ArticlePDF Available

The Relation of Economic Status to Subjective Well-Being in Developing Countries: A Meta-Analysis

Authors:

Abstract and Figures

The current research synthesis integrates the findings of 111 independent samples from 54 economically developing countries that examined the relation between economic status and subjective well-being (SWB). The average economic status-SWB effect size was strongest among low-income developing economies (r = .28) and for samples that were least educated (r = .36). The relation was weakest among high-income developing economies (r = .10) and for highly educated samples (r = .13). Controlling for numerous covariates, the partial r effect size remained significant for the least-educated samples (pr = .18). Moderator analyses showed the economic status-SWB relation to be strongest when (a) economic status was defined as wealth (a stock variable), instead of as income (a flow variable), and (b) SWB was measured as life satisfaction (a cognitive assessment), instead of as happiness (an emotional assessment). Findings were replicated with a meta-analysis of the World Values Survey data. Discussion centers on the plausibility of need theory, alternative explanations of results, interpretation of moderators, and directions for future research.
Content may be subject to copyright.
The Relation of Economic Status to Subjective Well-Being in Developing
Countries: A Meta-Analysis
Ryan T. Howell
San Francisco State University
Colleen J. Howell
University of California, Riverside
The current research synthesis integrates the findings of 111 independent samples from 54 economically
developing countries that examined the relation between economic status and subjective well-being
(SWB). The average economic status–SWB effect size was strongest among low-income developing
economies (r .28) and for samples that were least educated (r .36). The relation was weakest among
high-income developing economies (r .10) and for highly educated samples (r .13). Controlling for
numerous covariates, the partial r effect size remained significant for the least-educated samples ( pr
.18). Moderator analyses showed the economic status–SWB relation to be strongest when (a) economic
status was defined as wealth (a stock variable), instead of as income (a flow variable), and (b) SWB was
measured as life satisfaction (a cognitive assessment), instead of as happiness (an emotional assessment).
Findings were replicated with a meta-analysis of the World Values Survey data. Discussion centers on
the plausibility of need theory, alternative explanations of results, interpretation of moderators, and
directions for future research.
Keywords: poverty, income, subjective well-being, happiness, need theory
A feast is made for laughter, and wine makes life merry, but money
is the answer for everything.
—Ecclesiastes 10:19 (c. 300 BCE)
For thousands of years, people have believed that money could
buy happiness, and over the past 5 decades scholars have investi-
gated the extent to which this might be true. Survey research
consistently has demonstrated that the rich are typically happier
than the poor, and individuals living in wealthier economies are,
on average, happier than those living in poorer economies (Cantril,
1965; Diener, Sandvik, Seidlitz, & Diener, 1993; Veenhoven,
1991, 1994; Zavisca & Hout, 2005). This should not be surprising
given that higher economic status is associated with several desir-
able outcomes, such as increased life expectancy, reduced malnu-
trition, and lower infant mortality (Gillis, Perkins, Roemer, &
Snodgrass, 1996). Poverty, on the other hand, correlates with
“poor health, poor mobility, poor education, and poor access to
services” (Klasen, 1997, p. 89). Further, economic utility theory
suggests that as people gain income and wealth, they gain pur-
chasing power, which expands their bundle of affordable goods,
leading to increased consumption and, ultimately, to improved
utility or well-being.
Based on the common finding that financial resources associate
positively with well-being and/or welfare, one might expect a
strong positive correlation between income or wealth and life
satisfaction or happiness in life. Indeed, cross-national analyses of
countries’ average happiness and gross domestic product per cap-
ita have shown the correlation to be quite strong (r .60 to .84;
Diener, Diener, & Diener, 1995; Schyns, 1998; Veenhoven, 1991).
Yet, in cross-sectional analyses, correlations between wealth or
income and happiness tend to vary according to the economic
status of the country or sample (e.g., Diener & Biswas-Diener,
2002; Diener & Oishi, 2000; Diener et al., 1993; Diener, Suh,
Lucas, & Smith, 1999; Schyns, 1998, 2002; Veenhoven, 1991,
1996). Namely, much attention has been paid to the rather weak
economic status–subjective well-being (SWB) relations that tend
to be found within wealthy developed nations, such as the United
States, Australia, and countries within Western Europe (r .06 to
.15, r
2
.004 to .022; see Ahuvia & Friedman, 1998; Cummins,
2000; Diener & Oishi, 2000; Diener et al., 1993; Easterlin, 1995,
2001; Headey, Muffels, & Wooden, 2004; Rojas, 2004). In addi-
tion, while there is general agreement among researchers that
economic status–SWB relations are stronger within poorer and less
developed country samples (e.g., Arthad-Day & Near, 2005; East-
erlin, 1995; Veenhoven, 1991), no study has summarized the
relationship of economic status to SWB in developing countries,
and little is known of the sociodemographic or measurement-
related variables that may influence its strength.
These observations raise the following question, which we have
undertaken to answer through this meta-analysis: What is the
average relation between objective economic status (i.e., absolute
income, wealth, socioeconomic status [SES], etc.) and SWB
within developing economies, and is this correlation statistically
stronger than what has been reported in developed economies? The
second aim of the study is to test for moderators, to determine the
Ryan T. Howell, Department of Psychology, San Francisco State Uni-
versity; Colleen J. Howell, Department of Environmental Science, Univer-
sity of California, Riverside.
We are thankful for the numerous comments on drafts of this article
offered by Carol Graham, Ezequiel Morsella, Tanya Boone, Alex Wood,
and Sonja Lyubomirsky. We are also grateful to Tom and Sue James who
helped with editing, and to Katrina Rodzon and Thery Prok for assistance
with preparing the manuscript.
Correspondence concerning this article should be addressed to Ryan
Howell, Department of Psychology, San Francisco State University, 1600
Holloway Avenue, San Francisco, CA 94132. E-mail: rhowell@sfsu.edu
Psychological Bulletin Copyright 2008 by the American Psychological Association
2008, Vol. 134, No. 4, 536–560 0033-2909/08/$12.00 DOI: 10.1037/0033-2909.134.4.536
536
sample-specific factors that affect the economic status–SWB rela-
tion within developing countries.
Defining Economic Status and SWB in Economically
Developing Countries
Economic Status
We have chosen “economic status” as an overarching term to
refer to objective income (a flow variable) and objective wealth
1
(a
stock variable). Differences in household structure and the preva-
lence of unconventional income-earning activities within develop-
ing country samples (see Graham, 2005; Smith, 2003; Sta˜nculescu
et al., 2005; Tiliouine, Cummins, & Davern, 2006) may lead
researchers to choose different measures of economic status to
approximate objective income, assets, or wealth. These measures
may include personal income (Cheung & Leung, 2004; Tiliouine et
al., 2006), household income (Hayo, 2003; Yip et al., 2006), per
capita household income (Graham, Eggers, & Sukhtankar, 2004;
Smith, 2003), number of household or farm assets owned (Brink-
erhoff, Fredell, & Frideres, 1997; Graham & Pettinato, 2001), total
value of household assets (Howell, Howell, & Schwabe, 2006),
objective SES (Gitmez & Morcol, 1994; Lever, 2000), or house-
hold savings (Brinkerhoff et al., 1997; Howell et al., 2006).
SWB
The current project is concerned with what most researchers
term subjective well-being (SWB). SWB is understood broadly to
include both transient emotional phenomena (e.g., pleasant and
unpleasant affective experiences) as well as more enduring assess-
ments of life satisfaction (Cummins, 2000; Diener, 2000; Diener et
al., 1999; Moore, Leslie, & Lavis, 2005). SWB encompasses the
constructs of happiness, affect balance (the frequency of positive
vs. negative emotions; Bradburn, 1969), overall life satisfaction,
and domain satisfaction
2
(Arthaud-Day & Near, 2005; Diener,
2000; Diener et al., 1999). Happiness and affect balance are
distinguished as emotional appraisals that may be influenced by a
respondent’s present mood (Diener et al., 1999), while evaluations
of life satisfaction involve more cognitive assessments of the
quality of life as a whole and are generally independent of one’s
current emotional state (Diener, Emmons, Larsen, & Griffin, 1985;
Heller, Watson, & Ilies, 2004). Although SWB facets are theoret-
ically distinct (Diener et al., 1999), measures of life satisfaction,
happiness, and affect (both state and trait measures) are positively
correlated (r .48 –.66; Watson & Clark, 1994; also see Diener et
al., 1999; Tsou, & Liu, 2001). Because high ratings of happiness,
life satisfaction, and positive affect all indicate high levels of
subjective well-being, we will treat each construct as a variant of
subjective well-being and will use SWB as a generic term to
describe any facet of subjective well-being.
Understanding the Economic Status–SWB Relation
In the mid-1900s, Cantril (1965) pioneered a project to elicit the
aspirations of individuals living in various socioeconomic situa-
tions from countries in different phases of development. After 6
years of interviewing over 20,000 people in 13 nations, the data
pointed overwhelmingly in one direction: People perceived that
more money, more possessions, and an improved quality of life
would make them happier. Regardless of whether an individual
was from India making $13 a month or from the U.S. making
$8,000 a month, Cantril found that material prosperity was desired
by people of all economic circumstances.
Although individuals continue to express monetary wishes and
material aspirations regardless of economic standing (see Lever,
2000; Li et al., 1998), the relationship between economic status
and SWB is actually complex. First, impressive economic growth
within developed economies, at the country level, is typically not
associated with large increases in average SWB (Diener & Suh,
1997; Easterlin, 2001; see Frijters, Shields, & Haisken-DeNew,
2004, for an exception). Further, past research has found that the
economic status–SWB correlation tends to adhere to the economic
law of diminishing marginal utility: As income rises, each addi-
tional dollar contributes less additional satisfaction. This concave
3
effect between economic status and SWB has been replicated in
several cross-national studies (Diener et al., 1993; Schyns, 2002;
Veenhoven, 1991, 1994; Zavisca & Hout, 2005), across samples
varying by income and wealth both within developed economies
(Cummins, 2000) and developing economies (Graham & Pettinato,
2002), and within developing countries controlling for society and
institutions (Møller & Saris, 2001). In these studies, the strongest
economic status–SWB correlations tend to be reported in poor
developing countries, while the weakest correlations tend to be
demonstrated in the wealthier developed countries.
Psychologists, sociologists, and economists have proposed two
main theories to explain the two sides of this phenomenon. East-
erlin’s paradox—the counterintuitive finding that the economic
status–SWB relation is weak among developed country samples
and that SWB fails to rise with rising incomes—is explained by the
fact that, for wealthier individuals, SWB is impacted by relative
economic status (via comparisons with one’s past economic status,
others’ economic status, or aspirations for higher economic status)
rather than absolute economic status (Easterlin, 1974, 1995, 2001).
1
For ease of discussion, income will be used to refer to all measures of
objective household or personal income as well as to various proxies of
objective income (e.g., expenditures) that represent the flow of money or
liquid assets to the individual or household. Wealth will be used to refer to
all measures and proxies of objective wealth (e.g., savings, number or
value of durable household goods, objective socioeconomic status) that
represent a stock of monetary or capital resources.
2
In addition to these conventional constituents of SWB, Ryff (1989)
suggests that psychological well-being includes several components of
positive functioning (e.g., autonomy, environmental mastery, personal
growth, positive relations with others, purpose in life, and self-acceptance),
which may be neglected by the more frequently used hedonic measures of
affect and life satisfaction (also see Ryff & Keyes, 1995). Three of these
facets mirror the three basic psychological needs (autonomy, competence,
and relatedness) proposed by self-determination theory (Ryan & Deci,
2000). Similarly, Peterson and Seligman (2004) developed the Values in
Action (VIA) Signature Strengths Questionnaire, which measures 24 char-
acter strengths, some of which are associated with life satisfaction (Park,
Peterson, & Seligman, 2004). Although these eudemonic conceptualiza-
tions highlight important facets of well-being, they have rarely (if ever)
been used within studies examining the association between economic
status and SWB in developing countries.
3
Concave, here and elsewhere, implies that the correlation coefficient
between income and well-being decreases as income increases. This phe-
nomenon may also be called a curvilinear effect.
537
ECONOMIC STATUS AND SWB: A META-ANALYSIS
Easterlin posited that absolute income and assets strongly predict
SWB only until a “population is freed from subsistence level needs
for food, clothing and shelter” (Easterlin, 2001, p. 40). This posi-
tion, known as need theory, is often used to explain the stronger
economic status–SWB correlations often observed among poorer
samples.
Need Theory
A number of researchers have cited need theory to explain both
the strong economic status–SWB relation at lower income levels
as well the diminishing marginal effect of economic status on
SWB as income and wealth rise (e.g., Ahuvia, 2002; Arthaud-Day
& Near, 2005; Biswas-Diener & Diener, 2001; Brinkerhoff, Fre-
dell, & Frideres, 1997; Diener & Biswas-Diener, 2002; Diener &
Lucas, 2000; Diener et al., 1993; Frey & Stutzer, 2000; Fuentes &
Rojas, 2001; Lever, 2004; Oishi, Diener, Lucas, & Suh, 1999;
Schyns, 1998; Suhail & Chaudhry, 2004; Veenhoven, 1991; Za-
visca & Hout, 2005). As currently proposed, need theory posits
that income and assets have their strongest influence on SWB
when they are able to satisfy the most basic of physiological needs,
such as sufficient food, proper nutrition, clothing, sanitation, and
shelter (see Diener & Biswas-Diener, 2002; Diener & Lucas, 2000;
Møller & Schlemmer, 1983; Veenhoven, 1991). Once these needs
are fulfilled, additional economic resources directly affect SWB to
a lesser degree—this, again, is the Easterlin Paradox—perhaps
because higher-order needs are generally non-material (e.g., be-
longingness, love, esteem, self-actualization; see Arthaud-Day &
Near, 2005; Diener & Diener, 1995; Diener, Oishi, & Lucas, 2003;
Lever, Pinol, & Uralde, 2005). Another possible explanation for
this phenomenon “is that norms and expectations adapt upward at
about the same rate as income increases, and thus after basic needs
are met, more income does not make people happier” (Graham,
2005, p. 206).
Past studies have claimed support for need theory by using two
main research approaches. The most prevalent approach involves
comparing economic status–SWB correlations for wealthier and
poorer samples or nations. While not a robust test to prove need
theory, the logic behind correlation comparisons assumes that
people who fall into a category of low economic status are the least
likely to have their basic needs met and, thus, are most likely to
experience substantial gains in SWB as those needs become sat-
isfied via additional income or wealth (Camfield, Choudhury, &
Devine, 2006; Cummins, 2000; Diener et al., 1999; Easterlin,
1995; Fuentes & Rojas, 2001; Graham, 2005; Lever et al., 2005;
Royo & Velazco, 2006; Smith, 2003; Suhail & Chaudhry, 2004;
Veenhoven, 1991). Therefore, in a cross-sectional analysis, one
would expect to find a strong positive economic status–SWB
correlation for samples or subsamples comprising poor individuals
that live on or around the threshold of basic need satisfaction. For
samples or subsamples with relatively high economic status—and
for whom there is no concern over the ability to meet basic
needs— one would expect a weak relation between economic
status and SWB. Results confirming significantly stronger corre-
lations among poorer groups than among wealthier groups have
been cited as support for need theory within Bangladesh (Cam-
field, Choudhury, & Devine, 2006), India (Biswas-Diener & Die-
ner, 2001), Latin America (Graham & Pettinato, 2002), Mexico
(Fuentes & Rojas, 2001; Lever, 2004), Pakistan (Suhail &
Chaudhry, 2004), Russia (Zavisca & Hout, 2005), Korea (Kim,
1998; Lee, Kim, & Shin, 1982), South Africa (Møller & Saris,
2001), and Thailand (Royo & Velazco, 2006), among others (e.g.,
Arthaud-Day & Near, 2005; Cummins, 2000; Diener, Diener, &
Diener, 1995; Diener et al., 1993; Veenhoven, 1991).
Other research has approached support for need theory by
examining SWB as it relates to perceived need fulfillment at
different levels of income or wealth. Several studies demonstrated
that households expressing discontent with food consumption,
housing, hygiene, health, or clothing report significantly lower
SWB, on average, than do households within the same sample
whose basic needs are reported to be satisfied (Ahmed,
Chowdhury, & Bhuiya, 2001; Biswas-Diener & Diener, 2001;
Brinkerhoff et al., 1997; Camfield, Choudhury, & Devine, 2006;
Fuentes & Rojas, 2001; Gitmez & Morcol, 1994; Royo & Velazco,
2006). Further, poorer samples in developing economies define
SWB as including basic physical need satisfaction (e.g., food
intake, housing security), whereas wealthier respondents in these
same countries stress more non-material or higher-level needs
(e.g., autonomy, personal security, peace of mind, and status;
Cummins, 2000; Mahmuda, 2003; Oishi et al., 1999). Finally, in a
cross-national study of 39 countries, Diener, Diener, and Diener
(1995) reported that average national SWB correlates significantly
with the degree to which basic needs are met for the majority of
citizens.
Although individuals for whom basic needs are not fulfilled can
certainly be found in any type of society (developing or developed;
Cummins, 2000), there is agreement among economists, sociolo-
gists, and psychologists that “if there is a threshold beyond which
more money does not increase average levels of enhanced reported
well-being, most developing economies have not yet crossed it”
(Graham, 2005, p. 208; see also Argyle, 1999; Easterlin, 1995).
Alternatively, “wealthier nations tend to uniformly meet the phys-
ical needs of virtually all of their citizens” (Diener, Diener, &
Diener, 1995, p. 852; Biswas-Diener, Vitterso, & Diener, 2005;
Kenny, 2005). Easterlin (2001) argued that because individuals in
wealthy countries compare their economic standing with that of
their peers—a practice that is relatively less common in poorer
countries— being poor in a developing economy is very different
from being poor in a wealthy country (at least when examining the
causes and predictors of well-being). For example, Smith (2003)
demonstrated that Americans who are poor report significantly
higher levels of happiness when compared with poor individuals
from the former Soviet Union even after controlling for several
sociodemographic variables. Thus, due to the prevalence of pov-
erty within developing nations (Sachs, 2005; World Bank, 2007),
and due to the hypothesis that samples’ absolute incomes are a
stronger predictor of well-being in poor developing economies
than in wealthy developed economies, an assessment of the rela-
tionship between objective economic status and SWB for individ-
uals at or around the threshold of basic needs is best undertaken by
focusing on samples residing within developing countries.
Past Research Syntheses Examining the Economic Status–
SWB Relation in Poverty
Few research syntheses have attempted to confirm whether
economic status–SWB correlations are significantly stronger
within economically developing countries than within economi-
538
HOWELL AND HOWELL
cally developed countries. Cummins (2000) meta-analytically
combined effect sizes from 25 studies and reported the personal
income–SWB relation for low-income samples (r .26, k 9) to
be stronger than that for high-income samples (r .14, k 24).
This synthesis is limited in its generalizability to the extent that it
focused on low-income samples in economically developed coun-
tries (with the exception of one sample from South Africa), fo-
cused on only a single measure of income, and circumscribed its
search to articles filed within the Australian Centre on Quality of
Life (approximately 3,000 articles). Arthaud-Day and Near (2005)
conducted a qualitative review of the absolute income–SWB lit-
erature and concluded that the income–SWB relation is positive
and statistically significant— but weak. They state that the litera-
ture supported the assumption that income–SWB relations within
developing countries are stronger than those observed within de-
veloped countries. However, importantly, their claim demands
statistical validation.
Need for a Meta-Analysis
Despite the contribution of past research in providing a general
estimate of the strength of the economic status–SWB relation
within wealthy developed country samples, there is, to date, no
consensus over the extent to which economic status relates to SWB
among developing country samples. Both aforementioned reviews
supported the hypothesis that the economic status–SWB relation is
stronger in developing economies, yet neither attempted a com-
prehensive synthesis of the current economic status–SWB litera-
ture examining economically developing countries. Further, there
was little (if any) acknowledgement of some cross-disciplinary
research findings that have demonstrated weak, non-significant, or
negative economic status–SWB relations within developing coun-
try samples (e.g., Foley, 2005; Kousha & Mohseni, 2000; Møller,
Dickow, & Harris, 1999; Morawetz et al., 1977; Namazie &
Sanfey, 2001; Seik, 2000; Tan et al., 2006; Tsou & Liu, 2001).
The past 10 –15 years have witnessed a growing body of re-
search from several social science fields (e.g., economics, psychol-
ogy, sociology, urban services) in which the correlation between
economic status and SWB has been investigated within developing
countries. Yet, instead of uniting and organizing researchers in
these various fields to advance the study of economic status and
SWB among disparate populations, the interdisciplinary nature of
this scholarship has led to a rather disconnected body of work.
Indeed, the volume and range of research outlets in which findings
have been presented may result in relatively few “hits” from
subject-specific literature search techniques, thus misleading re-
searchers to conclude that little research has been conducted in this
area (e.g., Biswas-Diener & Diener, 2001; Fuentes & Rojas, 2001;
Tsou & Liu, 2001; Yip et al., 2006). Thus, a quantitative research
synthesis of the entire relevant literature, regardless of academic
discipline or study objective, is imperative for understanding the
economic status–SWB relation, and its moderators, among devel-
oping country samples.
Possible Moderators of the Economic Status–SWB
Relation
Past research within developing countries suggests that hetero-
geneity in the correlations between economic status and SWB may
be due to complex interactions with other sample-level socioeco-
nomic (e.g., country development status), demographic (e.g., ed-
ucation and gender), or geographic variables, as well as to how
certain variables (e.g., income or wealth, SWB) are operational-
ized (Aryee, 1999; Biswas-Diener & Diener, 2001; Brinkerhoff et
al., 1997; Diener et al., 1993; Fuentes & Rojas, 2001; Rojas, 2005;
Zavisca & Hout, 2005). Thus, a complete understanding of the
economic status–SWB relation requires investigating potential
sample-level moderators of the association.
SWB construct. Although researchers may use various SWB
terms interchangeably, the current study predicts that the construct
used to measure SWB may be a significant moderator of the
economic status–SWB relation within developing countries. For
example, because of the emotional foundation of happiness (Die-
ner, 1984), this construct may be more weakly related to income
than might be a more cognitive evaluation of life quality (e.g.,
satisfaction with life, quality of life). Kahneman, Krueger, Sch-
kade, Schwarz, and Stone (2004) argued that the task of evaluating
life satisfaction prompts individuals to focus on the quality of
objective circumstances (e.g., current income, marital status),
whereas evaluations of affective experience (e.g., happiness) are
less likely to prime such considerations. Lee et al. (1982) con-
tended that assessments of life satisfaction involve consistent
judgments of cognitive experiences— one aspect of which may be
income—whereas happiness assessments involve short-lived af-
fective judgments and, thus, can be influenced by transient emo-
tions. A second reason the economic status–happiness relation
may be weaker in developing countries pertains to translational
concerns. In other languages, the English term “happiness” may
have multiple meanings (Fuentes & Rojas, 2001; Oishi et al., 1999;
Padilla & Lindholm, 1995) or may convey an unintended meaning
when translated directly (Zavisca & Hout, 2005). Thus, we expect
a weaker relation between economic status and happiness than
between economic status and life satisfaction.
Economic status construct. The economic status construct em-
ployed in a particular study may also be critical for understanding
how income and assets relate to SWB, as not all economic mea-
sures are equally efficient at explaining variance in SWB. Within
developing country samples, intercorrelations between various
economic status measures (e.g., household income, consumption,
savings, household possessions) tend to be weaker (r .24–.26;
Headey et al., 2004; Howell et al., 2006) than the typical intercor-
relations found between various measures of SWB (r .44 –.72;
Kim, 1998; Lyubomirsky & Lepper, 1999; Suhail & Chaudhry,
2004). Also, separate measures of economic status can explain
unique variance in SWB even after controlling for other income
and asset constructs (Headey et al., 2004).
Single-item measures of personal or household income may fail
to capture important income sources beyond the household’s pri-
mary cash-income-generating activities and, thus, may lead to the
unintentional misreporting or underestimation of household eco-
nomic status (Diener & Biswas-Diener, 2002). For example, re-
mittances, in-kind income (e.g., agriculture, forest products, hunt-
ing, fishing), household production, or periodic income (e.g., odd
jobs, seasonal work) can contribute substantially to the budget of
a poor, rural household (e.g., Dercon, 2002; Howell, 2006). Thus,
the pervasiveness of unconventional income sources and in-kind
payments among poor developing country samples is the impetus
for many researchers in this field to rely on proxies of income and
539
ECONOMIC STATUS AND SWB: A META-ANALYSIS
wealth (Graham, 2005; Klasen, 1997; Smith, 2003; Sta˜nculescu et
al., 2005; Tiliouine et al., 2006).
These unconventional sources, combined with the unpredictable
nature of income for poor households, may result in economic
status estimates that are riddled with inaccuracies and unintended
omissions. Such inaccuracies would lead to increased measure-
ment error, which would then, following attenuation theory (Lord
& Novick, 1968), decrease the estimate of the true correlation
between economic status and SWB within a particular population
(for tests of attenuation for the economic status–SWB relation, see
Howell et al., 2006). Income constructs that are measured least
accurately will incur the greatest measurement error, and thus we
predict that these measures of income will demonstrate the weak-
est associations with SWB.
Sample gender composition. Several studies have proposed
that gender may moderate the economic status–SWB relation.
Adelmann (1987) found the income–happiness correlation for
male respondents to be stronger than for female respondents and
conjectured that men may find greater satisfaction in occupational
activities, while women may derive more happiness from relation-
ships and family. Mahmuda’s (2003) study in Bangladesh corrob-
orated Adelmann’s hypothesis by showing that women’s defini-
tions of SWB emphasize care for family, whereas men’s
definitions center more on income. George and Brief (1990) dem-
onstrated that men with high financial requirements report stronger
income–life satisfaction relations than do women in general. Sim-
ilarly, Aryee (1999) argued that the salience of one’s role as the
household breadwinner—typically male household members
within developing countries—may moderate the relation between
income and life satisfaction. Given these and other findings of
gender-related differences in perspectives on income and wealth
with SWB (e.g., Brinkerhoff et al., 1997; Zavisca & Hout, 2005),
we expect that the gender composition of the sample will moderate
the economic status–SWB relation. Specifically, we predict the
economic status–SWB association to be strongest for predomi-
nately male samples.
Objectives of the Current Study
The current study meta-analyzes past research examining the
economic status–SWB relation among populations living in devel-
oping economies. Our goal is first to estimate the overall r effect
size for developing economies and then to assess whether this
effect size is statistically stronger than that observed among de-
veloped country samples. As outlined above, this meta-analysis is
intended to improve our understanding of the sample-specific
factors that may influence the observed variation in reported
economic status–SWB effect sizes among developing countries.
Thus, this project extends past research, including qualitative and
quantitative syntheses, by (a) concentrating solely on studies con-
ducted within developing countries, (b) searching for studies in
disparate disciplines, (c) employing a number of meta-analytic
search techniques to identify as many relevant published and
unpublished studies as possible, and (d) testing for sample-level
moderators of the effect sizes, which may help to inform and
improve future research methodology.
Method
Literature Search Procedures
Several search techniques were used to retrieve applicable stud-
ies for inclusion. Only articles written in English were considered.
Studies were identified via electronic library databases
(PsychINFO, Web of Science—which included both forward and
backward searches, and Dissertation Abstracts International); a
Web-based search engine (Google Scholar); selected manuscript
reference lists (e.g., Argyle, 1999; Arthaud-Day & Near, 2005;
Diener et al., 1999); and manual searches of Social Indicators
Research (1980 –July 2006) and the Journal of Happiness Studies
(2000 –July 2006). Computerized searches involved all possible
combinations of terms reflecting SWB (life satisfaction, satisfac-
tion with life, happiness, quality of life, emotional well-being,
positive psychology, positive affect), income (wealth, income,
money, assets, possessions, poor, poverty, needs), and developing
economies (third-world, least developed, developing, low-income,
underdeveloped, transitional). Subsequently, in a series of addi-
tional searches involving Web of Science and Google Scholar, the
name of each developing country (using the World Bank [2007]
classification) was paired with the list of SWB terms in order to
retrieve relevant SWB studies that the previous searches may have
omitted. We examined the reference sections of the studies
obtained. Further, the title of each article was submitted to Web of
Science for a “times cited” search. All potentially relevant studies
published or posted through July 1, 2006, were evaluated for
inclusion.
Inclusion and Exclusion Criteria for Studies
Studies included. Only studies that met an established set of
criteria were chosen for inclusion in the meta-analysis. To be
selected, a study had to involve (a) participants of working age that
resided within a developing economy (see Footnote 5), (b) a
measure of SWB, and (c) an objective measure of income or
wealth. Acceptable constructs for SWB included single- and
multiple-item measures of life satisfaction, happiness, SWB (either
study-defined or the average of life satisfaction and happiness),
domain satisfaction (a composite of satisfaction within different
life domains such as work, family, social life, health, finances,
etc.), quality of life (study-defined), well-being (study-defined),
positive affect, and negative affect. Acceptable economic status
measures for this meta-analysis included absolute or objective
measures of the following: household income, per capita house-
hold income, personal income, household index (number or value
of durable goods or assets), expenditures, objective SES,
4
housing
quality, and standard of living.
For studies that met these criteria, it was necessary that associ-
ations between SWB and objective income or wealth be stated
4
SES was included as an acceptable measure of income only when the
original study authors combined multiple objective measures of absolute
income into a composite variable (e.g., using a standardized weighted
system that weighted an individual’s household income, number of durable
goods, and housing quality into a single composite variable). Thus, any
study that measured relative or subjective SES (i.e., asked the participant
to place themselves into an SES category) was excluded.
540
HOWELL AND HOWELL
directly or be computable from summary tables, descriptive sta-
tistics (e.g., means and standard deviations, or a table of counts),
or inferential statistics (e.g., t statistics, F ratios, odds ratios, or
chi-square statistics). For studies that used multiple regression or
probit analyses, r-equivalent effect sizes were computed from
exact p values (if available), conservative p cutoffs (e.g., .01, .05),
or zero-order correlations (when provided by authors on request). For
studies that reported results from multiple independent samples (typ-
ically multiple samples from different developing economies), each
independent sample was included and coded separately.
Studies excluded. Studies that examined the correlation be-
tween economic status and SWB for respondents in specific stages
of life (e.g., adolescents, university students, older samples) or for
clinical populations (e.g., samples with mental illness, physical
disabilities, or terminal illnesses) were excluded from the meta-
analysis because of the potential confounds associated with the
samples’ characteristics. Studies that measured only subjective or
relative wealth (e.g., “Compared to others in your country, how
wealthy are you?”) or income satisfaction were excluded. Finally,
studies that correlated national income or wealth with national
SWB (using mean SWB and an aggregate measure of income or
wealth, such as gross domestic product) were excluded.
Coding of Samples
Coding. All independent samples meeting the inclusion crite-
ria were coded for possible moderators of the economic status–
SWB relation (see Table 1). These moderators were (a) geographic
location (coded as region of the world); (b) economic stage of
development; (c) sampling technique (nationally representative,
urban, rural); (d) SWB construct; (e) economic status measure; (f)
proportion of sample with a secondary education; (g) proportion
male respondents; and (h) type of effect size reported or computed
(zero-order r or partial r).
The following criteria were established to code each moderator.
The geographic location (country) was determined from each
study’s Method section. Region of the world was coded by match-
ing study locations (by country) to the six geographic regions
designated by the World Bank (2007). Samples were grouped into
four economic development categories (low income, lower middle
income, upper middle income, and high income developing
5
)
based on each country’s World Bank classification (i.e., per capita
gross national income [GNI]) at the time of sampling. The cate-
gories are the following: low income (GNI per capita $905 or
less); lower middle income ($906 –$3,595); upper middle income
($3,596 –$11,115); and high income ($11,116 or more). Within the
high-income category are countries that are members of the Or-
ganisation for Economic Co-Operation and Development (OECD)
and those that are non-members. Although both OECD and non-
OECD countries are classified as “high-income” by the World
Bank, the two categories differ substantially on average per capita
GNI. The average per capita GNI for non-OECD countries is
$18,014 per year. The average per capita GNI for OECD countries
is $38,120 per year. Thus, only non-OECD countries are included
in the meta-analysis. For studies pre-dating 1987, we used the
author’s description of the economic status of the sample (using
keywords such as third-world) at the time of the study (e.g., Shinn,
1986). Sampling technique was categorized as nationally repre-
sentative, urban/city, or rural/village.
For each study, separately reported SWB constructs were coded
from key adjectives (e.g., satisfied, happy) and scale names (e.g.,
Satisfaction with Life Scale, Diener et al., 1985; Ladder Scale of
General Well-Being, Cantril, 1965). If item or response format
wording was not reported, the SWB term employed by the study
author was used for coding. SWB constructs that involved a
composite of items were coded as quality of life. For each study,
separately reported income and wealth constructs were coded from
author-reported variable definitions (e.g., household income, per-
sonal income, per capita household income, number of assets,
savings, objective SES). When multiple indicators of income and
wealth (e.g., household income, quality and/or location of house)
were used to form a composite economic status variable, the com-
posite was coded as objective SES. Sample demographics related to
education and gender were obtained from the Method section of
studies that reported these variables. Authors that provided the eco-
nomic status–SWB correlation via personal communication were
asked also to provide relevant demographic statistics, if available.
Both authors coded all of the studies, and the intercoder agreement
was 97%. All disagreements were resolved by discussion.
Effect sizes. On completion of coding, effect sizes for the eco-
nomic status–SWB relation were computed for each sample (see
Table 1). The specific types of effect sizes used in this meta-analysis
are from the r family, the most common of which is the Pearson
product–moment correlation. The r family also includes other biva-
riate measures of association (e.g., phi-coefficient, point-biserial cor-
relation, Spearman rank-order correlation coefficient; Rosenthal,
1994). For studies that did not report Pearson correlations between
income or wealth and SWB, r effect sizes were computed by using
Comprehensive Meta-Analysis 2.0 (Borenstein, Hedges, Higgins, &
Rothstein, 2005). For those samples that reported only p values or that
used multiple regression or probit models, exact p values were used to
determine r-equivalent effect sizes (see Rosenthal & Rubin, 2003).
Positive associations between economic status and SWB were ex-
pected for most SWB constructs. The one exception was the expected
negative association between income and negative affect (such as
with Ahmed et al., 2001). Positive effect sizes indicate the expected
direction of the economic status–SWB relation (i.e., positive correla-
tions between life satisfaction, happiness, quality of life, positive
affect, or domain satisfaction and income or wealth; a negative rela-
tion between negative affect and income or wealth).
Data Analysis
Unit of analysis. For the current study, the independent sample
was the primary unit of analysis. For each sample that reported
5
The World Bank (2007) classifies each country of the world into one
of five economic development groupings based on GNI per capita by using
the World Bank Atlas method. According to the World Bank (2007), GNI
per capita is the best single indicator of economic capacity, progress,
welfare, and success in development. OECD membership was chosen as
the criterion for determining whether or not a country was considered for
inclusion in the meta-analysis as a developing country. By grouping all
non-OECD countries into the four income categories based on GNI per
capita (low-income, lower middle income, upper middle income, and
high-income developing), we are able to observe the impact of “stage of
economic development” as a moderator of the economic status–SWB
relation as well as compare the relations within “wealthier” developing
economies to those within “poorer” developing economies.
541
ECONOMIC STATUS AND SWB: A META-ANALYSIS
Table 1
Effect Size Estimates and Sample Characteristics for Samples Ordered by Country and Then in Descending Order by Effect Size
Study N
Mean
age Country
Economic
development
stage
Sampling
technique
SWB
construct
Economic status
measure
Proportion
with
secondary
education
Proportion
male
Effect
size
Effect size
reported/
computed
Tiliouine et al. (2006) 731 25.00 Algeria
a
Lower middle Urban/city Domain
satisfaction
Personal income .94 .50 .04 r
Graham & Pettinato
(2001)
694 42.54 Argentina
b
Upper middle Nationally representative Life satisfaction Household index .48 .48 .28 r
Camfield et al. (2006) 915 Bangladesh
c
Low income Nationally representative Happiness SES .00 .34 r
Ahmed et al. (2001) 3,624 Bangladesh
c
Low income Rural/villages Worry SES .17 r
Smith (2003) 591 39.74 Belarus
d
Lower middle Nationally representative Happiness Per capita
income
.54 .46 .08 pr
Graham & Pettinato
(2001)
631 36.91 Bolivia
b
Lower middle Nationally representative Life satisfaction Household index .42 .49 .11 r
Cantril (1965) 2,168 Brazil
b
Low income Nationally representative Life satisfaction SES .10 .45 .38 r
Graham & Pettinato
(2001)
768 37.52 Brazil
b
Lower middle Nationally representative Life satisfaction Household index .65 .48 .07 r
Hayo (2003) 995 47.74 Bulgaria
d
Lower middle Nationally representative Life satisfaction Household
income
.54 .48 .20 r
Graham & Pettinato
(2001)
647 41.79 Chile
b
Upper middle Nationally representative Life satisfaction Household index .37 .45 .31 r
Cheung & Leung
(2004)
732 40.60 China
e
Lower middle Urban/city Life satisfaction Personal income .50 .22 r
Sirgy et al. (1995) 191 32.88 China
e
Lower middle Urban/city Life satisfaction Income .21 r
Jagodzinski (2005) 793 39.10 China
e
Lower middle Urban/city Combined Combined .18 r
Yang (2001) 308 40.00 China
e
Lower middle Urban/city Life satisfaction Household
income
.70 .49 .16 r
Yip et al. (2006) 1,195 45.00 China
e
Lower middle Rural/village Life satisfaction Household
income
.46 .14 pr
Foley (2005) 662 China
e
Lower middle Mixed Life satisfaction Household
income
.71 .48 .11 r
Foley (2005) 207 China
e
Lower middle Mixed Life satisfaction Household
income
.15 .50 .10 r
Graham & Pettinato
(2001)
823 37.83 Columbia
b
Lower middle Nationally representative Life satisfaction Household index .48 .49 .19 r
Graham & Pettinato
(2001)
688 39.42 Costa Rica
b
Upper middle Nationally representative Life satisfaction Household index .60 .50 .05 r
Cantril (1965) 992 Cuba
b
Low income Nationally representative Life satisfaction SES .38 .57 .16 r
Hayo (2003) 651 44.58 Czech
Republic
d
Upper middle Nationally representative Life satisfaction Household
income
.77 .46 .18 r
Cantril (1965) 2,416 Dominican
Republic
b
Low income Nationally representative Life satisfaction SES .07 .65 .42 r
Graham & Pettinato
(2001)
716 37.56 Ecuador
b
Lower middle Nationally representative Life satisfaction Household index .41 .50 .04 r
Graham & Pettinato
(2001)
619 37.44 El Salvador
b
Lower middle Nationally representative Life satisfaction Household index .46 .49 .02 r
Smith (2003) 470 39.73 Estonia
d
Upper middle Nationally representative Happiness Per capita
income
.60 .45 .14 pr
542
HOWELL AND HOWELL
Table 1 (continued)
Study N
Mean
age Country
Economic
development
stage
Sampling
technique
SWB
construct
Economic status
measure
Proportion
with
secondary
education
Proportion
male
Effect
size
Effect size
reported/
computed
Graham & Pettinato
(2001)
671 36.24 Guatemala
b
Lower middle Nationally representative Life satisfaction Household index .54 .51 .12 r
Graham & Pettinato
(2001)
616 37.57 Honduras
b
Lower middle Nationally representative Life satisfaction Household index .55 .49 .08 r
Aryee (1999) 255 40.00 Hong Kong,
China
e
High income Urban/city Life satisfaction Personal income 1.00 .47 .21 r
Leung & Lee (2005) 388 36.80 Hong Kong,
China
e
High income Urban/city Life satisfaction Personal income .67 .47 .11 pr
Liao et al. (2005) 860 40.80 Hong Kong,
China
e
High income Urban/city Combined Personal income .57 .49 .06 pr
Headey et al. (2004) 3,061 Hungary
d
Upper middle Nationally representative Life satisfaction Combined .58 .45 .19 r
Hayo (2003) 911 44.21 Hungary
d
Upper middle Nationally representative Life satisfaction Household
income
.17 r
Lelkes (2002) 5,365 Hungary
d
Upper middle Nationally representative Life satisfaction Household
income
.14 pr
Lelkes (2002) 3,802 Hungary
d
Upper middle Nationally representative Life satisfaction Household
income
.13 pr
Sta˜nculescu et al.
(2005)
2,979 Hungary
d
Upper middle Nationally representative Life satisfaction Household
income
.10 pr
Cantril (1965) 2,366 India
c
Low income Nationally representative Life satisfaction SES .06 .42 r
Biswas-Diener &
Diener (2001)
83 35.40 India
c
Low income Urban/city Life satisfaction Combined .35 .38 r
Banerjee et al. (2004) 1,024 India
c
Low income Rural/village Happiness Household
expenditure
.40 .23 pr
Brinkerhoff et al.
(1997)
341 35.00 India
c
Low income Rural/village Combined Combined .13 .91 .19 r
Jagodzinski (2005) 818 35.46 India
c
Low income Urban/city Combined Combined .81 .52 .10 r
Kousha & Mohseni
(1997)
187 35.60 Iran
a
Lower middle Urban/city Domain
satisfaction
SES .89 .00 .23 pr
Kousha & Mohseni
(2000)
1,026 32.90 Iran
a
Lower middle Urban/city Happiness Personal income .80 .41 .09 r
Kousha & Mohseni
(2000)
535 32.80 Iran
a
Lower middle Urban/city Happiness Personal income .78 .51 .05 r
Cantril (1965) 1,170 Israel
a
Low income Nationally representative Life satisfaction SES .12 .49 .55 r
Morawetz et al.
(1977)
30 38.00 Israel
a
Upper middle Rural/village Life satisfaction Per capita
income
.46 .44 .38 pr
Namazie & Sanfey
(2001)
3,797 35.00 Kyrgyzstan
d
Lower middle Nationally representative Life satisfaction Household
expenditure
.64 .46 .04 r
Lee et al. (1982) 1,500 Korea, Rep.
e
Lower middle Urban/city QOL SES .92 .51 .27 r
Shinn et al. (1983) 1,220 Korea, Rep.
e
Lower middle Nationally representative Combined Household
income
.25 r
Jagodzinski (2005) 752 37.21 Korea, Rep.
e
High income Urban/city Combined Combined .17 r
Kim (1998) 1,000 Korea, Rep.
e
High income Nationally representative Combined Combined .73 .11 r
(Table continues)
543
ECONOMIC STATUS AND SWB: A META-ANALYSIS
Table 1 (continued)
Study N
Mean
age Country
Economic
development
stage
Sampling
technique
SWB
construct
Economic status
measure
Proportion
with
secondary
education
Proportion
male
Effect
size
Effect size
reported/
computed
Smith (2003) 442 39.18 Latvia
d
Upper middle Nationally representative Happiness Per capita
income
.71 .40 .03 pr
Smith (2003) 413 41.52 Lithuania
d
Upper middle Nationally representative Happiness Household
income
.53 .46 .14 pr
Howell et al. (2006) 307 42.42 Malaysia
e
Upper middle Rural/village Life satisfaction Household index .02 1.00 .28 r
Jagodzinski (2005) 729 36.69 Malaysia
e
Upper middle Nationally representative Combined Combined .59 .49 .03 r
Lever (2004) 370 36.00 Mexico
b
Upper middle Urban/city Domain
satisfaction
Household
income
.68 .50 .43 r
Lever (2000) 768 Mexico
b
Upper middle Urban/city Domain
satisfaction
SES .47 .50 .22 r
Fuentes & Rojas
(2001)
339 30.60 Mexico
b
Upper middle Urban/city QOL Household
income
.21 r
Rojas (2004) 1,540 Mexico
b
Upper middle Nationally representative Happiness Combined .50 .17 r
Graham & Pettinato
(2001)
797 35.95 Mexico
b
Upper middle Nationally representative Life satisfaction Household index .16 r
Jagodzinski (2005) 800 35.23 Myanmar
e
Low income Urban/city Combined Combined .47 .49 .16 r
Graham & Pettinato
(2001)
642 35.26 Nicaragua
b
Low income Nationally representative Life satisfaction Household index .52 .50 .01 r
Cantril (1965) 1,200 Nigeria
b
Low income Nationally representative Life satisfaction SES .14 .81 .52 r
Suhail & Chaudhry
(2004)
983 Pakistan
c
Low income Urban/city Combined Combined .50 .36 r
Cantril (1965) 642 Panama
b
Low income Nationally representative Life satisfaction SES .20 .49 .52 r
Graham & Pettinato
(2001)
601 39.13 Panama
b
Upper middle Nationally representative Life satisfaction Household index .42 .49 .29 r
Graham & Pettinato
(2001)
390 39.18 Paraguay
b
Lower middle Nationally representative Life satisfaction Household index .44 .49 .15 r
Graham & Pettinato
(2001)
539 36.83 Peru
b
Lower middle Nationally representative Life satisfaction Household index .34 .50 .16 r
Cantril (1965) 500 Philippines
e
Low income Nationally representative Life satisfaction SES .44 r
Bulatao (1973) 941 Philippines
e
Low income Urban/city Combined Combined .18 r
Winter et al. (1999) 600 54.10 Poland
d
Upper middle Urban/city Domain
satisfaction
Combined .60 .46 .48 r
Hayo (2003) 1,103 47.36 Poland
d
Upper middle Nationally representative Life satisfaction Household
income
.06 .16 r
Sta˜nculescu et al.
(2005)
1,661 Romania
d
Lower middle Nationally representative Life satisfaction Household
income
.50 .47 .33 pr
Andre´n & Martinsson
(2006)
848 43.50 Romania
d
Lower middle Nationally representative Life satisfaction Household
expenditure
.57 .50 .15 pr
Hayo (2003) 985 43.02 Romania
d
Lower middle Nationally representative Life satisfaction Household
income
.12 r
Ferrer-I-Carbonell &
Van Pragg (2001)
1,845 Russian
Federation
d
Upper middle Nationally representative Life satisfaction Household
income
.37 r
Ferrer-I-Carbonell &
Van Pragg (2001)
2,074 Russian
Federation
d
Upper middle Nationally representative Life satisfaction Household
income
.22 r
Smith (2003) 401 40.90 Russian
Federation
d
Upper middle Urban/city Happiness Per capita
income
.21 pr
Graham et al. (2004) 4,524 40.67 Russian
Federation
d
Upper middle Nationally representative Life satisfaction Per capita
income
.42 .21 r
544
HOWELL AND HOWELL
Table 1 (continued)
Study N
Mean
age Country
Economic
development
stage
Sampling
technique
SWB
construct
Economic status
measure
Proportion
with
secondary
education
Proportion
male
Effect
size
Effect size
reported/
computed
Saris (2001) 2,000 Russian
Federation
d
Upper middle Nationally representative Life satisfaction Combined .65 .45 .16 r
Zavisca & Hout
(2005)
5,504 46.00 Russian
Federation
d
Upper middle Nationally representative Life satisfaction Household
income
.82 .43 .16 r
Ibrahim & Chung
(2003)
295 36.00 Singapore
e
High income Urban/city Life satisfaction Household
income
.55 .48 .16 r
Seik (2000) 931 37.50 Singapore
e
High income Urban/city Life satisfaction Household
income
.62 .48 .10 r
Tan et al. (2006) 987 37.00 Singapore
e
High income Urban/city Domain
satisfaction
Personal income .83 .51 .07 pr
Seik (2000) 1,935 32.00 Singapore
e
High income Urban/city Life satisfaction Household
income
.78 .52 .04 r
Keng & Hooi (1995) 242 27.00 Singapore
e
High income Urban/city Life satisfaction Combined .86 .57 .02 r
Hayo (2003) 289 43.15 Slovak
Republic
d
Lower middle Nationally representative Life satisfaction Household
income
.72 .48 .05 r
Hayo (2003) 894 50.70 Slovenia
d
High income Nationally representative Life satisfaction Household
income
.63 .51 .22 r
Møller et al. (1999) 1,530 South
Africa
f
Upper middle Nationally representative Combined Household
income
.19 .42 r
Klasen (2000) 8,724 South
Africa
f
Upper middle Nationally representative Life satisfaction Combined .31 r
Møller & Saris
(2001)
169 South
Africa
f
Upper middle Urban/city QOL Household
income
.27 pr
Møller & Saris
(2001)
540 South
Africa
f
Upper middle Urban/city QOL Household
income
.27 pr
Møller & Saris
(2001)
650 South
Africa
f
Upper middle Rural/Village QOL Household
income
.24 pr
Møller (1998) 765 South
Africa
f
Upper middle Nationally representative Combined Household
income
.21 r
Møller & Dickow
(2002)
1,990 South
Africa
f
Upper middle Nationally representative Combined Household
income
.43 .16 r
Møller (1998) 1,665 South
Africa
f
Upper middle Nationally representative Combined Household
income
.34 .12 r
Møller & Saris
(2001)
361 South
Africa
f
Upper middle Urban/city QOL Household
income
.01 pr
Møller et al. (1999) 1,290 South
Africa
f
Upper middle Nationally representative Happiness Household
income
.01 r
Jagodzinski (2005) 783 36.86 Sri Lanka
c
Lower middle Urban/city Combined Combined .97 .51 .22 r
Tsou & Liu (2001) 1,533 38.84 Taiwan
e
Upper middle Nationally representative Happiness Personal income .65 .50 .07 pr
Liao et al. (2005) 1,079 42.10 Taiwan
e
Upper middle Urban/city Combined Personal income .80 .44 .01 pr
Leelakulthanit & Day
(1992)
496 Thailand
e
Lower middle Urban/city Life satisfaction Household
income
.67 .49 .19 r
Royo & Velazco
(2006)
893 50.00 Thailand
e
Lower middle Rural/village Happiness Household index .52 .52 .08 pr
Jagodzinski (2005) 799 36.56 Thailand
e
Lower middle Nationally representative Combined Combined .77 .06 r
Gitmez & Morcol
(1994)
129 Turkey
d
Upper middle Urban/city Success in life SES .31 r
Sirgy et al. (1995) 131 32.32 Turkey
d
Upper middle Urban/city Life satisfaction Income .43 .13 r
(Table continues)
545
ECONOMIC STATUS AND SWB: A META-ANALYSIS
Table 1 (continued)
Study N
Mean
age Country
Economic
development
stage
Sampling
technique
SWB
construct
Economic status
measure
Proportion
with
secondary
education
Proportion
male
Effect
size
Effect size
reported/
computed
Olena (2005) 5,772 45.82 Ukraine
d
Lower middle Nationally representative Life satisfaction Combined .78 .42 .17 pr
Graham & Pettinato
(2001)
744 45.40 Uruguay
b
Upper middle Nationally representative Life satisfaction Household index .59 .45 .22 r
Jagodzinski (2005) 708 35.52 Uzbekistan
d
Low income Urban/city Combined Combined .94 .46 .14 r
Graham & Pettinato
(2001)
578 37.65 Venezuela
b
Upper middle Nationally representative Life satisfaction Household index .46 .50 .09 r
Jagodzinski (2005) 802 34.95 Vietnam
e
Low income Urban/city Combined Combined .56 .49 .10 r
Cantril (1965) 1,523 Yugoslavia
d
Low income Nationally representative Life satisfaction SES .22 r
Total M (SD) 131,935 38.88
(5.12)
.56(.24) .49(.13) .20
***
Note. Table organized by country, then by effect size. Positive effect sizes indicate the expected direction of the economic status–SWB relation (i.e., positive correlations between life satisfaction,
happiness, quality of life, positive affect, or domain satisfaction and income or wealth; a negative relation between negative affect and income or wealth). Sample sizes correspond to the number of
participants used to compute the effect size. The “combined” code for the economic status and SWB constructs implies that multiple indicators of the construct were aggregated. Socioeconomic status
(SES) is as defined by the author of the relevant article. Household index is based on a series of questions concerning the number of consumer goods (e.g., indoor plumbing, cars, refrigerators) owned
by the household. Domain satisfaction is a composite of satisfaction within different life domains (e.g., work, family, social life, health, finances). Each economy is classified by gross national income
(GNI) per capita, calculated using the World Bank (2007) Atlas method. The categories are low income (GNI per capita $905 or less); lower middle income ($906 –$3,595); upper middle income
($3,596 –$11,115); and high income ($11,116 or more). For comparison, the average GNI for the US was $44,970 in 2006. r Pearson r correlation coefficient; pr partial r (e.g., correlation r
controlled for one or more covariates); QOL quality of life.
a
Middle East and North Africa.
b
Latin America and Caribbean.
c
South Asia.
d
Europe and Central Asia.
e
East Asia and Pacific.
f
Sub-Saharan Africa.
***
p .001.
546
HOWELL AND HOWELL
multiple economic status–SWB correlations (i.e., multiple-
construct samples), a mean effect size was computed by Compre-
hensive Meta-Analysis 2.0 to form a within-sample aggregate
effect size for the sample. For example, two effect sizes were
calculated from Biswas-Diener and Diener (2001), who reported
correlations with life satisfaction for both objective housing (r
.30) and objective income (r .45). These two effect sizes were
aggregated together for a sample average r effect size of .38. Thus,
for all estimates of central tendency and all tests of homogeneity,
each independent sample contributed only a single effect size.
When testing for moderators, average effect sizes were com-
puted only from those samples that reported a potential moderating
variable. For moderator tests of economic status measures and
SWB constructs, two approaches were taken. First, the assumption
of independence was upheld and multiple-construct samples were
restricted to contributing only a single effect size (the within-
sample aggregate effect size) to the average effect size computed
for the appropriate construct moderator. Then, the assumption of
independence was relaxed and multiple-construct samples were
allowed to contribute a unique effect size to each average effect
size computed for the constructs on which they were measured.
Again, we consider the two effect sizes calculated from Biswas-
Diener and Diener (2001; see above): When examining the eco-
nomic status construct as a moderator, each of these effect sizes
was used to compute aggregate effect sizes for the two different
economic status constructs they represented (with objective hous-
ing representing a wealth variable and objective income represent-
ing an income variable). Yet, because these two average effect
sizes were computed with data from the same sample, they are not
independent of one another. Thus, when the assumption of inde-
pendence is relaxed, the total number of effect sizes listed can
exceed the total number of independent samples, as some samples
may be represented more than once. For these reasons, statistically
significant group differences are established only when indepen-
dent samples are significantly different and when there is no
overlap in the 95% confidence intervals of non-independent sam-
ples.
Fixed effects methods and models. Aggregate r effect sizes
and homogeneity tests of the effect sizes were estimated with
Comprehensive Meta-Analysis 2.0 by using a fixed effects ap-
proach. The fixed effects method of meta-analysis provides a more
precise and reliable estimate of the economic status–SWB relation
than might be obtained with a random effects approach (Cooper,
1998). All mean r effect sizes were calculated by averaging the
weighted (inverse variance weights
6
) correlation coefficients
across all independent samples. Homogeneity tests were used to
determine whether variance in the weighted effect sizes was ex-
plained by the proposed moderators. If a categorical moderator
explained significant variance in the effect sizes (Q
BET
.05) then
post hoc contrasts were performed to determine which groups were
statistically different. For continuous moderators, meta-regression
analyses were used to test whether variation in the effect sizes was
explained by the moderator.
7
Results
Description of the Literature Included
Publication statistics. The search techniques employed for
this meta-analysis identified a total of 56 studies that met the
established inclusion criteria. The typical study surveyed over
1,000 respondents (Mdn 1,109; M 2,357; SD 2,871). The
typical study focused on a country classified as an upper middle
income developing economy (where the GNI per capita was cal-
culated to be between $3,596 and $11,115), was published in a
peer-reviewed journal, focused on either East Asia and the Pacific
or Europe and Central Asia, and used a nationally representative or
urban/city sample. Of the 36 studies that reported gender compo-
sition, 52.8% had nearly equal male and female participants (45%
to 55% range between genders), and of the 28 studies that reported
sample education levels, an average of 59.3% (SD 26.4%) of
respondents had at least a secondary education (high school or
equivalent).
Characteristics of the independent samples. From these 56
studies, 111 independent samples were coded. The number of
independent samples reported per study ranged from 1 to 17, with
40 studies (71.4%) reporting a single independent sample, and 10
studies (17.9%) reporting only two independent samples. From
these samples, 185 distinct effect sizes were computed between
different measures of economic status and SWB. Most samples
6
The inverse variance weight is the reciprocal of the effect size variance,
which is computed as the square root of the standard error. Inverse variance
weights correlate positively with sample size.
7
Although all null hypotheses, fixed effects models, and post hoc
comparisons followed steps outlined by Hedges (1994; see Chapter 19 for
full details of fixed effects modeling), we provide some basic terminology
and background for readers not familiar with meta-analytic procedures. A
critical question raised in meta-analysis is whether the effect sizes in the
research studies are heterogeneous and, if so, whether sample-specific
characteristics explain this variation in effect sizes. Two general fixed
effects procedures attempt to answer these questions: one procedure is
analogous to analysis of variance, and one procedure is analogous to
multiple regression. When the study characteristic is categorical (e.g.,
SWB measure or wealth construct employed), then single-factor fixed
effects models partition the total heterogeneity of the effect sizes into
between-group heterogeneity (Q
BET
; df p 1) and within-group heter
-
ogeneity (Q
w
; df k p). This procedure essentially parallels the parti
-
tioning of total variance in an analysis of variance into SS
between
and
SS
within
. However, the Q statistics have chi-square distributions. Thus, the
Q
BET
tests the null hypothesis that the population effect sizes are equal
between different levels of the categorical moderator; significance demon-
strates that the categorical moderator explains significant variation in the
effect sizes. A statistically significant Q
w
demonstrates that significant
variation in the effect sizes exists within a level of a categorical moderator.
It should also be noted that when using a fixed effects model, the power to
detect significant differences is a function of the number of weighted
sample effect sizes, where each effect size is weighted by its corresponding
inverse variance weight (which is strongly influenced by sample size).
When the Q
BET
statistic is significant, it is typical to use post hoc contrasts
to determine which groups within the moderator demonstrated significantly
stronger or weaker average effect sizes; again this procedure parallels post
hoc t tests performed for significant omnibus F statistics. When the study
characteristic is continuous (e.g., education of the sample, age of the
sample, percentage of the male participants within a sample), then meta-
regression is used to model the effect sizes (the dependent variable) as a
function of the continuous predictors (where
0
is the predicted effect size
when the predictor equals 0, and
1
is the unit change in the effect size that
corresponds to a one-unit change in the predictor variable). Meta-
regression procedures are best explained by forming the regression equa-
tion (see Footnote 9).
547
ECONOMIC STATUS AND SWB: A META-ANALYSIS
reported a single economic status–SWB correlation. The 111 in-
dependent samples surveyed a total of 131,935 respondents from
54 developing countries. Of the 54 countries, 20 were indepen-
dently sampled more than once. The most frequently surveyed
countries were South Africa (k 10); China (k 9); the Russian
Federation (k 6); and Hungary, India, Mexico, and Singapore
(k 5 for each). Across the 185 distinct economic status–SWB
correlations, the two most commonly used SWB constructs were
life satisfaction (k 74) and happiness (k 32), with 7 samples
reporting both constructs. The typical economic status construct
was household income (k 53), followed by a sum of household
assets (k 35); 16 samples measured both of these economic
status constructs. Finally, most economic status–SWB relations
were measured without controlling for possible confounds (k
86).
Addressing Threats to Validity
The results of this meta-analysis and the implications of these
results are valid only if the included studies accurately represent
the population of studies measuring economic status and SWB in
developing countries. The major criticism facing any meta-
analysis involves publication bias: Not all completed studies and,
thus, not all data collected from samples surveyed have been
published. Further, unpublished studies may be more likely than
published studies to have found non-significant results ( p .05).
Theoretically, if a meta-analysis included all unpublished studies,
then the overall effect size would be reduced.
The first method used to test publication bias was a funnel plot.
In a funnel plot, effect sizes are plotted on the horizontal axis,
while the standard errors are plotted on the vertical axis in de-
scending order. The funnel plot depicts whether the overall effect
size computed from a given meta-analysis may be inflated due to
a failure to include studies in which the null hypothesis was
retained. The funnel plot demonstrated a rather symmetric distri-
bution of effect sizes and argued that the overall economic status–
SWB effect size was not a biased estimate of the population effect
size. Computation of the Begg and Mazumdar (1994) rank corre-
lation (Kendall’s tau b) between the r effect size and the standard
error further confirmed that sample size was not a predictor of the
effect size (tau b .058, p .364).
The second set of tests to examine publication bias involved
computing Orwin’s fail-safe N. Orwin’s fail-safe N calculates the
number of additional studies needed to reduce the current effect
size to a trivial magnitude (as determined by the researcher).
Because one objective of the current meta-analysis is to determine
whether the average economic status–SWB correlation is signifi-
cantly stronger for developing country samples than for developed
country samples, the trivial effect size needed to compute Orwin’s
fail-safe N was set at .12 (the expected average economic status–
SWB effect size for developed country samples). Orwin’s fail-safe
N was found to be 72. Thus, only on the addition of 72 unpublished
samples (more than the total number of studies identified for
inclusion in the meta-analysis) with average r effect sizes of .00,
would the average r effect size for developing economies match
the effect size commonly reported in developed economies. Thus,
the following meta-analysis provides a valid representation of the
strength of the association between economic status and SWB in
developing economies.
Meta-Analyzing the Samples
What is the overall economic status–SWB relation within devel-
oping countries? The mean weighted economic status–SWB r
effect size across independent samples (k 111) with fixed effects
analysis was .196 (95% confidence interval [CI] .191, .200). The
mean unweighted r effect size across independent samples with
random effects analysis was .183 (95% CI .160, .206). Both r effect
sizes were significantly different from zero (Z 77.74 and 15.22,
respectively). The unweighted median across independent samples
was .161, with 1st and 3rd quartiles of .100 and .234, respectively.
Is the economic status–SWB relation stronger in developing
economies? To establish a reasonable estimate of the economic
status–SWB relation in developed countries, effect sizes from the
following primary research articles were aggregated in Compre-
hensive Meta-Analysis 2.0: (a) Diener et al. (1993; N 8,753
nationally representative adults from the United States); and (b)
Headey et al. (2004; N 44,783 nationally representative adults
from Australia, Germany, and Britain). The average r effect size
from these two studies was .122 weighted and .118 unweighted.
The fixed effects homogeneity statistic, Q
BET
(1, k 115)
224.52, p .001, demonstrated that the economic status–SWB
association for developing countries was significantly stronger
than the association for these nationally representative developed
country samples.
Determining moderators of the effect sizes in the meta-analysis.
Prior to exploring potential moderators of the economic status–
SWB relation, an omnibus homogeneity test demonstrated sufficient
within-group variation across the 111 independent samples,
Q
w
(110) 2,359.94, p .001. Of the variables proposed as possible
moderators of the economic status–SWB relation, two moderators
(stage of economic development, and proportion of sample with a
secondary education) were specifically identified to test the hypoth-
esis that the r effect sizes will be largest for the poorest samples.
Moderator 1: Stage of economic development. We hypothe-
sized that the average r effect size would be strongest for the least
economically developed countries, as individuals within these
countries may be most likely to be at or around the threshold of
basic need fulfillment (Argyle, 1999; Easterlin, 1995; Graham,
2005). To test this hypothesis, we examined the average r effect
size within the four categories of economic development
8
(i.e., low
income, lower middle income, upper middle income, high income
developing). An omnibus homogeneity test revealed significant
differences in the average economic status–SWB r effect size
across the four stages of economic development, Q
BET
(3, k
111) 410.96, p .001 (see Table 2). Single degree of freedom
contrasts demonstrated that the economic status–SWB relation was
strongest within the low-income developing country category (r
.28) and weakest within high-income developing country category
(r .10). Interestingly, the average r effect size for the high-income
developing country category was not statistically different from the
average r effect size computed for developed countries (with data
from Diener et al., 1993, and Headey et al., 2004 —see above).
8
Studies are not evenly distributed across the four categories of eco
-
nomic development.
548
HOWELL AND HOWELL
Moderator 2: Sample education as a proxy for mean sample
income. Following the theory of diminishing marginal utility,
need theory suggests a negative effect of sample economic status
on sample r effect size. Although mean sample income or wealth
would have been the ideal moderator to test this hypothesis, few
samples reported this statistic. Studies that did report a measure of
central tendency for sample income or wealth employed a variety
of constructs (e.g., personal income, household income, SES,
number of assets, value of assets), and, thus, a common unit of
measurement (i.e., dollars) was absent. Therefore, education,
which was reported by more than 50% of the samples (k 72),
was chosen as a proxy for economic status in these analyses
(Easterlin, 2003). Indeed, education has been shown to correlate
positively with income and wealth at both the individual and
national level (Gillis et al., 1996; Graham & Pettinato, 2001; Royo
& Velazco, 2006; Sta˜nculescu et al., 2005; Tan, Tambyah, & Kau,
2006), and household income tends to be distributed closely along
the lines of educational attainment within developing countries
(Shinn, 1986). Klasen (2000) demonstrated that poverty rate, pov-
erty gap, and share of poverty gap all decreased with education.
Further, education is a positive predictor of income after control-
ling for several covariates (Graham, Eggers, & Sukhtankar, 2004).
A meta-regression confirmed the moderating effect of education
(percentage of sample with a secondary education) on the eco-
nomic status–SWB relation (see Table 3). As sample education
(and likely sample income and wealth) increased, the strength of
these effect sizes decreased.
9
To determine whether the observed
linear trend between the economic status–SWB r effect size and
education was constant across all levels of education, samples
were clustered into four groups based on the percentage of the
sample that reported a secondary education. An omnibus homo-
geneity test revealed significant differences in the r effect sizes
among the four education groups, Q
BET
(3, k 72) 962.43, p
.001 (see Table 4). Single degree of freedom contrasts demon-
strated that the average r effect size for the least-educated group
(less than 25% reporting a secondary education) was significantly
stronger than the r effect sizes calculated for the other three
education groups. No other statistical differences were observed.
These results suggest the possibility of a convex relationship
between education and the economic status–SWB effect size
among developing country samples. To test for a convex relation,
a post hoc contrast was performed with contrast weights of 8.5,
0.5, –3.5, and –5.5, respectively. The result of this contrast was
highly significant,
2
(1, N 131, 719) 792.19, p .001,
indicating that the average association between economic status
and SWB was strongest for the least-educated (and likely the
poorest) developing country samples and that the average r effect
size decreased at a diminishing rate.
Because both the economic stage of development and sample
education moderated the economic status–SWB relation, we ex-
amined the extent to which economic development status affected
the average economic status–SWB effect size, controlling for
education. Samples were sorted into two education groups based
on whether 50% or more of the sample had a secondary education.
Within each education group, average r effect sizes were computed
separately for each stage of economic development. However,
because many of the samples in the least-educated group (k 24)
9
The most straightforward way to interpret the meta-regression coeffi
-
cients is to write out the regression equation. The basic equation would be
as follows:
yˆ⫽␤
0
⫹␤
1
(X1)
where yˆ is the predicted effect size;
0
is the intercept (when
1
equals 0);
1
is the unstandardized slope (the change in the predicted effect size with
a unit change in the predictor); and X1 is the quantity of the predictor. For
example, if we consider the first significant slope for education (
1
–.327), we can use the meta-regression coefficients to predict the effect
size between income and SWB for a hypothetical sample with a given
proportion of educated respondents (where proportion of sample with a
secondary education is the predictor). In this case, the predictor runs from
0.00 (a sample devoid of participants with a secondary education) to 1.00
(a sample where all participants have at least a secondary education).
Thus, the meta-regression equation for this example (see Table 3) would
be
yˆ(predicted effect size).378.327(X1)
For example, if the proportion is 0.00 (none in the sample with a secondary
education), the predicted effect size between economic status and SWB
would be .378. If the proportion is 0.50 (half of the sample with a
secondary education), the predicted effect size would be .215. If the
proportion is 1.00 (all in the sample with a secondary education), the
predicted effect size is .051. Thus, we observe that as the participants in the
sample report higher levels of educational attainment, the strength of the
economic status–SWB effect size decreases.
Table 2
Fixed Effects Analysis of Categorical Model for Economic Development Stage
Economic development stage kN
Mean r effect
size SE
95% CI for r
Homogeneity within each
group (Q
w
)
LL UL
Low income 21 24,658 .28
a
.006 .27 .29 797.41
**
Lower middle income 34 32,969 .16
c
.005 .15 .16 355.59
**
Upper middle income 45 65,553 .19
b
.004 .19 .20 756.52
**
High-income developing 11 8,539 .10
d
.011 .08 .12 39.45
**
Note. Between-group effect (Q
BET
) for overall economic development stage was 410.96.
**
Significant Q statistic rejects null hypothesis of no variation
between groups (Q
BET
) or within specified group (Q
w
). k number of independent samples; N number of participants within independent samples; CI
confidence interval; LL lower limit; UL upper limit. Correlations with different subscripts differed significantly at p .01.
**
p .01.
549
ECONOMIC STATUS AND SWB: A META-ANALYSIS
were from low-income developing economies, we compared the
low-income group with the lower middle and upper middle income
groups combined. The average r effect size for these least-
educated and low-income developing economy samples was quite
strong (r .381; 95% CI .366, .395), while the average r effect
size for these least-educated and middle income developing econ-
omy samples was strong but significantly weaker (r .239; 95%
CI .225, .253). Interestingly, within the more-educated group
(k 48) average r effect sizes were not markedly different
between low-income developing economies (r .100; 95% CI
.072, .125), middle income developing economies (r .146; 95%
CI .137, .154), and high-income developing economies (r
.101; 95% CI .081, .122). Further, the average r effect sizes for
these more-educated groups varied little from the typical effect
sizes observed among developed country samples (r .06 –.15).
Thus, it appears that economic status is most strongly associated
with SWB when educational attainment is rare (e.g., samples are
poor) and individuals live in low-income developing countries.
Moderator 3: SWB construct. The construct used to measure
sample SWB was determined to be a moderator of the economic
status–SWB relation, Q
BET
(4, k 110) 123.48, p .001. For
independent samples, the SWB constructs having the strongest
relations with economic status were quality of life, domain satis-
faction, and life satisfaction (see Table 5). Post hoc contrasts
demonstrated no statistical differences among the average r effect
sizes for these three SWB constructs. However, the average eco-
nomic status–SWB relation for samples measured with a happiness
construct was significantly weaker than relations computed for
other SWB constructs. Even when the assumption of independence
was relaxed, the average r effect size for the 32 samples measured
on happiness (r .132; 95% CI .120, .143) remained weaker
than for samples reporting quality of life (r .251; 95% CI
.224, .277), life satisfaction (r .214; 95% CI .208, .220), or
domain satisfaction (r .161; 95% CI .144, .179).
Moderator 4: Economic status construct. The construct used
to measure sample economic status was also found to be a statis-
tically significant moderator of the economic status–SWB relation,
Q
BET
(6, k 111) 773.66, p .001. We began by grouping
economic status variables together into two groups, based on
whether wealth (e.g., stock variables) or income (e.g., flow vari-
ables) was measured. We found that the average effect size was
stronger when wealth variables were assessed (r .267; 95%
CI .257, .278) than when income variables were measured (r
.165; 95% CI .159, .172). A closer examination demonstrated
that these differences were due to two specific economic status
measures. As demonstrated in Table 5, post hoc contrasts using
independent samples showed that (a) the mean SES–SWB relation
was statistically stronger than all other income–SWB and wealth–
SWB associations and (b) the mean personal income–SWB rela-
tion was statistically weaker than all other associations. When the
assumption of independence was relaxed, the average r effect size
for the 21 samples that measured SES (r .316; 95% CI .307,
.325) remained larger than the average effect sizes computed for
all other economic status constructs, and the average r effect size
for samples that reported personal income remained smallest (r
.071; 95% CI .053, .089).
Moderator 5: Percent male respondents. A fixed effects ho-
mogeneity statistic was computed to determine whether there was
a significant difference in the average r effect size between sam-
ples that reported gender and those that did not. Samples that
reported gender were found to have a significantly weaker average
r effect size (r .179, k 82) than samples that did not report
gender (r .220, k 29), Q
BET
(1, k 111) 66.40, p .001.
Because samples that reported gender data are not fully represen-
tative of all samples in the data set, the following results should be
interpreted with some caution.
To test whether gender moderated the economic status–SWB
relation, the r effect size was meta-regressed onto the proportion of
male respondents in each sample. As shown in Table 6, sample
gender significantly moderated the economic status–SWB relation.
Table 3
Meta-Regression With Effect Size Regressed Onto Percent of
Sample With Secondary Education
Parameter Estimate SE z value
95% CI for
LL UL
0
.378 .008 49.16 .363 .393
1
.327 .012 26.34 .351 .303
Q
Model
(1, k 72) 693.90, p .001
Note. z value tests the null hypothesis that the parameter is zero in the
population. The 39 samples that did not report descriptive statistics con-
cerning sample education were not significantly different, Q
BET
(1, k
111) 2.83, p .05, from the 72 samples that did report education. CI
confidence interval; LL lower limit; UL upper limit.
Table 4
Fixed Effects Analysis of Categorical Model for Education
Percent with secondary education kN
Mean r
effect size SE
95% CI for r
Homogeneity within each
group (Q
w
)
LL UL
0%–25% 10 21,212 .36
a
.007 .35 .37 283.96
**
26%–50% 14 9,725 .16
b
.010 .14 .18 81.81
*
51%–75% 33 23,770 .14
b
.005 .13 .15 296.85
**
76%–100% 15 23,108 .13
b
.006 .12 .15 74.27
**
Note. Between-group effect (Q
BET
) for percent with secondary education was 962.43.
**
Significant Q statistic rejects null hypothesis of no variation either
between groups (Q
BET
) or within specified group (Q
wi
). k number of independent samples; N the number of participants within the independent
samples; CI confidence interval; LL lower limit; UL upper limit. Correlations with different subscripts differed significantly at p .01.
*
p .05.
**
p .01.
550
HOWELL AND HOWELL
An increase in the proportion of male respondents was associated
with a large increase in the strength of the economic status–SWB
relation. Because the gender variable was fixed between 0.00 (all
female respondents) and 1.00 (all male respondents), the meta-
regression could be used to predict the effect size for these two
extreme cases. For a hypothetical sample with all female respon-
dents, the correlation between economic status and SWB was
predicted to be relatively weak (r .06); the relation for a
hypothetically all-male sample was predicted to be much stronger
(r .31).
Exploring Methodological Choices to Explain
Heterogeneity in Effect Sizes
Because methodological choices (geographic location of the
study, sampling technique used, and statistical procedure selected)
could have affected the average correlation between economic
status and SWB, the next three sections examine how study design
may have influenced the aggregated effect sizes.
Region of the world. Region of the world was found to be a
significant moderator of the economic status–SWB effect size,
Q
BET
(5, k 111) 338.08, p .001. As demonstrated in Table
7, samples from Sub-Saharan Africa, the Middle East and North
Africa, and South Asia showed the strongest economic status–
SWB relations, while samples from Europe and Central Asia and
from East Asia and the Pacific showed the weakest average r effect
sizes.
Sampling technique. The moderating effect of sampling tech-
nique on the economic status–SWB relation was found to be
statistically significant, Q
BET
(2, k 109) 93.305, p .001. As
shown in Table 7, nationally representative samples demonstrated
significantly stronger economic status–SWB relations than were
found for rural/village or urban/city areas. No statistical difference
was found between the average r effect sizes for rural/village and
urban/city samples.
Because one criticism of a fixed effects meta-analysis concerns
the generalizability of findings, it was determined that results from
a reanalysis of the data using only the 57% of samples that were
nationally representative may most accurately reflect the economic
status–SWB relation in developing countries. Compared with re-
sults from the larger meta-analysis, effect sizes for nationally
representative samples were significantly stronger within low-
income developing economies (r .384) and high-income devel-
oping economies (r .148).
Statistical procedure. The research questions posed by each
study and the analytical approaches used to address those ques-
tions led to variation in the reported effect sizes. For example,
studies concerned only with assessing the zero-order relation be-
tween economic status and SWB within a sample (e.g., using
correlation, chi square, analysis of variance) demonstrated a stron-
ger average r effect size (r .213; k 86; 95% CI .208, .218)
than did studies concerned with assessing the strength of the partial
economic status–SWB relation (pr .134; k 25; 95% CI
Table 5
Categorical Moderators of the Economic Status–SWB Relation
Moderator
Between-group
effect (Q
BET
)
kN
Mean r
effect size SE
95% CI for r
Homogeneity within each
group (Q
w
)
LL UL
SWB construct
a
115.02
**
Quality of life 6 3,559 .23
a
.017 .21 .26 77.45
*
Domain satisfaction 6 3,643 .22
a,b
.017 .18 .25 123.28
**
Life satisfaction 65 90,462 .21
a
.003 .21 .22 1,504.06
**
Happiness 13 11,073 .11
c
.010 .09 .13 67.74
**
Combined 20 19,358 .18
b
.005 .17 .19 537.02
**
Economic status measure 773.66
**
SES 15 20,100 .35
a
.007 .34 .36 402.36
**
Household income 38 48,711 .18
c
.004 .17 .19 547.28
**
Per capita income 6 6,478 .18
c
.012 .15 .20 36.75
**
Household index 19 12,364 .14
d
.009 .12 .16 108.76
**
Household expenditures 3 5,669 .12
d
.016 .08 .15 26.85
**
Personal income 11 8,368 .07
e
.010 .05 .09 33.14
**
Combined 19 30,029 .20
b
.005 .19 .21 431.13
**
Note. Significant Q statistic rejects null hypothesis of no variation either between groups (Q
BET
) or within specified group (Q
w
). SWB subjective
well-being; k number of independent samples; N the number of participants within the independent samples; CI confidence interval; LL lower
limit; UL upper limit. Correlations with different subscripts differed significantly at p .01.
a
Negative affect was dropped from these analyses because it occurred in only one independent sample.
*
p .05.
**
p .01.
Table 6
Meta-Regression Analyses With Effect Size Regressed Onto
Proportion of Men in Sample
Parameter Estimate SE z value
95% CI for
LL UL
0
.063 .012 5.30 .040 .087
1
.247 .025 10.19 .199 .294
Q
Model
(1, k 82) 103.97, p .001
Note. z value tests the null hypothesis that the parameter is zero in the
population. There were 29 samples that did not report a gender breakdown;
these studies had a stronger average r effect size (.220) than the 82 samples
that did report gender, Q
BET
(1, k 111) 66.40, p .001. CI
confidence interval; LL lower limit; UL upper limit.
551
ECONOMIC STATUS AND SWB: A META-ANALYSIS
.123, .144). Partial effect sizes were estimated after controlling for
other covariates (as many as 10 additional demographic, eco-
nomic, political, geographic, cultural, and psychological variables)
by using ordered-probit models or OLS regression. Given these
differences, the major hypotheses were retested by examining the
average r effect sizes for those studies that reported partial r effect
sizes. For these studies, the economic status–SWB relation among
low-income developing economies ( pr .230) was larger than
among middle income developing economies ( pr .137) and
high-income developing economies ( pr .065). Also, samples
with the fewest respondents (less than 25%) reporting secondary
education showed the strongest partial relation between income
and SWB ( pr .183).
Testing the Robustness of the Meta-Analysis—A
Replication Using the World Values Survey
To test the robustness of these findings, we performed a repli-
cation meta-analysis with three waves of data from the World
Values Survey (WVS) database
10
(The European Values Study
Foundation and World Values Survey Association, 2006).
Differences between the meta-analysis and the WVS. The ma-
jor advantage of the WVS is its standardized methodology. Although
the inclusion of many different studies from various researchers and
journals allowed for a diversity of economic status–SWB associations
and moderators to be examined, the included articles also represent a
rather incongruous collection of operational definitions for both eco-
nomic status and SWB. Variables in the WVS are harmonized among
all countries examined and, thus, allow for a cleaner aggregation of
effect sizes. For example, participants in the 1990, 1995, and 2000
waves of the WVS were each asked a question about their current
level of happiness (“Taking all things together, would you say you are
[1] very happy, [2] quite happy, [3] not very happy, [4] not at all
happy”), life satisfaction (“All things considered, how satisfied are
you with your life as a whole these days: [1] dissatisfied to [10]
satisfied”), and a country-specific question regarding household in-
come. By virtue of its consistency across samples, the WVS data set
allows for tests of replication, for both the average economic status–
SWB effect size and for the moderator analyses conducted in the
initial meta-analysis, while minimizing any sample-level confounds
related to methodology.
Description of included WVS samples. We generated separate
country-level correlations between household income and happi-
ness, and household income and life satisfaction for all developing
country samples surveyed during the 1990, 1995, and 2000 waves
of the WVS. We coded each sample for the potential moderators
examined in the initial meta-analysis (e.g., economic stage of
development, education,
11
gender composition, region of the
10
We are appreciative of an anonymous reviewer who suggested using
data from the WVS to corroborate the findings from the meta-analysis. The
WVS is a longitudinal data collection project investigating cultural and
political change among diverse societies around the world. The project is
managed by an international network of university researchers and social
scientists. Thus far, there have been five survey waves: 1981, 1990, 1995,
2000, and 2005. (Only data from the first four waves were available from
the WVS Web site as of June 2007). To date, over 80 independent
countries representing nearly 85% of the world’s population have been
surveyed at least once across the five waves. For each country, the survey
is administered via oral interviews to nationally representative samples of
approximately 1,000 respondents. In addition to interviewing respondents
about household income, happiness, and life satisfaction, the questions in
the survey involve assessments and opinions regarding a number of topics:
personal values; personal health; volunteer activities; interest in politics;
feelings toward out-groups; locus of control; attitudes toward institutions
and organizations; opinions about poverty and democracy; religious affil-
iation and participation; voting attitudes and behavior; and demographics
(age, education, marital status, children, occupation, self-assessed SES,
size of town of residence, ethnicity). Additional information and WVS data
sets can be found at www.worldvaluessurvey.org
11
For the WVS data, the education variable used was the percentage of
the sample with a college degree, or equivalent. Although this is different
from the education variable used in the primary meta-analysis (percentage
of sample with secondary education, or equivalent), it was the most
consistently reported education category with a clear cutoff.
Table 7
Methodological Moderators of the Economic Status–SWB Relation
Moderator
Between-group
effect (Q
BET
)
kN
Mean r
effect size SE
95% CI for
r
Homogeneity within each
group (Q
w
)
LL UL
Region of the world 338.08
**
Sub-Saharan Africa 11 18,884 .26
a
.007 .25 .28 416.01
**
South Asia 9 10,937 .26
a
.008 .24 .28 187.09
**
Middle East and North Africa 6 3,699 .24
a,b
.016 .21 .27 256.61
**
Latin America and Caribbean 25 20,399 .23
b
.007 .22 .24 470.86
**
Europe and Central Asia 30 52,106 .17
c
.004 .16 .18 312.73
**
East Asia and Pacific 30 17,966 .15
d
.007 .14 .16 378.59
**
Sampling technique 93.305
**
Nationally representative 63 98,817 .21
a
.003 .20 .22 1,796.08
**
Urban/city 39 23,949 .16
b
.005 .15 .17 427.46
**
Rural/village 8 8,084 .17
b
.012 .15 .19 35.93
**
Note. Significant Q statistic rejects null hypothesis of no variation either between groups (Q
BET
) or within specified group (Q
w
). SWB subjective
well-being; k number of independent samples; N the number of participants within the independent samples; CI confidence interval; LL lower
limit; UL upper limit. Correlations with different subscripts differed significantly at p .01.
**
p .01.
552
HOWELL AND HOWELL
world). Only those samples measured on all of these potential
moderators were included in this replication effort.
Across the three waves of data, 74 independent samples repre-
senting a total of 102,224 participants from 54 developing coun-
tries (some countries were independently surveyed at multiple time
points) were identified as meeting our established inclusion crite-
ria. Because all 74 samples measured both happiness and life
satisfaction, 74 distinct effect sizes were computed between house-
hold income and happiness, and 74 separate effect sizes were
computed between household income and life satisfaction.
Meta-analyzing the WVS samples. The mean weighted house-
hold income–SWB r effect size for the WVS data with fixed
effects analysis was .195 (95% CI .189, .201; p .001). The
mean unweighted r effect size was only slightly stronger (r .197;
95% CI .176, .218; p .001). The unweighted median across
independent samples was .203, with 1st and 3rd quartiles of .134
and .251, respectively. The striking similarity of these aggregated
WVS effect sizes to the effect sizes calculated within the initial
meta-analysis (see Table 8) validates our estimate of r .20 as the
overall correlation coefficient between household income and
SWB in developing economies.
Replicating the moderators. For the purpose of comparison
with the aggregate effect size from developing countries, we
meta-analytically computed the average household income–SWB
relation for the developed countries included in the 1990, 1995,
and 2000 waves of the WVS. In all, 68,375 respondents from 24
economically developed countries reported their current level of
happiness, life satisfaction, and household income. The mean
weighted household income–SWB r effect size across independent
samples (k 53) with fixed effects analysis was .132 (95% CI
.124, .140). The mean unweighted r effect size was only slightly
larger (r .137; 95% CI .113, .161). These effect sizes are
similar to the average effect sizes (r .122 weighted; r .118
unweighted) estimated with the correlations from both Diener et al.
(1993) and Headey et al. (2004). Further, the fixed effects homo-
geneity statistic, Q
BET
(1, k 127) 156.56, p .001, demon
-
strated that the household income–SWB association for develop-
ing economies in the WVS was significantly stronger than the
association from developed countries in the WVS.
Also consistent with results from the initial meta-analysis (see
Table 8), the percentage of each WVS sample with a college
education (a proxy for income) was negatively associated with the
household income–SWB relation (
1
–.156, Z 3.91). Thus,
those WVS samples with the least education reported stronger
relations between household income and SWB. Given that each
WVS sample provided a separate household income–SWB corre-
lation for happiness and life satisfaction, it was possible to test the
moderating effects of the SWB construct more directly by mea-
suring the difference between these two effect sizes. Consistent
with the initial meta-analysis, the WVS correlation between house-
hold income and life satisfaction (r .223) was stronger than the
WVS correlation between household income and happiness (r
.167). We were unable to replicate the finding that the association
between household income and SWB was moderated by gender,
perhaps because there was too little variance in the gender vari-
able; most samples had near-equal numbers of male and female
participants. Finally, on examining the average WVS effect size
within different regions of the world, the average effect size
between household income and SWB was strongest in Sub-
Saharan Africa (r .277) and weakest in Latin America and the
Caribbean (r .120).
Interpretation of Correlational Results
To accurately and effectively interpret aggregate r effect sizes,
meta-analysts advocate the use of a binomial effect size display
(BESD; see Rosenthal, 1991 & 1994, for a full explanation of the
BESD procedure and rationale). We used the BESD to interpret the
average economic status–SWB effect size for low-income devel-
oping country samples with less than 25% of respondents reporting
secondary education (r .38). We presumed that these samples
would be most likely to represent individuals or households strug-
gling to meet basic needs. The BESD predicted that for a hypo-
thetical group of individuals from a low education and low-income
developing country sample, 69% of those in the “high economic
status” group would be likely to report comparatively high SWB,
whereas only 31% of individuals within the “low economic status”
group would be likely to report comparatively high SWB. Another
interpretative tool to understand the strength of correlations has
been to report the shared variance (r
2
) between the two variables.
In this case, the shared variance in the economic status–SWB
relation for those in the lowest education and least economically
Table 8
Comparison Between Meta-Analysis and World Values Survey (WVS) Effect Sizes
Statistic Meta-analysis WVS
Sample characteristic
Number of independent samples 111 74
Number of respondents 131,935 102,224
Number of countries included 56 54
Descriptive statistic
Weighted mean r effect size .196 .195
Unweighted mean r effect size .183 .197
Median r effect size .161 .203
1st and 3rd quartile r effect size .100 and .234 .134 and .251
Moderator of income–subjective well-being
Mean household income–life satisfaction relation .214 .223
Mean household income–happiness relation .132 .167
Mean household income–subjective well-being in Sub-Saharan Africa .263 .277
Standardized slope between % sample with education and effect sized from meta-regression
1
–.327
1
–.156
553
ECONOMIC STATUS AND SWB: A META-ANALYSIS
developed group (14.4%) is much larger than the typical shared
variance in developed economies (1%–2%).
Discussion
In recent years, a growing momentum to measure economic
status and SWB among disparate samples has spawned a discipli-
narily diverse collection of SWB research within developing econ-
omies. With the use of 56 studies containing a total of 111
independent samples from 54 developing countries, the aims of
this meta-analysis were to conduct a quantitative and statistically
powerful assessment of the average economic status–SWB relation
for developing country samples; to determine whether this effect
size is statistically stronger than the one typically observed in
developed economies; and to explore potential moderators of the
economic status–SWB relation within developing countries.
Results from both the primary meta-analysis and the replication
synthesis with the WVS data demonstrate that the overall eco-
nomic status–SWB relation within developing economies is ap-
proximately r .20 and significantly stronger than the average
relation estimated for developed country samples (r .13). The
association is strongest for low-income developing economy sam-
ples (r .28) and weakest for high-income developing economy
samples (r .10). Meta-regressions in both the initial meta-
analysis and with the WVS data demonstrated that the economic
status–SWB relation declined with sample-level increases in edu-
cational attainment (a proxy for sample wealth or income), such
that the effect size was strongest for the least-educated (and most
likely poorest) samples. In terms of moderators, the economic
status–SWB effect size was found to be moderated by the con-
structs used to measure SWB and economic status, as well as by
the gender distribution of the sample, with the relation being
stronger for predominately male samples.
Is Need Theory a Plausible Explanation for These
Results?
The main findings of this study relating to the statistical differ-
ences in average economic status–SWB relations observed in
developing versus developed country samples as well as low-
income developing versus high-income developing samples are
consistent with past correlational differences that SWB researchers
have identified as supporting need theory. With respect to results
of the meta-regression models, post hoc group comparisons re-
vealed a curvilinear trend by which the moderating effect of
education on the economic status–SWB relation diminished as
education increased. This finding mirrors the now-familiar con-
cave pattern between economic status and SWB, which corre-
sponds to the theory of diminishing marginal utility and has been
presented as one of the strongest arguments supporting the exis-
tence of a “threshold” of basic needs, assumed by both need theory
and relative standards theory (e.g., Cummins, 2000; Diener &
Diener, 1995; Diener, Diener, & Diener, 1995; Schyns, 2002).
Finally, our examination of the subset of low-income developing
country samples that were also low on educational attainment (k
14) revealed an effect size of r .38, implying that economic
status, as a predictor of SWB, is strongest for the poorest of the
poor. These results appear consistent with the possibility that
individuals within highly impoverished samples were at or around
the threshold of physiological need fulfillment, with those individ-
uals above the threshold reporting significantly higher SWB than
those below it.
Other Plausible Explanations for the Results
Although these results confirm correlational support for a basic
needs threshold, they are unable to verify a causal path from
economic status to SWB for developing country samples. For
example, there is the possibility that causation runs in the opposite
direction, with higher SWB leading to higher economic status (see
Frey & Stutzer, 2000; Lyubomirsky, King, & Diener, 2005), or
that the economic status–SWB relation is spurious, being affected
by a third unmeasured variable (e.g., family support; see Lever,
2004) that positively influences both economic status and SWB.
Even though our analysis of the 25 studies that controlled for many
possible confounds (e.g., age, gender, education, marital status,
social support, health status) produced a positive and significant
average partial r effect size, this result supports only the possibility
of a direct causal relationship. As noted by Cummins (2000),
“since it is obvious that money, of itself, can exert no direct
influence on SWB, the precise identification of the sources of
unique variance contributed by such intervening operational vari-
ables as hunger, perceived health, etc., is most uncertain” (p. 135).
Therefore, because few (if any) studies to date have tested medi-
ation models by using objective measures of need satisfaction over
time, we feel it is useful to conjecture other plausible explanations
for the statistical findings of this meta-analysis that may be tested
in future research.
Could the findings be due to methodological artifacts? To
investigate whether methodological artifacts (e.g., restriction of
range, differences in construct reliability) may have influenced the
effect size differences between wealthier and poorer samples, we
revisited the WVS data, which included countries in various stages
of economic development and which measured income, happiness,
and life satisfaction by using a standardized methodology. We
tested for the presence of range restriction by computing standard
deviations of the happiness, life satisfaction, and income variables
for both developing country samples as a whole and for developed
country samples as a whole. Standard deviations of the happiness
and life satisfaction measures were slightly smaller for the devel-
oped countries compared with those for the developing countries
(respectively, SD 2.05 vs. SD 2.66 for life satisfaction; and,
respectively, SD .69 vs. SD .78 for happiness). The standard
deviation for income was slightly larger among the developed
country samples (SD 2.64) than among the developing country
samples (SD 2.38). Because the WVS includes only single-item
measures of happiness and life satisfaction, the best estimate of
reliability we can provide is the correlation between happiness and
life satisfaction. This correlation was slightly stronger for the
developed country samples (r .56) than for the developing
country samples (r .46).
These results do not seem to support the argument that differ-
ences in the economic status–SWB relation are due solely to
554
HOWELL AND HOWELL
methodological artifacts.
12
Instead, it may be likely that method
-
ological artifacts (specifically construct reliability) are attenuating
the economic status–SWB relation among developing country
samples and that this relation could actually be stronger than
statistically observed (see Howell et al., 2006, for a demonstration
of greater construct attenuation in developing countries). This
argument was articulated by Biswas-Diener and colleagues (2005)
whose study involving Inughuit, Amish, and Maasai samples dem-
onstrated lower internal consistency estimates (Cronbach’s alphas)
for most measures than are typically found among samples in
Western societies. The authors conclude that “the consistency of
the findings is even more impressive, because of the random error
introduced” (p. 222).
In the current meta-analysis, a comparison of urban and rural
samples reveals statistically equal economic status–SWB effect
sizes, yet the average SWB construct reliability is higher for urban
samples (␣⫽.82) than for rural samples (␣⫽.67). This obser-
vation implies that measurement error in the SWB construct at-
tenuated the economic status–SWB relation for rural samples,
which may have been due to a lack of experience with surveys or
unfamiliarity with self-assessment (e.g., Biswas-Diener et al.,
2005). Howell et al. (2006) reported that when low reliability
resulting from measurement error was corrected, the wealth–SWB
relation for a sample of rural indigenous Malaysians increased
from r .23 to r .43. We suggest that future research report
construct reliability statistics in order to address this methodolog-
ical attenuation issue more directly.
Satisfaction of psychological needs. It is presumed that within
developing countries the absence of regular physical need fulfill-
ment results in lower SWB (e.g., Diener, Suh, Smith, & Shao,
1995). However, income and wealth may also be more strongly
related to SWB among poor samples if additional income and
assets are also able to satisfy individuals’ psychological needs.
Lever and colleagues (2005) suggested that due to the significant
relationships observed for certain psychological variables (e.g.,
coping strategies, locus of control, achievement motivation, self-
esteem, depression) with both SWB and purchasing and consump-
tion levels, “it can be hypothesized that such variables modulate or
modify the relationship between these two constructs” (p. 378).
Self-determination theory (see Ryan & Deci, 2000) has demon-
strated that individuals are happiest when the three ostensibly
universal psychological needs of autonomy, competence, and re-
latedness are met (Reis, Sheldon, Gable, Roscoe, & Ryan, 2000).
Therefore, if poverty interferes with individuals’ ability to (a)
pursue competency-producing daily behaviors (Brinkerhoff et al.,
1997), (b) work toward strengthening personal relationships
(Biswas-Diener & Diener, 2001), or (c) engage in activities of
intrinsic value, then perhaps the stronger economic status–SWB
relation within developing economies is mediated by the role of
income and assets in overcoming the barriers imposed by poverty
that hinder individuals from attaining psychological need fulfill-
ment (Diener et al., 1993). This said, much of the research to date
that has examined psychological need fulfillment across samples
with varying economic status has found that higher-level need
fulfillment is perceived to be more important and bears more
strongly on SWB in wealthier samples and nations than in poorer
ones (Lever et al., 2005; Li et al., 1998; Oishi et al., 1999). Still,
we suggest that determining whether or not psychological need
fulfillment may improve SWB among impoverished samples be-
fore, or in conjunction with, the satisfaction of physical needs
should be a continuing focus of future research.
The consumption of goods and leisure. Another possible ex-
planation for the larger economic status–SWB association ob-
served within developing countries extends the hypothesis of the
need fulfillment threshold beyond its focus on basic physiological
needs to include the consumption of non-essential goods and
leisure. Many of the modern comforts and conveniences standard
to life in developed economies are rarely present in very poor
developing country households (e.g., mattresses, indoor toilets, gas
stoves, refrigerators, motorized vehicles, automatic washing ma-
chines and dishwashers, heating and air conditioning; Biswas-
Diener & Diener, 2001; Howell et al., 2006). In some instances,
the purchase of such goods may facilitate certain chores, eliminate
arduous tasks, or increase time available for leisure. Thus, beyond
meeting basic physiological needs, it is possible that the use of
income or assets to purchase goods that make life more comfort-
able, convenient, restful, or enjoyable may, in turn, increase SWB
(Brinkerhoff et al., 1997; Diener et al., 1993; Kahneman et al.,
2004; Kousha & Mohseni, 1997; Shinn, 1986).
Because the consumption of goods and leisure adheres to the
law of diminishing marginal utility, the consumption-related gains
in SWB are assumed to decrease progressively as total consump-
tion continues to increase. Thus, it would be presumed that such
consumption would provide the greatest increase in SWB to those
initially consuming very little and then show a weaker relation
with SWB for high-consumption samples. Panel analyses with
longitudinal data would need to be conducted in order to test
whether non-essential consumption of modern comforts and con-
veniences significantly increases SWB among poor samples. Also
of interest would be whether this enhanced level of SWB among
poor samples can be sustained across time or whether adaptation
eventually occurs.
Influential Moderators and Implications for Future
Research
In addition to identifying the variables that may mediate the
economic status–SWB relation amidst poverty, it is also important
to disentangle any potential moderators of the relation. The current
meta-analysis tested several sample-level moderators. All moder-
ators were found to be statistically significant. The following
sections discuss those moderators that are most relevant to the
methodological designs of future research efforts investigating the
economic status–SWB relation within developing countries.
Should researchers be concerned with the constructs used to
measure SWB? Both the primary meta-analysis as well as the
replication study with the WVS data demonstrated the influence of
the SWB construct on the economic status–SWB relation. Specif-
ically, the economic statushappiness relation was found to be
significantly weaker than relations computed with more cognitive,
12
In order for this to have been the case, (a) we would have needed to
find that developed country samples had smaller standard deviations than
those of developing country samples on all income and SWB items, and (b)
we would have needed to find that the happiness–life satisfaction correla-
tion was weaker (e.g., demonstrated lower reliability) in the developed
countries than in the developing countries. Neither of these criteria was met
by the results of our analyses.
555
ECONOMIC STATUS AND SWB: A META-ANALYSIS
global assessments of SWB (e.g., life satisfaction, domain satis-
faction, quality of life). The results of the current meta-analysis
support past research arguments that current life circumstances
exert less influence on affective