The Benefits of Frequent Positive Affect:
Does Happiness Lead to Success?
Sonja Lyubomirsky
University of California, Riverside
Laura King
University of Missouri—Columbia
Ed Diener
University of Illinois at Urbana–Champaign and The Gallup Organization
Numerous studies show that happy individuals are successful across multiple life domains, including
marriage, friendship, income, work performance, and health. The authors suggest a conceptual model to
account for these findings, arguing that the happiness–success link exists not only because success makes
people happy, but also because positive affect engenders success. Three classes of evidence— cross-
sectional, longitudinal, and experimental—are documented to test their model. Relevant studies are
described and their effect sizes combined meta-analytically. The results reveal that happiness is associ-
ated with and precedes numerous successful outcomes, as well as behaviors paralleling success.
Furthermore, the evidence suggests that positive affect—the hallmark of well-being—may be the cause
of many of the desirable characteristics, resources, and successes correlated with happiness. Limitations,
empirical issues, and important future research questions are discussed.
Keywords: happiness, subjective well-being, positive affect, positive emotions, meta-analysis
“A merry heart goes all the day, Your sad tires in a mile-a.”
—William Shakespeare
“The joyfulness of a man prolongeth his days.”
—Sirach 30:22
“The days that make us happy make us wise.”
—John Masefield
Research on well-being consistently reveals that the character-
istics and resources valued by society correlate with happiness. For
example, marriage (Mastekaasa, 1994), a comfortable income
(Diener & Biswas-Diener, 2002), superior mental health
(Koivumaa-Honkanen et al., 2004), and a long life (Danner, Snow-
don, & Friesen, 2001) all covary with reports of high happiness
levels. Such associations between desirable life outcomes and
happiness have led most investigators to assume that success
makes people happy. This assumption can be found throughout the
literature in this area. For example, Diener, Suh, Lucas, and Smith
(1999) reviewed the correlations between happiness and a variety
of resources, desirable characteristics, and favorable life circum-
stances. Although the authors recognized that the causality can be
bidirectional, they frequently used wording implying that cause
flows from the resource to happiness. For example, they suggested
that marriage might have “greater benefits for men than for
women” (p. 290), apparently overlooking the possibility that sex
differences in marital patterns could be due to differential selection
into marriage based on well-being. Similarly, after reviewing links
between money and well-being, Diener and his colleagues pointed
out that “even when extremely wealthy individuals are examined,
the effects [italics added] of income are small” (p. 287), again
assuming a causal direction from income to happiness. We use
quotes from one of us to avoid pointing fingers at others, but such
examples could be garnered from the majority of scientific publi-
cations in this area. The quotes underscore the pervasiveness of the
assumption among well-being investigators that successful out-
comes foster happiness. The purpose of our review is not to
disconfirm that resources and success lead to well-being—a notion
that is likely valid to some degree. Our aim is to show that the
alternative causal pathway—that happy people are likely to ac-
quire favorable life circumstances—is at least partly responsible
for the associations found in the literature.
A PRELIMINARY CONCEPTUAL MODEL
In this article, we review evidence suggesting that happy peo-
ple—those who experience a preponderance of positive emo-
tions—tend to be successful and accomplished across multiple life
domains. Why is happiness linked to successful outcomes? We
propose that this is not merely because success leads to happiness,
but because positive affect (PA) engenders success. Positively
Sonja Lyubomirsky, Department of Psychology, University of Califor-
nia, Riverside; Laura King, Department of Psychological Sciences, Uni-
versity of Missouri—Columbia; Ed Diener, Department of Psychology,
University of Illinois at Urbana–Champaign and The Gallup Organization,
Omaha, Nebraska.
This work was supported in part by grants from the Positive Psychology
Network. We are grateful to Fazilet Kasri, Rene Dickerhoof, Colleen
Howell, Angela Zamora, Stephen Schueller, Irene Chung, Kathleen Jamir,
Tony Angelo, and Christie Scollon for conducting library research and
especially to Ryan Howell for statistical consulting.
Correspondence concerning this article should be addressed to Sonja
Lyubomirsky, Department of Psychology, University of California, River-
side, CA 92521. E-mail: sonja@citrus.ucr.edu
Psychological Bulletin Copyright 2005 by the American Psychological Association
2005, Vol. 131, No. 6, 803– 855 0033-2909/05/$12.00 DOI: 10.1037/0033-2909.131.6.803
803
valenced moods and emotions lead people to think, feel, and act in
ways that promote both resource building and involvement with
approach goals (Elliot & Thrash, 2002; Lyubomirsky, 2001). An
individual experiencing a positive mood or emotion is encounter-
ing circumstances that he or she interprets as desirable. Positive
emotions signify that life is going well, the person’s goals are
being met, and resources are adequate (e.g., Cantor et al., 1991;
Carver & Scheier, 1998; Clore, Wyer, Dienes, Gasper, & Isbell,
2001). In these circumstances, as Fredrickson (1998, 2001) has so
lucidly described, people are ideally situated to “broaden and
build.” In other words, because all is going well, individuals can
expand their resources and friendships; they can take the oppor-
tunity to build their repertoire of skills for future use; or they can
rest and relax to rebuild their energy after expending high levels of
effort. Fredrickson’s model (Fredrickson, 2001) suggests that a
critical adaptive purpose of positive emotions is to help prepare the
organism for future challenges. Following Fredrickson, we suggest
that people experiencing positive emotions take advantage of their
time in this state—free from immediate danger and unmarked by
recent loss—to seek new goals that they have not yet attained (see
Carver, 2003, for a related review).
The characteristics related to positive affect include confidence,
optimism, and self-efficacy; likability and positive construals of
others; sociability, activity, and energy; prosocial behavior; immu-
nity and physical well-being; effective coping with challenge and
stress; and originality and flexibility. What these attributes share is
that they all encourage active involvement with goal pursuits and
with the environment. When all is going well, a person is not well
served by withdrawing into a self-protective stance in which the
primary aim is to protect his or her existing resources and to avoid
harm—a process marking the experience of negative emotions.
Positive emotions produce the tendency to approach rather than to
avoid and to prepare the individual to seek out and undertake new
goals. Thus, we propose that the success of happy people rests on
two main factors. First, because happy people experience frequent
positive moods, they have a greater likelihood of working actively
toward new goals while experiencing those moods. Second, happy
people are in possession of past skills and resources, which they
have built over time during previous pleasant moods.
This unifying framework builds on several earlier bodies of
work—the broaden-and-build model of positive emotions
(Fredrickson, 1998, 2001), the notion that positive emotions con-
vey specific information to the person (Ortony, Clore, & Collins,
1988), the idea of positivity offset (Ito & Cacioppo, 1999), work
on the approach-related aspects of PA (Watson, 2000), and, fi-
nally, Isen’s (e.g., 2000) groundbreaking research on the behaviors
that follow positive mood inductions. We extend the earlier work
in predicting that chronically happy people are in general more
successful, and that their success is in large part a consequence of
their happiness and frequent experience of PA. Although the vast
majority of research on emotions has been on negative states, a
body of literature has now accumulated that highlights the impor-
tance of positive emotions in people’s long-term flourishing.
Classes of Evidence
Figure 1 displays our general conceptual model, which proposes
that successful outcomes are caused by happiness and do not
merely correlate with it or follow from it. Specifically, below the
conceptual model, we display four classes of evidence that can be
used to test it. The first type of evidence (Type A) represents
positive correlations derived from cross-sectional studies. Al-
though it is a truism that correlation does not imply causation,
correlations must generally be positive to be consistent with prop-
ositions about causality. Except in the rare case in which strong
third-variable suppressor effects exist across studies, an absence of
correlation between two variables indicates an absence of causality
in either direction. Thus, correlational evidence is germane to our
argument because the absence of positive correlations suggests
that happiness does not cause success.
The second class of evidence (Type B) is based on longitudinal
research, and is somewhat more informative about causal direction
than cross-sectional correlations. If one variable precedes another
in time and other potential causal variables are statistically con-
trolled, the resulting causal model can be used to reject a causal
hypothesis. In cases in which changes in variable X are shown to
precede changes in variable Y, this form of evidence is even more
strongly supportive of a causal connection, although the influence
of third variables might still contaminate the conclusions and leave
the direction of cause in doubt. Evidence of Type C, the classic
laboratory experiment, is commonly believed to represent the
strongest evidence for causality, although even in this case it can
be difficult to determine exactly what aspect of the experimental
manipulation led to changes in the dependent variable. Finally,
long-term experimental intervention studies (Type D evidence)
would offer the strongest test of our causal model, although again
the active ingredients in the causal chain are usually not known
with certainty.
Empirical Tests of Model and Organizational Strategy
Because no single study or type of evidence is definitive, an
argument for causality can best be made when various classes of
evidence all converge on the same conclusion. Therefore, we
document several types of evidence in our article in order to most
rigorously test the idea that happiness leads to success. Our review
covers the first three classes of evidence (Types A, B, and C) and
is organized around five focal questions arising from these three
categories:
1. Cross-sectional studies (Type A)
Question 1: Are happy people successful people?
Question 2: Are long-term happiness and short-term
PA associated with behaviors paralleling success—
that is, with adaptive characteristics and skills?
2. Longitudinal studies (Type B)
Question 3: Does happiness precede success?
Question 4: Do happiness and positive affect precede
behaviors paralleling success?
3. Experimental studies (Type C)
Question 5: Does positive affect lead to behaviors
paralleling success?
First, we document the extensive cross-sectional correlational
evidence (Type A), as shown in Figure 1. The first question
addressed by this evidence is the one that forms the basis of our
causal hypothesis—that is, are happy people more likely to suc-
804
LYUBOMIRSKY, KING, AND DIENER
ceed at culturally valued goals (e.g., concerning work, love, and
health) than their less happy peers? However, the large number of
available correlational studies in this category also includes rele-
vant research examining behavior and cognition that parallel suc-
cessful life outcomes—that is, the characteristics, resources, and
skills that help people succeed (e.g., attributes such as self-
efficacy, creativity, sociability, altruism, immunity, and coping).
Accordingly, the second question addressed by this evidence ex-
plores the relations of behavior paralleling success to long-term
happiness and short-term PA. Because we define happiness as the
Figure 1. Empirically testing the conceptual model. PA ⫽ positive affect; Grp. ⫽ group.
805
BENEFITS OF FREQUENT POSITIVE AFFECT
frequent experience of positive emotions over time (see below),
our model assumes that the correlations involving long-term hap-
piness are parallel to those of short-term positive moods. In con-
clusion, only if the correlations generated by Questions 1 and 2 are
generally positive will we consider our causal hypothesis further.
Second, we consider longitudinal studies, which address two
further questions. Is happiness at Time 1 associated with success-
ful outcomes at Time 2 (Question 3)? Is happiness and PA at Time
1 correlated with behaviors paralleling success at Time 2 (Ques-
tion 4)? In summary, prior levels of happiness and positive affect
must correlate with later levels of successful outcomes and behav-
ior for our causal hypothesis not to be rejected.
In laboratory experimentation, the third type of evidence, cau-
sality is put to a stronger test. In this case, however, because of the
limits of the laboratory, only short-term changes in behavior and
cognitions that parallel successful life outcomes are assessed.
Thus, the fifth and final question we address is whether PA causes
the cognitive and behavioral characteristics paralleling success.
Again, because positive affect is defined here as the basic constit-
uent of happiness, our model requires that the outcomes of short-
term positive moods are parallel to the successful outcomes in our
conceptual model. Furthermore, this question is critical, as it
speaks to whether PA may be a mediator underlying the relation-
ship between happiness and flourishing—that is, whether PA
causes the adaptive characteristics that help happy people succeed.
Although the fourth type of evidence shown in Figure 1 (Type
D) would provide the strongest type of data for our model, unfor-
tunately, to our knowledge no studies of this type exist. Neverthe-
less, support for our conceptual model from all three of the
previously described types of evidence, while not definitive, will
suggest a likelihood that our causal model is correct. Furthermore,
combining the three types of evidence represents an advance
beyond laboratory experimentation alone, because the relatively
greater rigor and control provided by experimentation are supple-
mented by the relatively greater ecological validity provided by the
other types of studies. Thus, the first two classes of evidence
(Types A and B) speak to the plausibility of generalizing the causal
laboratory findings to the context of success and thriving in ev-
eryday life. Meanwhile, by revealing the processes uncovered in
the laboratory, the experimental evidence (Type C) illuminates the
possible causal sequence suspected in the correlational data. Taken
together, consistent findings from all three types of data offer a
stronger test than any single type of data taken alone.
After describing our methodology and defining our terms, we
address each of the five focal questions in order, documenting the
three classes (A, B, and C) of relevant empirical evidence. Then,
we turn to a discussion of several intriguing issues and questions
arising out of this review, caveats and limitations, and important
further research questions.
Methodological Approach
To identify the widest range of published papers and disserta-
tions, we used several search strategies (Cooper, 1998). First, we
searched the PsycINFO online database, using a variety of key
words (e.g., happiness, satisfaction, affect, emotion, and mood).
Next, using the ancestry method, the reference list of every em-
pirical, theoretical, and review paper and chapter was further
combed for additional relevant articles. To obtain any papers that
might have been overlooked by our search criteria, as well as to
locate work that is unpublished or in press, we contacted two large
electronic listserves, many of whose members conduct research in
the area of well-being and emotion—the Society of Personality
and Social Psychology listserv and the Quality of Life Studies
listserv. Twenty-four additional relevant articles were identified
with this method.
The final body of literature was composed of 225 papers, of
which 11 are unpublished or dissertations. From these 225 papers,
we examined 293 samples, comprising over 275,000 participants,
and computed 313 independent effect sizes. A study was included
in our tables if it satisfied the following criteria. First, measures of
happiness, PA, or a closely related construct had to be included, in
addition to assessment of at least one outcome, characteristic,
resource, skill, or behavior. Second, the data had to include either
a zero-order correlation coefficient or information that could be
converted to an r effect size (e.g., t tests, F tests, means and
standard deviations, and chi-squares). If a study did not report an
r effect size, we computed one from descriptive statistics, t statis-
tics, F ratios, and tables of counts (see Rosenthal, 1991). If no
relevant convertible statistics were presented, other than a p value,
we calculated the t statistic from the p value and an
r-sub(equivalent) (Rosenthal & Rubin, 2003). When a paper re-
ported p ⬍ .05, p ⬍ .01, or ns, we computed rsub(equivalent) with
p values of .0245, .005, and .50 (one-tailed), respectively, which
likely yielded a highly conservative estimate of the effect size.
Finally, the sample size had to be available. When possible, we
also contacted authors for further information.
Descriptions of the critical elements of each study (i.e., authors,
year, sample size, happiness/PA measure or induction, related
construct, and effect size [r]) are included in Tables 1, 2, and 3,
which present cross-sectional, longitudinal, and experimental
work, respectively. Table 2 additionally presents the length of time
between assessments, and Table 3 includes the comparison groups
used in the studies. Studies with subscripts after their name are
those that appear in more than a single section or table, usually
because multiple outcome variables are included.
Furthermore, mirroring our documentation of the literature pre-
sented in this paper, Tables 1–3 are subdivided into substantive
categories (or panels). For example, Table 1 is subdivided into
nine categories—work life, social relationships, health, percep-
tions of self and others, sociability and activity, likability and
cooperation, prosocial behavior, physical well-being and coping,
and, finally, problem solving and creativity. The mean and median
effect size (r), weighted and unweighted by sample size, as well as
a test of heterogeneity, is provided for each category for the three
classes of data (cross-sectional, longitudinal, and experimental) in
Table 4.
Tables 1, 2, and 3 report all effect sizes of interest to readers—
including instances of two or more effect sizes generated from the
same sample or dataset. For example, the relation of happiness
with income and marital status derived from a single study may
appear in two different panels of a table (i.e., work life and social
relationships). Alternatively, the correlation between happiness
and coping derived from a single longitudinal study may appear in
two different tables (e.g., the cross-sectional table and the longi-
tudinal table). However, in order to meta-analytically combine the
464 effect sizes listed in Tables 1–3, we had to ensure a degree of
(text continues on page 816)
806
LYUBOMIRSKY, KING, AND DIENER
Table 1
Study Information and Effect Sizes for Nine Categories of Cross-Sectional Research
Study n Happiness/PA measure Correlated construct
Effect size
(r)
Work life
Crede´ et al., 2005 959 PANAS Organizational citizenship behavior .37
Crede´ et al., 2005 959 PANAS Counterproductive work behavior ⫺.25
Crede´ et al., 2005 959 PANAS Job withdrawal ⫺.25
Cropanzano & Wright, 1999
a
(first assessment)
60 Index of Psychological Well-Being Supervisory evaluations .29
Cropanzano & Wright, 1999
a
(second assessment)
60 Index of Psychological Well-Being Supervisory evaluations .34
DeLuga & Mason, 2000 92 Affectometer 2 Job performance .22
Donovan, 2000 188 Current Mood Report Organizational citizenship behavior .20
Donovan, 2000 188 Current Mood Report Turnover intentions ⫺.38
Donovan, 2000 188 Current Mood Report Work withdrawal ⫺.20
Donovan, 2000 188 Current Mood Report Organizational retaliatory behavior ⫺.22
Donovan, 2000 188 Current Mood Report Satisfaction with work .50
Foster et al., 2004 41 Job Affect Scale Organizational climate for performance .32
Foster et al., 2004 41 Job Affect Scale Employee health and well-being .29
Frisch et al., 2004 3,638 Quality of Life Inventory Academic retention absenteeism .18
George, 1989 254 Job Affect Scale ⫺.28
George, 1995 53 PANAS (leader) Judged customer service .41
George, 1995 53 PANAS (aggregated group) Judged customer service .35
Graham et al., in press
a
(1995 assessment)
4,524 One-item happiness Income .20
b
Graham et al., in press
a
(2000 assessment)
5,134 One-item happiness Income .16
b
Howell et al., in press 307 SWLS Material wealth .23
Jundt & Hinsz, 2001 164 Seven-point semantic differentials Task performance .19
Krueger et al., 2001
a
397 MPQ positive emotionality Self-reported altruism .44
Lucas et al., 2004 24,000 One-item happiness Income .20
Magen & Aharoni, 1991
a
260 Four-item positive affect Transpersonal commitment .21
Magen & Aharoni, 1991
a
260 Four-item positive affect Involvement in community service .36
Miles et al., 2002 203 Job-Related Affective Well-Being
Scale
Organizational citizenship behavior .23
Seligman & Schulman, 1986
a
(Study 1)
94 Attributional Style Questionnaire Quarterly insurance commissions .18
Staw & Barsade, 1993
a
83 Three-measure composite of positive
affectivity
Judged managerial performance .20
Staw et al., 1994
a
272 Experience and expression of
positive emotion on the job
Job autonomy, meaning, and variety .22
Staw et al., 1994
a
272 Experience and expression of
positive emotion on the job
Gross annual salary .12
Staw et al., 1994
a
272 Experience and expression of
positive emotion on the job
Supervisory evaluations (creativity) .30
Thoits & Hewitt, 2001
a
3,617 One-item happiness Time spent volunteering .09
Totterdell, 2000* 17 One-item happiness (12 times over
4 days)
Cricket batting average .36
Van Katwyk et al., 2000
a
(Study 3)
111 PANAS Interpersonal conflict ⫺.12
Van Katwyk et al., 2000
a
(Study 3)
111 PANAS Intention to quit ⫺.33
Weiss et al., 1999
a
24 Fordyce HM Scale Job satisfaction .29
Wright & Cropanzano, 1998 52 PANAS Emotional exhaustion ⫺.39
Wright & Cropanzano, 2000 47 Index of Psychological Well-Being Job performance .32
(Study 1)
Wright & Cropanzano, 2000 (Study 2) 37 Index of Psychological Well-Being Supervisory evaluations .34
Wright & Staw, 1999
a
(Study 1,
second assessment)
45 Index of Psychological Well-Being Supervisory evaluations .33
Wright & Staw, 1999
a
(Study 2,
first assessment)
62 Index of Psychological Well-Being Supervisory evaluations .25
Wright & Staw, 1999
a
(Study 2,
second assessment)
64 Index of Psychological Well-Being Supervisory evaluations .43
Social relationships
Baldassare et al., 1984 202 Four-item happiness Instrumental support .17
Baldassare et al., 1984 202 Four-item happiness Emotional support .15
Baldassare et al., 1984 202 Four-item happiness Companionship .30
Berry & Willingham, 1997 127 PANAS Commitment to current relationship .27
Cooper et al., 1992
a
(Study 1 & Study 2)
118 SWLS Satisfaction with friends .31
Cooper et al., 1992
a
(Study 2)
118 SWLS Satisfaction with social activities .37
(table continues)
807
BENEFITS OF FREQUENT POSITIVE AFFECT
