagine (Kahneman & Miller, 1986; Kahneman
& Tversky, 1982; Roese & Olson, 1997). People are often content
to focus on what comes to mind easily, without making the effort
to think about alternative explanations, scenarios, outcomes, or
beliefs (Gilbert, 1991).
Similarly, when people forecast their future happiness after an
emotional event, they focus too much on that event. People could,
in principle, go beyond what is accessible and think about the
1
Schkade and Kahneman (1998) have independently called this a fo-
cusing illusion.
FOCALISM: A SOURCE OF DURABILITY BIAS 823
many other things that will occupy their future lives. Consistent
with the research just mentioned, however, we hypothesize that
when making affective forecasts, people focus too much on the
focal event and too little on other events that will also transpire and
require their attention.
If so, then it should be possible to reduce the durability bias by
inducing people
to
think about the many other events that will
transpire
in the
future. That
is, if
focalism
is a
cause
of the
durability bias, then reducing focalism
(by
inducing people
to
think about nonfocal events) should reduce this bias.
If
people are
thinking not only about their tenure decision but also about what
their future teaching schedule will be like and how often they will
have to attend committee meetings, they should make more accu-
rate estimates
of
the extent
to
which their tenure decision will
influence their happiness. This hypothesis follows directly from
studies in other areas that have asked people to go beyond the most
accessible explanation or hypothesis that comes to mind, by think-
ing about alternative explanations and hypotheses (e.g., Hirt
&
Markman, 1995). Our studies followed the same logic, by asking
people
to
think about other events that would transpire
in the
future,
in
addition to the focal event.
Specifically, we asked people
to
predict their overall level
of
happiness after an emotional event, with the expectation that they
would overestimate how long that event would have an impact on
their happiness (the durability bias). Before making their predic-
tions,
some participants completed a prospective "diary" (ostensi-
bly as part
of
another study), in which they rated how much time
they would spend on
a
variety of everyday activities on
a
specific
future date. We hypothesized that people who completed the diary
would predict that they would think less about the focal event
in
the future
and
that
it
would have less impact
on
their future
happiness than would people who did not complete the diary.
In
studies 1-3 we tested these hypotheses with college football fans
who predicted how happy they would be after a win and a loss by
their college football team, whereas in Studies 4-5 we examined
people's predicted happiness after hypothetical national events
such as
a
space tragedy in which several astronauts were killed.
Study
1:
They Foresaw
a
Game
Method
Overview
College football fans at the University of Virginia (UVA) and Virginia
Polytechnic Institute and State University (Virginia Tech) predicted what
their level
of
overall happiness would
be
immediately after the UVA-
Virginia Tech football game and on each
of
the succeeding few days
if
their school lost and if their school won the game. They also predicted how
much they would think about the game. Prior to making these predictions,
some participants completed
a
prospective diary questionnaire, on which
they rated how much time they would spend
on a
variety
of
everyday
activities in the days after the football game. We hypothesized that people
in the diary condition, relative to people in
a
no diary control condition,
would predict that their happiness would
not be as
influenced
by the
outcome of the game and that they would think less about the game.
Participants
Participants were 36 students (19 women, 17 men) from UVA and
52
students from Virginia Tech (27 women, 25 men) who indicated that they
were football fans and cared about the outcome of their school's football
games (that is, they were above the median on the average
of
these two
measures, which were highly correlated,
r =
.71). The students participated
for partial fulfillment
of a
requirement
in an
undergraduate psychology
course.
Procedure
Participants completed
a
prediction questionnaire
1-2
months prior
to
the 1995 football game between UVA and Virginia Tech in small groups
or during class meetings. They were told that the packet contained ques-
tionnaires from different research projects and that they should go through
the packet one page
at a
time without looking ahead.
Diary manipulation. Approximately half the participants (randomly
assigned) first received
a
questionnaire labeled "diary study,"
on
which
they were asked to think about
a
specific day later in the semester and to
estimate what they would be doing that day. They estimated the number of
hours they would spend on 10 activities (e.g., going to class, socializing
with friends, studying, eating meals) on a 7-point scale that ranged from no
time to four or more hours. They then filled in 24 blanks, 1 for each hour
of
the
day, according to what they thought they would be doing at that time.
Participants completed the measures for either Monday, November 20
(2
days after the UVA-Virginia Tech football game)
or
Tuesday, Novem-
ber
21
(3 days after the football game). Participants in the control condition
did not receive these measures. To equalize the length of the study and to
be consistent with the cover story that they would be completing different
questionnaires, control participants received
a
personality scale after com-
pleting the dependent measures.
Dependent
measures,
As a baseline measure of happiness, people were
first asked, "How happy would
you say you are
these days?" They
responded on
a
9-point scale that ranged from
1
(not happy)
to 9
(very
happy).
This question has been used in past studies and has been found to
correlate highly with other scales
of
happiness and life satisfaction. For
example, Gilbert
et
al. (1998) found that this item was correlated with
items from Diener, Emmons, Larsen,
and
Griffin's (1985) Satisfaction
With Life Scale
(r =
.86) and with Kamman and Flett's (1983) Affecto-
meter
2
scale
(r
.83). Participants were then reminded that UVA and
Virginia Tech would play each other
in
football
on
Saturday, Novem-
ber 18, and were asked to predict what their "general level of happiness"
would be right after the game and on each of the following
7
days,
if
their
team lost and if their team won. Participants made their predictions on the
same scale on which they had rated their current happiness. People also
predicted how much they would think about the game right after
it
ended
and on each of
the
following 7 days, if their team won and if their team lost.
These predictions were made on 9-point scales that ranged from 1 (not at
all) to
9
(very often). The order
of
the happiness and thought predictions
was counterbalanced.
Results and Discussion
We asked people to predict their happiness and thoughts on the
day
of
the football game and on each
of
the subsequent
7
days.
This length
of
time was arbitrary because we did not know how
long people believed
a
football game would impact their happi-
ness.
As
it
happened, people's predictions started
to
level
off
on
the fourth day after the game,
and
there were few differences
between conditions from this day onward. To simplify the presen-
tation of the data, and to be consistent with subsequent studies, we
report only people's predicted happiness right after the game and
on the next
3
days. (The significant effects involving the diary
manipulation remain significant when all days are entered into the
analyses.) Initial analyses also revealed that there were no
signif-
icant interactions between
the
diary manipulation
and
gender,
824
WILSON, WHEATLEY, MEYERS, GILBERT, AND AXSOM
whether participants were students at UVA or Virginia Tech, or
whether happiness or thought predictions were made first. We thus
collapsed across these variables in subsequent analyses.
Happiness Predictions
There were no significant differences in baseline happiness
between the control and diary conditions (M = 6.47, SD = 1.51 vs.
M = 6.67, SD = 1.41, respectively), *(86) < 1, ns. To control for
individual variation in initial happiness, we subtracted partici-
pants'
baseline scores from their predictions to create an index of
predicted change in happiness.
The scores were analyzed with a 2 (diary: diary vs. control) X 2
(outcome: predictions following a win vs. loss) X 4 (time: predic-
tions for after the game and for the next 3 days) between-within
analysis of variance (ANOVA).
2
Not surprisingly, there was a
strong main effect of outcome, reflecting the fact that people
predicted they would be happier if their team won than if their
team lost, F(l, 85) = 103.64, p < .001 (see Figure 1). There was
also a main effect of
time,
reflecting the fact that people predicted
that their happiness would improve as time passed, F(3>,
255) = 11.62, p < .001, and a significant Outcome X Time
interaction, reflecting the fact that the difference in predicted
happiness after a win versus a loss became smaller over time, F(3,
255) - 102.29, p< .001.
Of greater theoretical interest was the significant Diary X Out-
come X Time interaction, F(3, 255) = 4.23, p < .03. As seen in
Figure 1, this interaction reflects the fact that people in the diary
conditions made more moderate (i.e., closer to baseline) affective
predictions, at least on some days. A closer look at this interaction
revealed that the diary had its strongest effects on people's pre-
dicted happiness on the 3 days after the football game. On predic-
O
ft
9 -i
7 -
6-
i
5
3 -
Win,
Control
Win,
Diary
Lose,
Control
Lose,
Diary
Saturday Sunday Monday Tuesday
Figure
2.
Study
1:
Effects of diary on thought predictions. The higher the
number, the greater the amount of predicted thought about the football
game.
tions for the day of the game (Saturday), neither the main effect of
diary nor the Diary X Outcome interaction was significant, Fs(l,
86) < 1.33, ns. On predictions for the 3 days after the game, the
Diary X Outcome interaction was significant, F(l, 85) = 4.05,
p < .05.
CO
CO
o
Q.
a.
co
X
a>
.c
"5)
8
GO
I
o
CD
£
Win,
Control
Win,
Diary
Lose,
Diary
Lose,
Control
Saturday Sunday Monday Tuesday
Figure I. Study 1: Effects of diary on affective predictions. The higher
the number, the happier people predicted they would be, relative to their
baseline level of happiness.
Thought Predictions
People's predictions about how much they would think about
the game were analyzed with the same 2 (diary) x 2 (out-
come) X 4 (time) between-within ANOVA. Not surprisingly, the
main effect of time was significant, F(3, 258) = 2\2.\l,p < .001,
reflecting the fact that people said they would think less about the
game as time passed (see Figure 2). The main effect of outcome
was also significant, F(l, 86) = 99.82, p <
.001,
reflecting the fact
that people said they would think about the game more after a win
than a loss. These main effects were qualified by an Outcome x
Time interaction, F(3, 258) = 8.51, p < .001, reflecting the fact
that the difference in predicted thought following a win versus a
loss got larger over time.
Of greater theoretical interest was the main effect of the diary
manipulation, F(l, 86) =
9.65,
p < .005 and a significant Diary X
Time interaction, F(3, 258) = 1.20, p <
.001.
As seen in Figure 2,
there was no effect of the diary on the amount of predicted thought
right after the game (on Saturday). As expected, however, people
2
Jn a repeated measures design with several measurements, it is possible
to have an inflated Type I error due to a violation of symmetry assump-
tions.
To avoid this problem, thep levels for all effects involving a repeated
measures factor with more than 2 degrees of freedom, in all analyses in this
article, were corrected using the Greenhouse-Geisser adjustment.
FOCALISM: A SOURCE OF DURABILITY BIAS
825
in the diary condition predicted that they would think about the
game less on succeeding days.
3
Mediation Analyses
We tested the hypothesis that the effect of the diary manipula-
tion on predicted happiness was mediated by its effects on how
much people said they would think about the game, using multiple
regression (as described by Kenny, Kashy, & Bolger, 1998). As an
overall index of predicted thought, we averaged people's thought
forecasts over time and outcome. As an index of predicted happi-
ness,
we averaged people's happiness forecasts over time and then
subtracted their predictions after a loss from their predictions after
a win. (Thus, the larger the score on the happiness index, the more
people thought that their happiness would be influenced by the
outcome of the football game.) As seen in the first column of
Table 1, there was a significant negative relationship between the
dummy-coded diary condition {control = 0, diary = 1) and
predicted thought (this is similar to the main effect of diary on
predicted thought already reported, whereby people in the diary
condition predicted that they would think less about the football
game).
The relationship between predicted thought and predicted
happiness was also significant, adjusting for the effects of the diary
condition. Table 1 also shows the relationship between the diary
condition and predicted happiness, before and after adjusting for
the amount of predicted thought. The reduction in beta weight after
controlling for predicted thought was significant (z = 2.12, p =
.03),
supporting the hypothesis that predicted thought mediated the
effects of the diary manipulation on predicted happiness (Kenny et
al,
1998).
Table 1
Tests of
the
Mediating Role of Predicted Thought
Path Study 1 Study 3 Study 4 Study 5
Diary condition ->
predicted thought
Predicted thought
»
predicted happiness
Diary condition*
predicted happiness
(unmediated)
Diary condition -*
predicted happiness
(mediated)
-0.530*
(.171)
0.461*
(.151)
-0.369
(.250)
-0.124
(.252)
2.12*
-0.574*
(-214)
0.943*
(.490)
-1.613*
(.551)
-1.073
(.594)
1.50
-1.043*
(.467)
-0.412*
(.104)
-1.473*
(.564)
0.256
(.295)
1.057*
-1.475
(.449) (.894)
0.627
(.423)
1.90
-1.086
(1.003)
0.77
Note. The values shown are the unstandardized beta weights and their
standard errors (in parentheses). The beta between predicted thought and
predicted happiness controls for diary condition. The first beta weight
between diary condition and predicted happiness does not control for
predicted thought, whereas the second beta weight does. A meta-analysis
of the mediation effect across the four studies was significant, z = 3.15,
p = .002.
a
The test of the reduction of the unmediated path from diary condition
to predicted happiness, when predicted thought is added as a mediator
(Kenny, Kashy, & Bolger, 1998).
*p < .05.
Summary of Results
As hypothesized, people in the diary condition predicted that, as
time passed, they would think less about the football game than did
people in the control condition. They also predicted that their
happiness would return more quickly to baseline levels. These
results, and the mediation analyses just reported, are consistent
with the focalism hypothesis. One reason that people make endur-
ing predictions about the impact of future events on their happiness
may be that they overestimate how much they will think about
those events.
Because Study 1 did not include a measure of people's actual
happiness after the football game, we cannot conclude that people
in the diary condition made more accurate affective forecasts. On
the basis of our previous research, we assumed that people in the
control condition overestimated how long the football game would
have an impact on their happiness and that this durability bias was
corrected, at least in part, by completing the prospective diary. The
purpose of Study 2 was to test this assumption by seeing whether
college student football fans really are happier in the days after
their team's victories than in the days after their team's losses. We
assessed fans' happiness on several Mondays or Tuesdays in a
subsequent fall semester after Saturdays on which their school had
won a football game, lost a football game, or had an open date.
Study 2: They Experienced a Game
Method
Overview
College football fans at UVA and Virginia Tech rated their overall
happiness on Mondays or Tuesdays for 5 weeks in the fall of 1997, after
several Saturdays on which their schools played football games. The last
ratings were made after the game between UVA and Virginia Tech. Some
participants also listed what they had been thinking about each day. We
predicted that the students' happiness would not be related to the fate of
their school's football game on the previous Saturday, that the fans of the
winner of the game between UVA and Virginia Tech would not be happier
than the fans of the losers, and that the football games would not be
prominent in the students' thoughts.
Participants
Participants were 96 students (52 women, 44 men) at UVA and 167
students (114 women, 53 men) at Virginia Tech who indicated that they
were football fans, using the same criterion as in Study 1. Because the data
were collected on 5 consecutive weeks in psychology classes, and several
students missed at least one class, there were missing data on each of the
measures.
Method
On Wednesday, October 29, 1997, or Thursday, October 30, 1997,
participants rated their overall level of happiness on the same scale as used
3
The diary manipulation appears to have had a somewhat larger effect
when people were predicting a win than when they were predicting a loss,
though neither the Diary X Outcome nor the Diary X Outcome x Time
interactions reached conventional levels of significance, F{\, 86) = 1.13,
ns,
and F(3, 258) =
2.56,
p =
.08,
respectively. The latter trend may be due
to a floor effect, given that there was less room to move down on the scale
when people made predictions about a loss than when people made
predictions about a win.
826 WILSON, WHEATLEY, MEYERS, GILBERT, AND AXSOM
in Study 1. Because people filled out this measure 4 or 5 days after the
previous Saturday's football game, we considered it a baseline measure of
happiness. Participants rated their overall happiness again on the Monday
or Tuesday of each of the succeeding 5 weeks. (Virginia Tech students did
not complete the measures on the fourth week because they were on
Thanksgiving break.) On the Saturdays during this time period, each
school's football team either won a game, lost a game, or had an open date
with no game. The last time of measurement followed the football game
between UVA and Virginia Tech, which was won by UVA.
Some participants received only the overall happiness question. Others
first rated their happiness and then answered an open-ended question
asking them to "jot down the things that you have been thinking about the
most today." These participants were instructed to write down "the thing
you have been thinking about the most first, the thing you have been
thinking about second-most next, and so on."
Results and Discussion
Happiness Ratings
Preliminary analyses found no significant differences in
happiness ratings on Mondays or Tuesdays of each week,
te(153)< 1.23, ns. Nor did the day people fill out the questionnaire
interact significantly with whether a school's football team won,
lost, or had an open date the previous Saturday, Fs < 1.17, ns.
Therefore, we collapsed the data across these days. The relation-
ship between students' happiness and the fate of their school's
football team was assessed in two ways. First, we compared the
average happiness ratings of students at each university following
Saturdays on which their team won a game, lost a game, or had an
open date, after subtracting people's baseline level of happiness
from these ratings. Positive numbers thus mean that people re-
ported greater happiness than their baseline level and negative
numbers mean they reported less happiness than their baseline
level. (Very similar results were obtained when we analyzed the
ratings without subtracting baseline happiness.)
UVA students were slightly happier on the days following a loss
than on the days after a win by their team (A/ = 0.13, SD = 1.15,
vs.
M = -0.04, SD =
1.12).
They were happiest on the days
following an open date (M = 0.40, SD ~
1.30),
which happened
to be the week of the Thanksgiving holiday. The effect of game
outcome was nearly significant, F(2, 42) = 3.42, p = .07, though
in the direction opposite to that predicted by participants in Study 1
(i.e.,
contrary to these predictions, people reported less happiness
after a win than a loss). Virginia Tech students had similar levels
of happiness following a win by their team, a loss by their team, or
an open date (M -
-0.15,
SD = 1.22; M =
-0.01,
SD = 1.68;
and M =
-0.03,
SD = 1.72, respectively), F(2, 206) = 1.12, ns.
Second, we tested whether UVA students were happier than
Virginia Tech students following UVA's victory over Tech. UVA
students were only slightly higher than their baseline level of
happiness (M
0.18, SD =
1.47),
whereas Virginia Tech students
were about as happy as they were at baseline (M
—0.01,
SD =
1.68).
The difference between these means was not
signif-
icant, t(\9S) = 0.81, ns.
Thought Ratings
Some participants in Study 2 answered an open-ended question
about what they had been thinking about that day. A research
assistant coded the number of times people made any mention of
the prior Saturday's football game versus the number of
times
they
mentioned other thoughts. (A second assistant independently
coded a subset of the thoughts, and her ratings agreed perfectly
with the first assistant's ratings.) The results were straightforward:
Football was not focal in people's thoughts. In the UVA sample,
there were a total of 90 responses to the open-ended question on
the four Mondays or Tuesdays that followed a football game. None
of these responses mentioned the prior Saturday's game. In the
Virginia Tech sample, there were a total of 193 answers to the
open-ended question over the four weeks that followed football
games. People mentioned the prior Saturday's game only three
times.
Thus, the football game was mentioned only 3/283 (1%) of
the time in people's thought listings.
Summary of Results
The results of Study 2 suggest that by the Monday and Tuesday
after a college football game, football fans' happiness was not
affected by the outcome of the prior Saturday's game. This result
is consistent with prior research on happiness and subjective
well-being, which has found that happiness is a fairly stable state
that is relatively unaffected by minor past events (e.g., Suh et al.,
1996).
These results further suggest that people in the diary con-
dition of Study 1 made more accurate affective forecasts than
people in the control conditions, because they predicted that their
happiness following a win or loss by their college's football team
would be closer to their baseline level of happiness. People in the
diary condition of Study
1
also predicted that they would think less
about the game than people in the control condition did. Judging
by the open-ended responses of people in Study 2, they were more
correct about these predictions as well.
Comparisons across Studies
1
and 2 must be tentative, of course,
because we cannot be certain how comparable the samples or
football games were. For example, we cannot rule out the possi-
bility that students' actual happiness was affected more by the
outcome of the 1995 than the 1997 game. In Study 3, we used a
within-subjects design in which the same participants predicted
how happy they would be after a football game and reported their
actual happiness after that game. We hypothesized that people in
the diary condition would make more moderate affective forecasts
and would predict that they would think less about the football
game than would people in the control condition. Further, we
hypothesized that the forecasts of people in the diary condition
would be more accurate.
Study 3: They Foresaw and Experienced a Game
Method
Overview
College football fans at UVA predicted how happy they would be and
what they would be thinking about on the days following a football game
between UVA and the University of North Carolina at Chapel Hill (UNC).
As in Study 1, some participants completed a prospective diary question-
naire prior to making these predictions. On the day after the football game,
participants rated their actual happiness.
Participants
Participants were 27 students (19 women, 8 men) from UVA who
indicated that they were football fans, on the same measures used in
FOCALISM: A SOURCE OF DURABILITY BIAS
827
Studies I and 2. The students participated in an initial session for partial
fulfillment of a requirement in an undergraduate psychology course.
Procedure
Predicted happiness and thought. The procedure was identical to that
of Study 1 with the following exceptions: Students participated approxi-
mately 2 months before the football game between UVA and UNC that was
played on November 16, 1996. Those randomly assigned to the diary
condition first completed the same diary questionnaire as in Study 1. The
day they were asked to think about was Monday, November 18. All
participants then predicted what their overall level of happiness would be
following the football game (November 16) and on each of the succeed-
ing 3 days (November 17, 18, and 19) if UVA lost and if UVA won the
game.
We changed the way in which people predicted their thoughts in two
ways.
First, people were asked to predict not only how much they would
think about the football game but also how much they would think about
other matters. Half the participants were randomly assigned to make these
predictions on the same 9-point scales used in Study 1. After rating how
much they would think about the game, people rated how much they would
think about their school work, social life, leisure time activities, paying
job,
family, and any other things not listed. The remaining participants received
an open-ended question that asked them to complete the sentence, "I will
be thinking mostly about these things." All participants predicted what they
would think about after a UVA loss and a UVA win for each of three time
periods (after the football game on Saturday, November 16, and each of the
next two days).
Actual happiness. We assessed people's actual happiness on the day
after the football game. Beginning a week before the game, participants
were contacted and asked to pick up an envelope containing a question-
naire they could complete for credit or payment. We made no mention of
the prediction part of the study that students had already completed.
Participants were instructed to open the envelope on Sunday, November 17
(the day after the football game), after
5
p.m. Participants rated their overall
level of happiness on the same scale on which they made their predictions
at Time 1.
On subsequent pages of the questionnaire we included exploratory
measures of how much people had been thinking about the game and their
recall of their happiness immediately after the game. By necessity, these
questions followed a mention of the football game (e.g., the thought
question asked people how much they had been thinking about the UVA-
UNC game). Because the reference to the football game might have
triggered people's memory of the earlier, prediction part of the study,
thereby contaminating their answers to the questions, these measures were
exploratory.
Results and Discussion
Happiness Predictions
People in the diary and control conditions reported similar levels
of baseline happiness (M = 7.46, SD ~ 1.27 and M = 6.71,
SD = 1.54, respectively), t(25) - 1.37, p > .18. To control for
individual variation in initial happiness, we subtracted baseline
ratings from predictions to create an index of predicted change in
happiness. As in Study 1, we performed a 2 (diary: diary vs.
control) X 2 (outcome: predictions following a win vs. loss) X 4
(time: predictions for after the game and for the next 3 days)
between-within ANOVA on these scores. Not surprisingly, there
was a strong main effect of outcome, reflecting the fact that people
predicted they would be happier if their team won than if their
team lost, F(1, 25) = 53.94, p < .001 (see means in Figure 3).
3i
tn
05
'5.
a.
a
0)
8
CO
I
-2
Win,
Control
Win,
Diary
Control,
Actual
Diary,
Actual +
Lose,
Diary
Lose,
Control
Saturday Sunday Monday Tuesday
Figure
3.
Study 3: Effects of diary on affective predictions. The higher
the number, the happier people predicted they would be, relative to their
baseline level of happiness.
There was also a main effect of
time,
reflecting the fact that people
predicted that their happiness would improve as time passed, F(3,
75)
8.75, p < .001, and a significant Outcome x Time inter-
action, reflecting the fact that the difference in predicted happiness
after a win versus a loss got smaller over time, F(3, 75) = 60.68,
p < .001.
Of greater theoretical interest were the significant Diary X
Outcome and Diary X Outcome x Time interactions, F(l,
25) = 8.30, p < .01, and F(3, 75) = 4.00, p < .03, respectively.
As m Study 1, these interactions reflect the fact that people in the
diary conditions made more moderate