Events and Subjective Well-Being: Only Recent Events Matter

Abstract

The effect of life events on subjective well-being (SWB) was explored in a 2-year longitudinal study of 115 participants. It was found that only life events during the previous 3 months influenced life satisfaction and positive and negative affect. Although recent life events influenced SWB even when personality at Time 1 was controlled, distal life events did not correlate with SWB. SWB and life events both showed a substantial degree of temporal stability. It was also found that good and bad life events tend to covary, both between individuals and across periods of the lives of individuals. Also, when events of the opposite valence were controlled, events correlated more strongly with SWB. The counterintuitive finding that good and bad events co-occur suggests an exciting avenue for explorations of the structure of life events.

Full text

Journal of Personality and Social Psychology
1996.
Vol. 70. No. 5, 1091-1102
Copyright 1996 by the American Psychological Association. Inc.
0022-3514/96/S3.00
Events and Subjective Weil-Being: Only Recent Events Matter
Eunkook Suh and Ed Diener
University of Illinois at Urbana-Champaign
Frank Fujita
Indiana University at South Bend
The effect of life events on subjective well-being (SWB) was explored in a 2-year longitudinal study
of 115 participants. It was found that only life events during the previous 3 months influenced life
satisfaction and positive and negative affect. Although recent life events influenced SWB even when
personality at Time 1 was controlled, distal life events did not correlate with SWB. SWB and life
events both showed a substantial degree of temporal stability. It was also found that good and bad
life events tend to covary, both between individuals and across periods of the lives of individuals.
Also,
when events of the opposite valence were controlled, events correlated more strongly with
SWB.
The counterintuitive finding that good and bad events co-occur suggests an exciting avenue
for explorations of the structure of life events.
Despite the claims of astrologers, psychologists are keenly
aware of their limited ability to predict and control the occur-
rences of various life events. Scientists have instead focused
their attention on the psychological consequences of life events
and how individuals cope with the stresses created
by
these "ex-
ogenous shocks" (e.g., Dohrenwend & Dohrenwend, 1981;
Rabkin
&
Struening, 1976; Selye, 1976; Vaillant, 1977). Nega-
tive events in particular have received a great amount of atten-
tion, largely because of their threatening potential impact on
one's well-being. Decades of research show that individuals'
cognitive interpretation styles (Davis, Lehman, Wortman, Sil-
ver, & Thompson, 1995; Peterson & Seligman, 1987; Taylor,
1983;
Thompson, 1981), coping patterns (Folkman, Lazarus,
Dunkel-Schetter, DeLongis, & Gruen, 1986; Lazarus & Folk-
man,
1984;
Zeidner& Hammer, 1990), and personality factors
(Bolger, 1990; Kobasa, 1979; Kobasa, Maddi, & Kahn, 1982;
Mullen
&Suls,
1982;
Ormel, Sanderman,
&
Stewart, 1988) can
mediate the ways in which these stressful life events will be ex-
perienced. Despite the large volume of life events research, the
potential influence of events on positive outcomes, such as sub-
jective well-being
(SWB),
remains relatively unexplored.
Although recent studies have shown that most people are
happy (Diener & Diener, in press) and consider positive emo-
tions more normative than negative emotions (Sommers,
1984),
the extent to which life events influence individual levels
of
SWB
is not fully understood. In addition to the prominent
influence of coping models on life events research, several SWB
research
findings
may have unintentionally discouraged further
research on this topic. For instance, the effects of objective life
circumstances, such as income (Diener, Sandvik, Seidlitz, &
Eunkook Suh and Ed Diener, Department of Psychology, University
of Illinois at Urbana-Champaign; Frank Fujita, Department of Psy-
chology, Indiana University at South Bend.
Correspondence concerning this article should be addressed to Ed
Diener, Department of Psychology, University of Illinois, 603 East Dan-
iel Street, Champaign, Illinois 61820. Electronic mail may be sent via
the Internet to ediener@s.psych.uiuc.edu.
Diener, 1992), health (Okun
&
George, 1984), years of educa-
tion (Diener, 1984), and physical attractiveness (Diener, Wol-
sic,
&
Fujita, 1995), are often found to be small. Similarly, the
findings of Campbell, Converse, and Rogers (1976) indicate
that the summed effects of demographic variables on SWB are
small.
Other researchers
have shown
that well-being
is
primarily
determined by enduring individual characteristics rather than
by external life circumstances (Costa & McCrae, 1980, 1984;
Costa, McCrae, & Zonderman, 1987; Diener, Sandvik, Pavot,
& Fujita, 1992). Finally, on the basis of the framework of
adap-
tation level theory, Brickman, Coates, and Janoff-Bulman
(1978) discounted the influence of life events on happiness by
suggesting that the effects of salient life events eventually wear
off(habituation) and, by comparison, make mundane experi-
ences seem less extreme (contrast). They found that lottery
winners were not significantly happier than controls and also
that people with spinal cord injuries were only slightly less
happy than other people.
The studies just mentioned share the common methodologi-
cal problems of
a
cross-sectional design. By limiting the mea-
surement of people to a single point in time, they fail to show
the process of adaptation or habituation among participants.
Moreover, there is difficulty in assessing the degree of psycho-
logical impact of the incident on individuals. Silver (1982),
however, conducted a longitudinal study to understand how
people cope during the first 8 weeks after the trauma of
a
dis-
abling accident (spinal cord injury).
She
found that
the
patients
adapted surprisingly well to their tragic experiences during the
course of the next several months. This study implies that ad-
aptation indeed does occur, and in a surprisingly short time.
Longitudinal
findings
reported by Costa et al. (1987) and Die-
ner, Sandvik, Pavot, and Fujita (1992) are supportive of Silver's
finding in suggesting that life events or changes do hot have a
large influence on
SWB
over
long
periods of time (10 years).
The aforementioned studies can be more or less grouped as
"top-down" theories of
SWB
that suggest that SWB is essen-
tially determined by one's personality rather than by external
life circumstances or events. Recent studies
(Brief,
Butcher,
George, & Link, 1993; Feist, Bodner, Jacobs, Miles, & Tan,
1091

1092
SUH, DIENER,
AND
FUJITA
1995) suggest, however, that
SWB is
best explained
by
inte-
grative models that consider both
the
effects
of
immediate life
circumstances (e.g., objective health)
and
global personality
di-
mensions. Such findings, along with
the
earlier adaptation stud-
ies (Brickmanetal., 1978;Silver, 1982), lead
us to
believe that
life events
and
personality
may
differ
in
terms of their duration
of effect
on
SWB; personality
is
likely
to
have long-term effects
on SWB, whereas
the
effects of life events
on
SWB
may be
more
short term. Thus,
the
current study
was
based
on a
conceptual
framework suggesting that one's current level of SWB
is a
prod-
uct
of
both long-term personality (e.g., extraversion
and neu-
roticism)
and
recent events that temporarily modify one's
ex-
perience of well-being.
The current study represented
a
further extension
of
Silver's
(1982) findings
to
life events
in
general.
We
desired
to
track
people during
a
transitional period
of
their life (college
graduation)
and
examine whether events gradually lose impact
on SWB over
time.
We particularly wanted
to
assess the amount
of time required
for
people
to
adapt
to
life events.
Do
life events
lose their impact
on SWB in a
matter
of
months,
as
Silver's
(1982) findings suggest,
or do
they last
for
years?
We
also
wanted
to
examine
the
adaptation
to
both good
and bad
events.
Do people adapt
to
good events
as
well
as
they adapt
to bad
experiences?
If
so,
do
they adapt
to
both with equal rapidity?
Finally, we examined whether life events influence SWB beyond
the effects
of
personality.
Several additional factors about
the
relation of life events
and
SWB were explored.
A
number
of
studies (Block
&
Zautra,
1981;
Headey
&
Wearing, 1989; Magnus, Diener, Fujita,
& Pa-
vot, 1993) have shown that positive
and
negative life events
cor-
relate positively, suggesting that more active people experience
both more good
and bad
events because
of
their greater degree
of involvement with
the
world.
In
contrast,
the
lives
of
some
people
are
less dramatic; they experience fewer
of
both types
of
events. Following
the
lead
of
previous findings,
we
addressed
two further questions. First, we sought to determine whether
the
coupling
of
positive
and
negative events moderates SWB. That
is,
if people experience both types
of
events,
will control
of
one
type increase
the
impact
of the
other
on
SWB? This would
be
the case
if
negative events
do in
fact suppress
the
impact of
pos-
itive events
on
SWB,
and
vice versa. Second,
we
tried
to
ascer-
tain whether this covariation
is
strictly
a
function
of
activity
level
or
whether
it
also occurs within individuals.
In
other
words,
is the
coupling
of
positive
and
negative events primarily
an individual-differences phenomenon,
or
does
it
also occur
within individuals (depending
on the
life period)
?
We also sought
to
explore
the
stability
of
personality,
SWB,
and events. Costa
and his
colleagues (Costa,
1994;
Costa
&
McCrae,
1988;
McCrae
&
Costa,
1990)
have argued that
per-
sonality
is
extremely stable. They claimed that
it is
basically
set
by
the
time adults reach
the age of
30
and is no
longer reactive
to external input, except under unusual circumstances.
SWB,
by comparison,
is
presumably somewhat less stable than
per-
sonality because
it is a
function
of
both personality
and
events.
Schwartz
and
Strack (1991) have convincingly demonstrated
that momentary situational factors
can
influence
SWB
judg-
ments.
But are
there also stable influences that create
a
tempo-
ral stability
in
SWB? Headey
and
Wearing (1989) argued that
life events
are
stable because they are,
in
part, generated
by the
individual's personality. Although events were believed
to be
largely caused
by
random causes outside
of
the individual,
re-
cent studies show that they also flow
in
part from personality.
In
support
of
this claim, Magnus
et al.
(1993) showed that extra-
version
and
neuroticism predicted
the
occurrence of future
ob-
jective events
for
individuals. Thus,
one
would expect life
events, both positive
and
negative,
to
show temporal stability
because they
are in
part influenced
by
personality, which
is
very
stable. Furthermore, there
are
stable aspects of individuals' life
circumstances that might also predispose them
to
stable levels
of positive
and
negative life events.
Finally,
a
number of researchers have suggested
an
alternative
view
on the
effects
of
personality
in the
relation between stress-
ful,
negative life event reports
and
well-being (Brett,
Brief,
Burke, George,
&
Webster,
1990; Brief,
Burke, George, Robin-
son,
&
Webster, 1988; Watson
&
Pennebaker, 1989). These
au-
thors suggest that
the
personality disposition
of
negative affec-
tivity
and
neuroticism introduces
a
spurious effect
in the
life
event
and
well-being relation
by
reasons other than actual
differences
in the
occurrence
of
events (e.g., report style
and
attribution style). Zautra, Reich,
and
Guarnaccia (1990),
how-
ever, found that neuroticism
did not
have
a
substantive effect
on
the reporting
of
the frequency
of
life events.
We
attempted
to
study this issue
by
examining whether
the
correlations between
negative life event measures
and SWB are
affected when
neu-
roticism is partialed
out.
In summary, life events
and
circumstances,
in
comparison
with personality, appear
to
have
a
surprisingly small effect
on
SWB (Costa
et al., 1987;
Diener, Sandvik, Pavot,
&
Fujita,
1992;
McCrae
&
Costa, 1988).
In an
attempt
to
understand
the
theoretical significance
of
this finding,
we
examined
the
effects
of life events
on SWB in
several directions. First,
we
explored
whether proximal events influence SWB, whereas distal events
do
not. On the
basis
of
past findings,
we
hypothesized that
the
correlation between life events
and
SWB will
be
significant only
when recent events
are
involved. Thus,
we
sought
to
determine
the duration
of
various life events' influence
on
individuals'
SWB.
In
addition, we explored whether the decay period is iden-
tical
for
both positive
and
negative experiences. According
to
Frijda (1988), negative emotions require more adaptation time
than
do
positive emotions.
In
light of this intriguing suggestion,
we investigated whether good
and bad
events
in
general have
the
same decay period
or
whether
one
type
of
event
has
much
shorter lived effects than
the
other.
Second, we examined
the
interplay between positive
and neg-
ative life events. If good
and bad
events co-occur
to
some extent,
possibly because individuals
go
through many life changes that
have both good
and bad
aspects,
is
their influence
on SWB mit-
igated because they suppress
the
effects of each other? Psychol-
ogists have traditionally regarded
the
emotional consequences
caused by desirable versus undesirable life events as distinguish-
able:
Desirable events
are
more linked
to
psychological well-
being,
and
undesirable events have adverse effects
on
mental
health (Vinokur&Selzer, 1975; Zautra
&
Reich, 1983; Zautra
et
al.,
1990).
If,
however, there
is a
tendency
for the
effects
of
one type
of
event
to be
canceled
by
events
of the
opposite
va-
lence,
the
impact
of bad
events
on SWB may
increase
if one

EVENTS AND SUBJECTIVE WELL-BEING
1093
controls for the effects of good events. Similarly, one can specu-
late that the impact of good events on SWB might increase if
the influence of bad events is held constant.
Third, we assessed the stability of personality, good and bad
life events, and SWB. We hoped to determine the stability of
SWB
in the midst of
the changes people
experience
in
their
lives.
We examined whether life events are somewhat stable, which
should
be
true if they
do,
in fact, follow in part from personality.
Finally, we sought to explore whether life events have effects on
SWB beyond the influence of personality and prior level of
SWB.
Method
Overview and Sample Description
This study was carried out as part of a longitudinal research project
on SWB and personality. The data for Time
1
were collected from 222
individuals (110 men and 112 women) enrolled in a semester-long
course on SWB either in the fall semester of 1991 or in the spring se-
mester of 1992. Virtually all participants were either junior or senior
college students (20-21 years old) majoring in psychology, and most
were White. Personality, SWB, and life events data were collected as
class exercises, completed either at home or in laboratory sessions.
The time interval between measurement occasions is crucial in de-
tecting effects in longitudinal studies. We selected a period of 2 years,
broken into smaller
intervals.
This period is reasonably short, to prevent
severe memory distortions of past events, but also long enough to trace
the adaptation processes during the interval. In the fall of 1993, 155 of
the 222 participants in the Time 1 study were located. The follow-up
questionnaire package was mailed to the participants in December of
1993.
One hundred nineteen individuals (78%) responded during the
spring of 1994 (Time 2). To determine whether there was any differen-
tial attrition of participants, we compared the Time
1
Satisfaction with
Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985), extra-
version, neuroticism, and positive and negative life event scores of those
who participated in the follow-up study with the scores of those who
were involved in the Time 1 study only. No significant difference was
found between the two groups among the five measures (all ts < 1).
Questionnaires sent out in the Time 2 study included a number of per-
sonality measures, a 100-item life events checklist, happiness and life
satisfaction questions, a current mood measure, and an affect frequency
measure. Because of the incomplete data set obtained from 4 partici-
pants,
the final follow-up sample consisted of 72 women and 43 men
(115 in total). The number of participants differed slightly in several of
the analyses as a result of occasionally missing data. Participants were
paid $20 for their responses. The check was mailed with a thank you
letter that summarized our findings and future research plans.
Measures
Personality. Participants completed the extraversion and neuroti-
cism subscales of the NEO Personality Inventory (NEO-PI; Costa &
McCrae, 1985) at both Time 1 and Time 2. Extraversion and neuroti-
cism at Time
1
were assessed by using each of the 48-item scales of the
NEO-PI. As a means of measuring temporal stability, extraversion and
neuroticism were measured by the NEO Five Factor Inventory (NEO-
FFI) at Time 2. The NEO-FFI (12-item scales) is a shortened version
of the NEO-PI that allows a
brief,
comprehensive measure of the five
domains of personality. The Time 2 alpha coefficients for the short ex-
traversion and neuroticism scales were .77 and .86, respectively.
Subjective
well-being.
The SWLS (Diener, Emmons, et
al.,
1985), doc-
umented as having adequate psychometric properties, was used to measure
the participants' global, cognitive assessment of their life as a whole (Pavot
& Diener, 1993; Pavot, Diener, Colvin, & Sandvik, 1991). The SWLS was
administered twice at Time
1,
with a separation of a few
weeks.
As a means
of gai ning a more stable Time
1
SWLS score, the mean of the two scores
was used as the participants' Time
1
life satisfaction
score.
The total SWLS
score was obtained by summing the ratings for the five 7-point items,
which produced a possible range of
5
to 35. The alpha coefficients of the
SWLS were
.91
at Time
1
and .82 at Time 2.
On the basis of major emotion theories, 24 discrete emotions were
selected for measuring the frequency of the participants' positive and
negative emotional experience at both Time I and Time 2 (see Diener,
Smith, & Fujita, 1995). The 24 emotions used to obtain positive affect
(PA) and negative affect (NA), respectively, represented two positive
and four negative emotion
categories:
joy (joy, happiness, contentment,
and pride), love (love, affection, caring, and fondness), fear (fear, worry,
anxiety, and nervous), anger (anger, irritation, disgust, and rage), sad-
ness (sadness, unhappiness, depression, and loneliness), and shame-
guilt (shame, guilt, regret, and embarrassment). Diener, Smith, &
Fujita (1995) identified six coherent factors from the 24 emotion adjec-
tives,
including higher order PA and NA factors. The participants indi-
cated how often they experienced each of the 24 emotions by using a 7-
point scale ranging from never (I) to always (7). The PA score was
obtained by averaging the eight joy and love items. Similarly, the mean
sum of the emotion adjectives in the four negative categories, fear, anger,
sadness, and shame-guilt, was used as the NA score. The Time
1
Cron-
bach's alphas for PA and NA were .82 and
.91,
respectively. At Time 2,
the alphas were .88 for PA and .86 for N A.
Life
events.
At Time 1, participants were asked to indicate their ex-
periences during the previous 4 years with various events on an 88-item
life events checklist (35 positive
events,
46 negative
events,
and
7
neutral
events). Items were selected from widely used life events measures: the
List of Recent Events (Henderson, Byrne, & Duncan-Jones, 1981),the
Social Readjustment Rating Scales (Holmes & Rahe, 1967), and the
Life Experiences Survey (Sarason, Johnson, & Siegel, 1978). The new
checklist was intended to maximize base rates of responses from the
sample (mostly college juniors and seniors) and also to balance the dis-
tribution of event valence and significance. As can be seen in the exam-
ples in Table 1, the checklist consisted of very significant events (e.g.,
"death of a close friend" and "marriage") as well as relatively mundane
events (e.g., "joined a club/group" and "gained weight") of each
valence.
At Time 2, the events checklist was increased to a total of 100 items
by adding 10 positive events, 3 negative events, and 1 neutral event (45
positive events, 49 negative events, and 6 neutral events) and deleting 2
neutral items. This revision was made so as to match the number of
positive and negative events more evenly and improve base rate re-
sponses by adding items such as "acquired a new TV or stereo." Among
the total items, 73 were objective life events
{e.g..
had an operation) and
27 were more subjective in nature (e.g., received an unfairly low grade
in a course). In this new checklist, participants were asked not only to
indicate the occurrence of each event but also to report how long ago
the event had occurred. Participants were instructed to place a number
from a 7-point scale next to each event they had experienced during the
past 4 years. The scale was coded as follows: past 3 months (1), 4-6
months ago (2), 7-12 months ago (3), I-2 years ago (4), 2-3 years ago
(5),
3-4 years ago (6), and did not happen (7).
Results
The means, standard deviations, and stability coefficients of
the major variables are summarized in Table 2. The large
differences of means and the standard deviations for
the
person-

1094
SUH, DIENER, AND FUJITA
Table 1
Sample of Items
From the
Life
Events Checklist
Event(1991-I993)
Number of
participants
reporting
%of
total
Positive
Marriage (self)
Parent/relative gave you
a
start in business/job
Engagement (sell)
Became an uncle/aunt
Sibling got
married
Performed in an artistic event (concert, etc.)
Got into graduate school
Promotion/ raise
Improvement in work hours
or
conditions
Reconciliation with romantic partner
Improvement in
financial
status
Got a car
Joined
a
club or group
Involvement in a steady romantic relationship (at least
2
months)
Made a new close friend
Received an
A
for
a
college course
Negative
Divorce/marital separation (self)
Parents divorced or separated
Had an abortion (self/spouse/romantic partner)
Victim of a violent crime (rape/assault/etc.)
Did not get into graduate school
Fired/laid off
Death of
a
close friend
Serious illness
or
accident requiring hospitalization
Had
problems getting
along with coworkers
Had an operation
Death of
a
close family member
Troublesome neighbors
Family had
financial
problems
Had a project or assignment overdue
Unable to locate job
Gained weight (at least
10
pounds)
Long-term (at least
3
months) romantic relationship ended
Had a difficult time deciding on career or life goals
8
11
15
20
23
27
32
40
47
49
61
69
77
94
104
HE
1
2
5
7
8
9
10
13
23
26
33
35
36
42
48
60
63
93
7
10
13
17
20
23
28
35
41
43
53
60
67
82
90
97
I
2
4
6
7
8
9
1!
20
23
29
30
31
37
42
52
55
81
ality and life events measures across Time
1
and Time 2 are due
to changes in the number of items used in the follow-up study
rather than actual changes in the participants' scores. The total
numbers of positive and negative events that each participant
reported to have happened during the previous 4 years were
used, respectively, as the individuals' positive and negative event
scores.
Stability
As can be seen in Table 2 (in boldface), SWB, personality, and
life events showed a significant degree of stability. Although statis-
tically significant and strong, the temporal stabilities of extraver-
sion and neuroticism were somewhat lower than those reported by
Costa and McCrae (1988). It may be that the use of the NEO-
FFI instead of the longer version of the NEO-PI at Time 2 slightly
lowered the stability in our data. When the SWB measures were
corrected for attenuation, the stability of the SWLS increased to
.76,
the stability of PA increased to .66, and the stability of NA
increased to .69. At the same time, the fact that the unattenuated
stability coefficients of SWB measures were noticeably lower than
their respective alpha values implies that some of the unaccounted
variance could be attributed to other external sources, such as
changes in events and life circumstances. The moderately strong
stabilities for SWB measures are consistent with the idea that SWB
can be influenced by immediate situational factors (Schwartz &
Strack, 1991) as well as the notion that there are stable personality
influences on SWB (Costa & McCrae, 1980, 1984; Headey &
Wearing, 1991).
We hypothesized that if events are in part "created" by indi-
viduals, life events should also have a significant degree of sta-
bility. Our prediction was confirmed by both types of events. As
shown in Table 2, individuals who experienced good events
more often at Time 1 continued to do so after 3 years at Time 2
(r = .27, p < .01). A similar pattern was observed for bad
events, which tended to occur more persistently for those indi-
viduals who had previously experienced more such events (r =
.52,
p< .001). An interesting point to note, however, is that bad
events seemed to be experienced with much more consistency
than good events, /(114) = 3.48, p < .01. Contrasting results

EVENTS AND SUBJECTIVE WELL-BEING
1095
Table 2
Means,
Standard
Deviations,
and
Correlations
of Major
Vhriables
Period and variable
Time 1
l.Extraversion
2.
Neuroticism
3.
Positive life events
4.
Negative life events
5.
Positive affect
6. Negative affect
7.
Life satisfaction
Time 2
8. Extra version
3
9. Neuroticism
8
10.
Positive life events
b
11.
Negative life events
0
12.
Positive affect
13.
Negative affect
14.
Life satisfaction
M
SD
1
—
.41
.19
.00
.41
-.40
.41
.67
-.41
.25
-.06
.33
-.39
.36
118.47
20.35
2
—
.01
.22
-.40
.69
-.48
-.33
.59
-.14
.13
-.21
.47
-.22
91.88
22.63
3
—
.64
.05
.03
.09
.15
-.03
.27
.17
.05
-.08
.17
17.75
4.96
4
—
-.20
.32
-.32
-.10
.19
.23
.52
-.14
.19
-.10
15.30
5.84
5
—
-.31
.46
.43
-.22
.17
-.11
.56
-.20
.29
17.51
3.61
6
—
-.48
-.40
.53
.05
.35
-.19
.61
-.31
10.04
2.05
7
—
.48
-.41
.18
-.23
.28
-.42
.66
23.87
5.87
8
—
-.53
.18
-.23
.54
-.46
.55
30.33
6.29
9
—
-.15
.25
-.39
.69
-.48
19.62
8.16
10
—
.50
.19
-.07
.19
23.30
5.63
11
—
-.08
.30
-.21
15.85
6.37
12
—
-.35
.52
17.78
1.92
13
—
-.50
10.11
3.82
14
—
24.56
5.83
Note. N = 111 to 115. Correlations of .19 and higher were significant at p < .05, and correlations of .25 and higher were significant at p < .01.
Boldface values indicate stability coefficients.
a
Used the NEO Five Factor Inventory scales (12 items).
b
Total number of items was increased from 35 at Time
1
to 45 at Time
2.
Total number
of items was increased from 46 at Time
1
to 49 at Time 2.
were previously obtained by Headey and Wearing (1989):
Good events were slightly more stable than bad events. At this
point, it
is
difficult to offer
a
substantive reason for the contrast-
ing results. One possibility, however, is that the difference was
due to an age-related factor(s). All of the participants in the
present study at Time
1
were either college juniors or seniors,
whereas the participants in Headey and Wearing's study were
sampled from a more general population (an age range of
18
to
65 years). The present results suggest the need for future re-
search on this
issue:
whether positive and negative events actu-
ally differ in terms of stability, and, if they
do,
what the possible
reasons may be
(e.g.,
recall style or activity level).
Adaptation
As summarized in Table 3, the overall correlations between
events in various periods and SWB supported our suggestion
that only recent life events influence SWB and that the impact
of magnitude drops quickly afterward. Although the exact du-
ration of time is likely to vary across individuals and different
life events, the sample of events used in our study affected indi-
viduals' life satisfaction and PA for less than 6 months. An im-
portant implication of our finding is that the time distances of
events should be considered as a crucial factor in determining
the degree of
the
influence of life events on SWB. Overall, the
findings of the present study support Headey and Wearing's
(1989) "dynamic equilibrium model" of SWB. The model
claims that deviations from normal patterns of life events mod-
ify individuals' SWB, but only temporarily. Our findings sup-
port the model's prediction that SWB reverts to a level prede-
termined by individuals* personality as soon as the pattern of
life events regains its equilibrium. According to our findings,
this regression process probably takes several months.
The results for NA were similar to the life satisfaction and
PA
findings.
The sole exception to this general pattern was the
unexpected correlation between NA and bad events that had
occurred 3 to 4 years in the past (r
— ,21
f
p < .05). Further
examination of our data showed that the events that were re-
Table 3
Correlations Between Subjective
Welt-Being
Measures
and
Life
Event Period
Event period
and valence
3 months ago
Positive events
Negative events
6 months ago
Positive events
Negative events
1 year ago
Positive events
Negative events
2 years ago
Positive events
Negative events
3 years ago
Positive events
Negative events
4 years ago
Positive events
Negative events
All periods
Positive events
Negative events
Life
satisfaction
.25**
-.28**
.16
.12
-.01
-.02
.01
.00
-.04
-.08
-.10
-.03
.19*
-.21*
Positive
affect
.02
-.30**
.26**
.04
.09
.06
-.02
.01
.05
,06
.03
.04
.19*
-.09
Negative
affect
.00
.30**
-.07
.11
-.09
-.04
-.07
.11
.05
.11
.07
.21*
-.07
.30**
Note.
A^
=115.
*p<.05.
**/?<.01.

1096
SUH, DIENER, AND FUJITA
ported to have happened most frequently during
this
particular
time period were also supposedly the most negative items on
our checklist. For instance, "death of
a
close family member,"
"parents divorced," and "abortion"
were
among the
11
negative
events that happened most frequently during this time period.
However, because the rest of the distal periods did not correlate
with NA, it seems unlikely that negative events in general are
more difficult to adapt to than positive events. It seems more
reasonable to restrict this possibility to certain exceptionally
undesirable events. Nevertheless, this speculation needs to be
clarified by future studies.
Finally, to examine whether neuroticism had a marked in-
fluence on the report of negative life events and SWB, we par-
tialed out the neuroticism scores in the correlations among the
six negative event periods and life satisfaction,
PA,
and
NA.
The
overall pattern of adaptation was similar to the zero-order cor-
relation results. However, the partial correlations between the
most recent bad events (past
3
months) and life satisfaction, PA,
and NA dropped, respectively, to -.18, -.22, and .16. Implica-
tions of this result are mentioned in the Discussion session.
Event
Structure
To
explore whether
positive
and negative events suppress each
other's effects on SWB over time, we analyzed the correlations
between aggregated life
events
and
SWB
after holding the effects
of the opposite valenced events constant. After the effects of neg-
ative events had been controlled, the correlations between ag-
gregated positive life events and life satisfaction, PA, and NA,
respectively, increased to .35 {p < .001), .27 (p < .01), and
-.27 (p < .01). Similarly, when the effects of positive events
were held constant, the correlations between aggregated nega-
tive events and life satisfaction, PA, and NA, respectively, in-
creased to -.36 {p < .001), -.22(p< .05), and .39(p< .001).
As can be seen by comparing the partial correlations with the
zero-order correlations presented in Table 3, the relation be-
tween life events and SWB became more pronounced when
events of the opposite valence were controlled. Moreover, when
the opposite type of event
was
controlled, events of both valence
seemed to influence both PA and NA. The theoretical implica-
tions of this important
finding
are discussed later.
Previous findings (Block & Zautra, 1981; Magnus et al.,
1993)
have
also shown that positive and negative life
events
cor-
relate across individuals; some people tend to experience both
good and bad events more frequently than others. This pattern
was
strongly replicated in our study
{r =
.50, n= 115,p < .001).
We
extended the previous
findings
by
examining whether events
of both
types also
correlate within
individuals.
That
is,
are there
periods within the idiographic course of life when both good
and bad events happen more or
less
often?
To conduct a within-subject correlational analysis of positive
and negative events, we standardized both the good event and
bad event scores for the six time periods within subjects. This
allowed us to eliminate individual differences between partici-
pants who experienced many life events and participants who
experienced few events within the same time periods.
A
signifi-
cant within-subject correlation was obtained between the peri-
ods of good events and bad events (r
=
A1, p < .001), implying
that good and bad events are indeed more likely to occur to-
gether during the
same
periods
in the lives
of most young adults.
One potential artifact might explain this intriguing finding.
Regardless of the actual times of occurrence, people might sim-
ply recall more recent events, whether bad or good, and fewer
distal events. Thus, good and bad events might appear to co-
occur because of a memory artifact.
We
checked
this
possibility
by partialing out
the
temporal order of events from the good and
bad event
correlation.
The partial correlation barely dropped in
comparison with the zero-order correlation (r =
.39,
p
<
.001),
suggesting that positive and negative events do happen closely
together within individuals' lives. This counterintuitive finding
becomes more understandable if one thinks about the charac-
teristics of life events that occur during transitional periods in
life:
They usually create more good as well as more bad experi-
ences.
Major transitions, such as moving, divorce, a new
job,
or
getting married, are likely to have both positive and negative
outcomes.
Personality
In general, the results of our study support the position that
SWB is predictable from individuals' stable personality (Costa
& McCrae, 1980, 1984; Diener, Sandvik, Pavot, & Fujita,
1992).
However, we wanted to explore whether recent events
influence SWB beyond the effects of Time 1 personality. We
performed a hierarchical multiple regression analysis, predict-
ing Time 2 life satisfaction, PA, and NA by entering three sets
of predictors in the following order: Time 1 personality
(extraversion and neuroticism), recent events (positive and
negative), and distal
events
(positive and negative). In addition,
we performed another hierarchical multiple regression analysis
in which individuals' initial SWB baseline level at Time
1
was
controlled. Instead of controlling for Time
1
personality,
we
en-
tered individuals' initial
SWB
level at Step
1
in the second anal-
ysis.
In both cases, it was predicted that if recent life events in-
fluenced Time
2
SWB measures beyond the effects of one's per-
sonality or initial SWB level, significant F changes would be
observed when the recent events block was entered in Step 2.
On the basis of the correlational
findings
in Table 3, we opera-
tionalized recent events as incidents that happened during the
previous 6 months; distal events were operationalized as those
that occurred from 7 months to 4 years in the past.
As
summarized in Table
4,
results indicated that Time
1
per-
sonality and recent life events
were
significant predictors of cur-
rent SWB. As expected, recent events predicted SWB beyond
the effects of personality (Fchanges for life satisfaction,
PA,
and
NA were all significant at p < .05 at Step 2). However, events
that had occurred more than 7 months previously failed to add
significant increments to the prediction of current SWB level.
More specifically, NA was predicted by Time
1
personality and
recent negative events. Life satisfaction and PA, however, were
predicted
by
Time
1
extraversion and both positive and negative
recent
events.
Discussions on this "crossover" effect from nega-
tive life events to life satisfaction and PA follow in the next
section.
As shown in Table 5, results of the regression analysis ob-
tained after controlling initial
SWB
level resembled the findings

EVENTS AND SUBJECTIVE WELL-BEING
1097
Table
4
Hierarchical
Regression
Analysis for
Predicting
Time
2
Subjective
Well-Being:
Controlling
Time
I
Personality
Variable
R
2
entered
R
2
change Fchange
Time
1
personality
Extraversion
Neuroticism
Recent events
Positive
Negative
Distal events
Positive
Negative
Life satisfaction
.369
8.59***
.560 .191
.564 .004
13.82*
0.34
.245**
-.028
,369***
-.330***
.087
-.072
Time I personality
Extraversion
Neuroticism
Recent events
Positive
Negative
Distal events
Positive
Negative
Positive affect
.335 6.90*
.422 .087
.430
.008
4.30*
0.41
.237*
-.059
.191
-.234*
.045
.048
Time
1
personality
Extraversion
Neurolicism
Recent events
Positive
Negative
Distal events
Positive
Negative
Negative affect
.513 19.50***
.564 .051
.586 .022
4.30*
1.97
-.196*
.324***
-.079
.230**
-.121
.203
Note. N
=
112, Recent events = less than 6 months ago; distal events
= from 7 months to 4 years ago.
*p< .05.
**p
<
.01.
m
**p
< .001.
of
the
first analysis. Similar
to the
previous analysis, recent life
events significantly increased
the
prediction
of
Time
2
SWLS
score
and PA,
whereas distal events
did not.
Both recent
and
distal life events, however, failed
to add
significant increments
to
the
prediction of Time 2 NA when initial NA level was
con-
trolled. Overall, results
of the
regression analyses once again
supported
our
hypothesis that,
at
least
for a
short period
of
time,
recent life events
do
affect SWB beyond
the
influence
of
stable personality or baseline levels of
SWB.
Discussion
A majority of the participants
in
the present study
(81 %)
in-
dicated that they
had
a difficult time deciding on their career or
life goals
in
recent periods. As regards the study of the relation
between life events
and
SWB, such individuals
who
were
at a
transitional period of life provided interesting information that
might
not
have been detectable during more stable life periods.
On
the
other hand, because
of
the
relatively homogeneous
na-
ture of our sample (recent college graduates),
it
remains
to be
shown whether our
findings
can be
extended to middle-aged and
older
adults,
as well
as to
less educated
and
high-income popu-
lations.
As
Sears (1986) has cautioned, data obtained from col-
lege student participants
who
have relatively weak self-defini-
tions
and
uncrystallized attitudes
may not
accurately portray
the social behavior
of
the general population
in
everyday life.
Although this problem
is not
unique
to the
present research,
this relatively unstable nature
of
college student data
has im-
portant implications
in
generalizing
the
findings
to
personality
stability
in
adulthood. Haan, Millsap,
and
Hartka (1986),
for
instance, suggested that marked shifts
in
personality occur dur-
ing
the end of
adolescenc