You are on page 1of 11

Qual Life Res (2009) 18:179–189

DOI 10.1007/s11136-008-9426-2

The relative contributions of race/ethnicity, socioeconomic status,


health, and social relationships to life satisfaction in the United
States
Steven D. Barger Æ Carrie J. Donoho Æ
Heidi A. Wayment

Accepted: 17 November 2008 / Published online: 12 December 2008


 Springer Science+Business Media B.V. 2008

Abstract with life satisfaction, with emotional support having the


Purpose To evaluate racial/ethnic disparities in life satis- strongest association with life satisfaction.
faction and the relative contributions of socioeconomic
status (SES; education, income, employment status, wealth), Keywords Quality of life  Health status disparities 
health, and social relationships (social ties, emotional sup- Hispanics  Blacks  Social support  Socioeconomic status
port) to well-being within and across racial/ethnic groups.
Methods In two cross-sectional, representative samples of
U.S. adults (the 2001 National Health Interview Survey and Introduction
the 2007 Behavioral Risk Factor Surveillance System;
combined n [ 350,000), we compared life satisfaction Subjective well-being is a lynchpin of population health.
across Whites, Hispanics, and Blacks. We also evaluated For example, the World Health Organization (WHO)
the extent to which SES, health, and social relationships defines well-being as an integral part of health [1] and the
‘explained’ racial/ethnic group differences and compared number one public health priority in the United States
the magnitude of variation explained by life satisfaction (U.S.) is to increase the number and quality of years of life
determinants across and within these groups. [2]. Thus, subjective well-being is a critical component of
Results Relative to Whites, both Blacks and Hispanics psychological and health assessments, and a rigorous
were less likely to be very satisfied. Blacks were somewhat understanding of well-being determinants is a precondition
more likely to report being dissatisfied. These differences for policy interventions to promote health and well-being
were reduced or eliminated with adjustment for SES, [3, 4].
health, and social relationships. Together, SES and health Well-being surveillance at the population level is essen-
explained 12–15% of the variation in life satisfaction, tial to more precisely identify levels of health status in the
whereas social relationships explained an additional 10– U.S., as well as to monitor the conditions contributing to or
12% of the variance. detracting from one’s well-being [5]. Despite a maturing
Conclusions Racial/ethnic life satisfaction disparities literature regarding this dimension of life quality, studies of
exist for Blacks and Hispanics, and these differences are nationally representative samples are urgently needed
largest when comparing those reporting being ‘satisfied’ to because much well-being research is conducted in ‘‘…small,
‘very satisfied’ versus ‘dissatisfied’ to ‘satisfied.’ SES, accidental samples of respondents’’ [4]. One undesirable
health, and social relationships were consistently associated consequence of convenience sampling is the paucity of work
on well-being among minorities, particularly Hispanics. In
2006, there were an estimated 44.3 million Hispanics in the
U.S., comprising 14.8% of the population [6], yet, the well-
being literature provides almost no data regarding this largest
S. D. Barger (&)  C. J. Donoho  H. A. Wayment
U.S. minority population. The evaluation of Hispanic
Department of Psychology, Northern Arizona University, P.O.
Box 15106, Flagstaff, AZ 86011, USA well-being is essential to improve U.S. public health efforts
e-mail: steven.barger@nau.edu to identify and reduce health and well-being disparities [2],

123
180 Qual Life Res (2009) 18:179–189

i.e., unnecessary, avoidable, and unjust health differences To address these gaps in the literature, we sought to
[7], and to accurately characterize well-being determinants advance our understanding of well-being by evaluating
for policy decisions [3]. We use life satisfaction as an inte- markers of important life satisfaction domains across two
grative measure of well-being. Both constructs encompass very large, nationally representative samples of non-His-
pleasure, engagement, and other positive emotions [4]. Life panic White, non-Hispanic Black, and Hispanic U.S.
satisfaction is also considered as a marker of quality of life in residents. Specifically, the present study: (1) evaluated
Healthy People 2010 [2]. racial/ethnic disparities in life satisfaction; (2) partitioned
the relative contribution of health, SES, and social rela-
tionships to these disparities; and (3) compared the relative
Well-being predictors strength of association among life satisfaction predictors,
both within and across racial/ethnic groups.
Health status, as denoted by perceived health and physical
disability, is an important predictor of well-being [8, 9]. To
better understand the contribution of health to well-being, Methods
it is important to compare the relative contribution of
health when simultaneously considering other well-being Data sources
correlates, such as income, education, and unemployment.
Although studies have addressed the relative importance of We analyzed two U.S. population-based public health
life satisfaction predictors among racial/ethnic groups [9, surveys, the 2001 National Health Interview Survey
10], we are aware of none that have done so with repre- (NHIS) and the 2007 Behavioral Risk Factor Surveillance
sentative samples that include Hispanics. Relative System (BRFSS). Although the focus of these surveys was
comparisons of life satisfaction predictors help to prioritize not life satisfaction, a life satisfaction assessment was
research resources. Further, representative multi-ethnic available for these years. Informed consent was obtained
within-country samples can more precisely test hypotheses from all participants.
that cultural factors moderate well-being [4]. We, there-
fore, compared the relative predictive strength of health National Health Interview Survey
and socioeconomic status (SES) variables for life satis-
faction across diverse White, Black, and Hispanic The NHIS is the primary U.S. health surveillance instrument
population-based samples. for a broad range of health conditions [18]. The survey uses
Social relationships are also important determinants of stratified, multistage cluster sampling to obtain a probability
life satisfaction [4, 11, 12], and forming and maintaining sample of U.S. households in the 50 states and the District of
social relationships are theorized to be a fundamental Columbia [19]. Households are contacted and randomly
human motive [13]. Social relationships are central to chosen adults are interviewed in person in their residence
theoretical models of SES [14], and adequate testing of using computer-assisted interviewing technology. Black and
these models requires the assessment of a variety of social Hispanic participants are oversampled to provide reliable
relationship markers. One established social relationship estimates for these populations. The 2001 survey included
indicator, marital status, is a robust life satisfaction corre- 33,326 adult respondents (73.8% conditional response rate;
late [8]. However, qualitative and quantitative markers of [20]). Of those respondents, 32,121 participants described
social relationships, such as emotional support and social themselves as Hispanic (n = 5,615; 29.4% interviewed in
integration (‘‘…active engagement in a wide range of Spanish), non-Hispanic Black (n = 4,622), or non-Hispanic
social activities or relationships’’ [15]) are theoretically White (n = 21,884). Life satisfaction data were missing
important, but have received less attention in the context of for 1.6, 2.4, and 1.8% of these groups, respectively
life satisfaction [10, 16]. Emotional support and social (n = 31,537). Missing values on covariates in fully adjusted
integration should be potent predictors of well-being multivariate models reduced the sample to 92.7% of those
because they are theorized to buffer stressful experiences with life satisfaction data (final n = 29,243).
and to promote positive psychological states, respectively
[15]. Simultaneously modeling marital status, emotional Behavioral Risk Factor Surveillance System
support, and social integration will help clarify the poten-
tial mechanisms through which being married confers The BRFSS is an annual telephone survey of over 350,000
greater well-being. Finally, variation in the predictive adults (18? years of age) in the 50 United States, the
strength of social relationships across racial/ethnic groups District of Columbia, Puerto Rico, the U.S. Virgin Islands,
may determine whether the benefits of social relationships and Guam. It is a population-based stratified probability
vary across sociocultural contexts [17]. sample that provides the primary source of information on

123
Qual Life Res (2009) 18:179–189 181

health-related behaviors in the U.S. [21]. The median Socioeconomic status


cooperation rate across the 50 states in 2007 was 72% [22].
Of the 430,912 participants, 403,137 described themselves Socioeconomic status (SES) was assessed by education
as Hispanic (n = 31,310; 41.1% interviewed in Spanish), (dummy variables representing \high school; high school
non-Hispanic Black (n = 33,216), or non-Hispanic White graduate or equivalent; some college; college graduate;
(n = 338,611). Life satisfaction data were missing for 6.7, postgraduate degree [NHIS]), household income ([ or
7.1, and 4.0% of these groups, respectively, leaving \$20,000 per year [NHIS]; eight categories ranging from
385,163 participants (95.5% of the three racial/ethnic \$10,000 to [$75,000 per year [BRFSS]), employment
groups). Missing data on covariates in fully adjusted mul- status, and wealth. Employment was represented by
tivariate models reduced the sample to 85.4% of those with dummy codes representing retired, unemployed, and never
life satisfaction data (final n = 329,004). worked [NHIS] versus working (the referent group in both
Census-based weights for respondents were calculated surveys). The BRFSS analyses included dummy categories
to adjust for nonresponse and to ensure representativeness. for student, homemaker, and unable to work. Wealth was
Our analyses incorporated the weights, sample strata, and coded dichotomously (own home vs. rent/other [NHIS]).
clusters, and, therefore, estimate parameters for the adult
civilian, non-institutionalized population of the United Health status
States [20]. We used Stata 10.0 (Stata Corp., College
Station, TX) for all of the analyses. Health was assessed by the presence or absence of any
reported disability (‘‘Are you limited in any way in any
Measures activities because of physical, mental, or emotional prob-
lems?’’ [BRFSS]; any limitations reported for physical
There are modest differences in item wording across sur- activities, such as carrying groceries, grasping small
veys and, generally, the BRFSS has a less comprehensive objects, walking a quarter of a mile, etc. [NHIS]). Partic-
set of health, social ties, and SES variables (e.g., fewer ipants rated their health as poor, fair, good, very good, or
chronic disease indicators, no measures of wealth or social excellent. Finally, we created a summary measure of
contacts). Nonetheless, similar associations across samples reported chronic disease diagnoses (hypertension, diabetes,
and variable definitions increase generalizability and pro- heart attacks, coronary heart disease, cancer [NHIS], and
vide greater confidence in the associations. Items available other heart disease [NHIS], summed and grouped into
only in one survey are identified in brackets. none, 1, or 2 or more).

Social relationships
Life satisfaction
These resources were assessed by marital status (married/
In both surveys, life satisfaction was measured with the cohabiting or not), emotional support (How often do you
question ‘‘In general, how satisfied are you with your life?’’ get the social and emotional support you need? Would you
The response options included very satisfied, satisfied, say always, usually, sometimes, rarely, or never?), and
dissatisfied, or very dissatisfied. Because the very dissat- social integration. Social integration [NHIS only] was the
isfied category was reported by 1% of participants in both sum of six questions covering the 2-week prevalence of
samples, we combined it with the dissatisfied category. contacts with friends or relatives over the phone or in
person, as well as whether they attended a group social
Demographic variables activity or a religious service. Although they are not perfect
indicators, these social relationship markers capture ele-
Gender and age (dummy codes for six age categories) were ments of the general human motive to have stable, lasting,
used as covariates. The respondents reported a racial cat- and positive social interactions with others [13].
egory (e.g., White, Black/African American) and whether
or not they were of Hispanic ethnicity [23]. Participants Analytic strategy
were categorized into non-Hispanic White, non-Hispanic
Black, and Hispanic (henceforth referred to as White, We evaluated well-being disparities by comparing levels of
Black, or Hispanic, respectively). Racial and ethnic clas- life satisfaction for White, Hispanic, and Black participants.
sifications encompass a number of social categories related Next, we examined, in turn, the relative contributions of
to ancestral origins, language, history, and customs [24], SES, health status, and social relationships to racial/ethnic
but the present classification structure reflects the minimum disparities in life satisfaction. This strategy characterizes
administrative standards of the U.S. Government [23]. the relative importance of these domains for life satisfaction

123
182 Qual Life Res (2009) 18:179–189

and addresses possible mediating factors for between-group of 1.33 would reflect a 33% greater likelihood of being
variation in life satisfaction. dissatisfied, presuming that the confidence intervals did not
We then repeated these multivariate analyses within include 1.0. Odds are on a log scale, where the distance
each of the three racial/ethnic groups using ordinary least from 1.0 to 2.0 is the same as from 0.5 to 1.0. Below, we
squares regression. We report R2 effect size estimates to report NHIS and BRFSS ORs, respectively, with 95% CIs
characterize the variance in life satisfaction accounted for in brackets.
by the blocks of SES, health, and social relationship vari- Relative to Whites, Blacks were more likely to be dis-
ables. Variance was estimated with Taylor series satisfied (OR = 1.31 [1.22, 1.52] and 1.33 [1.20, 1.47])
linearization [25] to accommodate the clustered sampling and less likely to be very satisfied (OR = 0.68 [0.63, 0.74]
design. and 0.65 [0.61, 0.68]). Black/White differences in dissat-
isfaction were eliminated with SES adjustment
(OR = 0.92 [0.78, 1.09]) and 0.90 [0.81, 1.01]). For very
Results satisfied, Black/White differences persisted following
adjustment for SES, health, and social relationships
An overview of the demographic characteristics for each covariates in the NHIS but not the BRFSS (OR = 0.87
sample is presented in Table 1. [0.79, 0.96] and 0.99 [0.92, 1.05]).
Hispanics were equally likely to report being dissatisfied
Life satisfaction in the U.S. by race/ethnicity relative to Whites (OR = 1.0 [0.86, 1.16] and 0.88 [0.78,
1.0]), but were less likely to report being very satisfied
Whites had higher life satisfaction relative to Blacks and (OR = 0.80 [0.74, 0.86] and 0.69 [0.65, 0.73]). A pattern of
Hispanics (see Fig. 1). This pattern was largely explained lower odds of being dissatisfied for Hispanics was observed
by racial/ethnic differences in the very satisfied and satis- with adjustment for SES (OR = 0.72 [0.60, 0.86] and 0.66
fied categories, with a greater proportion of Whites in the [0.57, 0.77]) and this Hispanic advantage persisted after
very satisfied category. Within-group life satisfaction was adjusting for all covariates, although the confidence inter-
consistent across the two surveys (Fig. 1). vals approached 1.0 in the NHIS (OR = 0.82 [0.68, 1.00]).
To characterize the extent of poor life satisfaction across Comparing the odds for very satisfied versus satisfied, SES
racial/ethnic groups, we used the person-level weights in covariates removed Hispanic/White differences in both
the BRFSS to estimate the number of adults in the U.S. who samples (OR = 1.09 [1.00, 1.18] and 1.02 [0.95, 1.09]), a
are dissatisfied with their lives. We estimate that, in 2007, pattern which persisted, except in BRFSS analyses that
1.6 million Blacks (95% confidence interval [CI] 1.5– included health (OR = 1.11 [1.03, 1.19]) and, additionally,
1.8 million), 7.7 million Whites (95% CI 7.4–7.9 million), social relationship covariates (OR = 1.16 [1.07, 1.25]).
and 1.6 million Hispanics (95% CI 1.5–1.8 million) were Thus, moderate Black/White differences in life satisfaction
dissatisfied or very dissatisfied with their lives. were observed, but were attenuated after adjustment for
potential confounders. Smaller initial Hispanic/White life
Racial/ethnic disparities in life satisfaction satisfaction differences were observed and were eliminated
or reversed in multivariate models.
To contextualize the regression analyses, correlations
among the key predictor variables are presented in Table 2. The relative contributions of SES, health status, and
These estimates reveal independence among predictors and social relationships
moderately higher life satisfaction–income correlations
than that reported previously (r = 0.22–0.25 vs. 0.13) [26]. Across all participants, SES indicators accounted for 6–9%
Racial/ethnic disparities were evaluated using two sets of the variance in life satisfaction judgments. Adding
of regressions, each including a binary variable; one vari- health variables to these equations increased the explained
able compared Hispanics with Whites, the other Blacks variance to 12–15%. Adding social relationship variables
with Whites. To more precisely characterize patterns of substantially increased the explained variance to up to 24–
disparities, we compared dissatisfied and very satisfied 25% of the variance across both samples (see Table 3).
ratings with the satisfied category using multinomial logit
models. Our measure of association was the exponentiated Magnitude and consistency of individual life
regression coefficient or odds ratio (OR), which reflects satisfaction predictors
differences in the likelihood of being in a particular life
satisfaction category. For example, when comparing the In general, the largest life satisfaction predictors were
likelihood of being dissatisfied (relative to satisfied), a ratio unemployment, disability, self-rated health, and the three
of 1.0 would reflect no group differences, whereas a ratio social relationship markers. These patterns were consistent

123
Qual Life Res (2009) 18:179–189 183

Table 1 Percentages and means for the sociodemographic characteristics of the 2001 National Health Interview Survey (NHIS) and the 2007
Behavioral Risk Factor Surveillance System (BRFSS)

2001 National Health Interview Survey 2007 Behavioral Risk Factor Surveillance System
White Black Hispanic Total White Black Hispanic Total
(n = 21,884) (n = 4,622) (n = 5,615) (n = 32,121) (n = 338,611) (n = 33,216) (n = 31,310) (n = 403,137)

Age
18–24 11.8 16.3 19.0 13.2 9.5 12.9 16.8 11.0
25–34 16.3 20.8 24.8 17.8 15.9 20.1 27.1 18.1
35–44 21.2 23.3 23.4 21.7 19.4 22.1 23.3 20.3
45–54 19.6 17.9 15.1 18.9 20.0 18.6 15.1 19.1
55–64 12.8 10.0 8.5 12.0 15.6 13.4 9.1 14.3
C65 18.2 11.7 9.2 16.4 19.5 12.9 8.5 17.1
Gender
Female 51.9 55.5 50.8 52.2 52.0 54.0 49.6 51.8
Male 48.1 44.4 49.2 47.8 48.0 46.0 50.4 48.2
Education
\High school 12.7 24.3 43.7 17.6 6.8 13.5 32.3 11.5
High school or GED 30.4 30.3 23.1 29.6 28.6 33.9 29.4 29.3
Some college 30.0 30.4 21.3 29.1 27.1 28.9 20.3 26.2
College graduate 17.2 10.0 7.2 15.2 37.3 23.3 17.1 32.7
(or higher)a
Postgraduatea 8.9 3.9 2.4 7.6 – – – –
Missing 0.8 1.1 2.3 1.0 0.2 0.5 0.7 0.3
Annual income (US $)
\$10,000 – – – – 2.4 7.6 9.6 4.1
$10,000–15,000 – – – – 3.0 6.2 9.1 4.3
$15,000–19,999 – – – – 4.4 9.2 11.2 6.0
\$20,000b 15.6 29.2 28.1 18.7 – – – –
$20,000–24,999 – – – – 6.2 9.4 11.8 7.4
[$20,000b 78.5 63.9 65.7 75.3 – – – –
$25,000–34,999 – – – – 9.2 13.0 12.7 10.2
$35,000–49,999 – – – – 13.6 13.6 11.9 13.4
$50,000–74,999 – – – – 16.9 12.2 8.8 15.1
C$75,000 – – – – 31.9 14.9 12.3 27.1
Missing 5.9 6.9 6.2 6.0 12.5 13.9 12.4 12.6
Own home
Yes 76.4 49.2 49.3 29.7 – – – –
No 23.6 50.8 50.7 70.3 – – – –
Employment status
Employed 66.4 63.8 67.0 66.2 60.9 57.6 61.4 60.6
Retired 16.7 9.8 6.9 14.7 18.7 13.6 7.4 16.4
Never worked 2.8 6.0 10.7 4.1 – – – –
Unemployed 13.9 19.9 15.0 14.8 3.8 8.8 6.1 4.7
Student – – – – 3.8 5.9 5.1 4.2
Homemaker – – – – 8.0 3.6 13.6 8.4
Unable to work – – – – 4.5 9.8 5.5 5.2
Missing 0.2 0.5 0.3 0.2 0.3 0.7 0.9 0.5
Self-rated health
Poor 2.8 4.7 3.4 3.1 4.1 5.7 5.2 4.4
Fair 8.1 13.1 9.4 8.8 9.6 15.4 22.3 12.3

123
184 Qual Life Res (2009) 18:179–189

Table 1 continued
2001 National Health Interview Survey 2007 Behavioral Risk Factor Surveillance System
White Black Hispanic Total White Black Hispanic Total
(n = 21,884) (n = 4,622) (n = 5,615) (n = 32,121) (n = 338,611) (n = 33,216) (n = 31,310) (n = 403,137)

Good 23.6 27.6 26.8 24.5 28.3 34.6 35.7 30.2


Very good 33.2 27.4 31.5 32.3 36.0 26.4 20.3 32.5
Excellent 32.1 27.0 28.8 31.1 21.6 17.3 16.1 20.3
Missing 0.1 0.2 0.0 0.1 0.3 0.5 0.4 0.4
Disability
No 64.2 68.1 78.1 66.3 78.7 79.3 84.2 79.6
Yes 35.4 31.5 21.7 33.4 19.7 18.3 12.5 18.4
Missing 0.3 0.4 0.2 0.3 1.6 2.5 3.3 2.0
Chronic diseases
0 63.9 62.8 75.4 65.0 66.2 58.6 73.2 66.5
1 22.7 23.7 16.5 22.1 23.0 26.7 18.4 22.7
C2 13.4 13.5 8.1 12.9 10.8 14.7 8.5 10.8
Marital status
Married/cohabitating 66.4 45.1 64.3 63.7 67.6 56.7 64.4 64.5
Other 33.5 54.9 35.7 36.3 32.1 42.8 35.3 35.2
Missing – – – – 0.3 0.5 0.3 0.3
Social relationships Mean (SE)

Social ties 4.53 (0.012) 4.64 (0.029) 4.42 (0.029) 4.53 (0.011) – – – –
Emotional support 4.24 (0.008) 4.16 (0.018) 4.19 (0.019) 4.23 (0.007) 4.24 (0.003) 3.98 (0.015) 3.92 (0.019) 4.17 (0.004)
Life satisfaction 1.40 (0.005) 1.28 (0.011) 1.34 (0.010) 1.38 (0.045) 1.42 (0.002) 1.28 (0.007) 1.33 (0.007) 1.39 (0.002)
a
College graduate and above was a combined category in the BRFSS
b
Income was coded as greater than or less than US $20,000 in the NHIS
– denotes an item not included in the survey. SE = standard error. All standard errors for demographic variables were less than or equal to 0.02.
The variable percentages may not sum exactly to 100 due to rounding

Fig. 1 Life satisfaction by race/


ethnicity: United States, 2001 2001 NHIS N = 31,537 2007 BRFSS N = 385,163
and 2007. NHIS = National
Health Interview Survey;

.6
.6

Dissatisfied
BRFSS = Behavioral Risk Satisfied
Very satisfied .56 .56 .56
Factor Surveillance System. The .54
proportions may not sum to 1.0
.49 .48 .47
due to rounding .46
.4
.4

.40
Proportion

.38
Proportion

.36 .36
.2
.2

.08 .08
.06 .06 .05 .05
0
0

White Hispanic Black White Hispanic Black

across both surveys and within each racial/ethnic group bivariate analyses, but persisted in explaining [8% of the
(see Fig. 2). Emotional support had the largest association variance in fully adjusted models. Despite this consistency
of any predictor. It accounted for 15% of the variance in in the full sample, there were appreciable racial/ethnic

123
Qual Life Res (2009) 18:179–189 185

Table 2 Population-based correlations among life satisfaction, sociodemographic, health, and social relationship variables
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

1. Life satisfaction – 0.00 0.15 0.25 –0.15 – -0.20 0.31 -0.09 0.18 0.39
2. Male 0.01 – 0.00 0.08 -0.03 – -0.03 0.02 0.02 0.05 -0.02
3. Education 0.16 0.02 – 0.49 -0.14 – -0.07 0.29 -0.11 0.12 0.16
a
4. Family income 0.22 0.07 0.39 – -0.26 – -0.19 0.37 -0.18 0.35 0.22
5. Unemployed -0.14 -0.15 -0.15 -0.24 – – 0.12 -0.12 0.05 -0.13 -0.09
6. Own home 0.14 0.00 0.11 0.40 -0.06 – – – – – –
7. Disability -0.18 -0.10 -0.13 -0.16 0.22 0.05 – -0.39 0.27 -0.07 -0.12
8. Self-rated health 0.29 0.04 0.28 0.29 -0.26 0.05 -0.43 – -0.39 0.09 0.21
9. Chronic diseases -0.09 -0.01 -0.12 -0.14 0.16 0.08 0.36 -0.43 – -0.03 -0.06
10. Married 0.18 0.06 0.09 0.36 -0.05 0.24 -0.03 0.05 0.00 – 0.13
11. Emotional support 0.39 0.01 0.08 0.14 -0.07 0.10 -0.13 0.16 -0.04 0.12 –
12. Frequency of social contacts 0.25 -0.08 0.19 0.15 -0.08 0.08 -0.14 0.21 -0.09 0.06 0.22
Estimates are based upon Hispanic, non-Hispanic Black, and non-Hispanic White participants in the 2001 National Health Interview Survey
(NHIS: below diagonal) and the 2007 Behavioral Risk Factor Surveillance System (BRFSS: above diagonal). Sample sizes range from 20,071–
32,121 in the NHIS to 333,350–403,137 in the BRFSS. Unemployment compares currently employed to unemployed, but not retired, students,
etc. Unemployed correlations are based upon 20,071–25,295 (NHIS) and 217,403–380,229 (BRFSS) observations. Social contacts and home
ownership were not assessed in the BRFSS
a
The NHIS variable was last year’s family income in 11 levels ranging from $0–$4,999 to C$75,000 (n = 24,078). This measure was not used
in the primary analyses because a substantial number of participants (23.5%) did not report income in these graded categories. The correlation
with the binary family income variable was smaller (r = 0.15)

Table 3 Multivariate estimates of variance (R2) in life satisfaction accounted for by racial/ethnic disparities, socioeconomic status (SES), health
status, and social relationships by national health survey
Age and gender ?SES ?Health ?Social relationships Emotional support omitted
NHIS BRFSS NHIS BRFSS NHIS BRFSS NHIS BRFSS NHIS BRFSS

Overall 0.2 0.2 6.2 9.1 12.7 14.8 25.5 24.5 17.5 15.9
Disparitiesa
Hispanic vs. White 0.3 0.5 6.2 9.0 13.1 15.0 26.0 24.9 17.8 15.0
Black vs. White 0.6 0.8 6.6 10.0 13.1 15.9 26.5 26.6 18.0 15.7
Within groups
White 0.1 0.2 6.6 10.0 13.5 16.0 27.1 27.4 18.4 17.3
Black 0.6 0.6 4.6 7.8 9.0 12.7 21.4 20.7 14.0 13.2
Hispanic 0.8 0.1 4.7 4.5 10.8 9.4 19.3 15.7 14.3 13.2
NHIS = National Health Interview Survey (n = 29,243 in fully adjusted models). BRFSS = Behavioral Risk Factor Surveillance System
(n = 329,004 in fully adjusted models). The baseline model included age and gender, while subsequent models added, in turn, socioeconomic
status (SES; education, income, home ownership [NHIS only], employment status), health status (functional limitations, self-rated health, and
diagnosed chronic diseases), and social ties (marital status, emotional support, and frequency of social contacts [NHIS only])
a
Disparities analyses compare only Black/White or Hispanic/White participants using a binary race/ethnicity variable

differences in explanatory power for emotional support marital status was 0.3–1.2% (median = 0.7%) when added
(approximately 9–10, 7, and 2.5–5% of the variance to models containing the social relationship variables.
explained for Whites, Blacks, and Hispanics, respectively;
Table 3). Similar gaps in explained variance were observed
for the fully adjusted regressions, in which 27, 21, and 16– Discussion
19% of the variance was explained for Whites, Blacks, and
Hispanics, respectively. Across racial/ethnic groups and This study evaluated racial/ethnic disparities in life satis-
surveys, the incremental increase in explained variance for faction, and explored the relative contributions of SES,

123
186 Qual Life Res (2009) 18:179–189

Fig. 2 Partial regression 2007 Behavioral Risk Factor Surveillance System


coefficients for life satisfaction All Participants Hispanics Whites Blacks
predictors by race/ethnicity and
U.S. public health survey Male
High school
Some college
College grad
Income
Unemployed
Homemaker
Student
Retired
Unable to work
Disability
Chronic diseases
Self-rated health
Married
Emotional support

-.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2


a) Regression Coefficient (b)

2001 National Health Interview Survey


All Participants Hispanics Whites Blacks

Male
Own home
High school
Some college
College grad
> College
>20K income
Retired
Unemployed
Never worked
Disability
Chronic diseases
Self-rated health
Married
Emotional support
Social contacts

-.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2 -.2 -.1 0 .1 .2


b) Regression Coefficient (b)

health status, and social relationships to life satisfaction through which these disparities may be reduced. Further,
among two very large, diverse probability samples of U.S. this information provides a foundation for improving
adults. This is the first major evaluation of Hispanic life quality of life, a major priority for both U.S. and interna-
satisfaction in the U.S. and is the largest U.S. population- tional health policy.
based life satisfaction study to date. We found that Blacks The relative importance of life satisfaction predictors
and Hispanics have lower life satisfaction than Whites, but varied across self-reported race and ethnicity, accounting
controlling for SES and health status attenuated these dif- for the least variance among Hispanics, with increased
ferences for Blacks and eliminated them for Hispanics. We explanatory power for Blacks and still more so for
also found a modest Hispanic benefit for being very sat- Whites. These differences became particularly marked for
isfied in multivariable models. Statistical control aside, Hispanics in models lacking measures of wealth and
these data clearly show that disparities exist for the two social integration. The consistently higher explained life
largest racial/ethnic groups in the U.S., and falsify claims satisfaction variance among Whites could represent sub-
that ‘‘…knowing someone’s race or ethnic group … gives stantive cultural variation in the types of support relevant
little clue to the person’s psychological well-being’’ [27] to well-being judgments [30] or reflect methodological
(for more evidence of racial well-being disparities, see [9, variance, e.g., differential measurement error and/or
28, 29]). This study addresses a central goal of U.S. health interactions of ethnicity with other predictors. Disentan-
policy [2] by identifying well-being disparities and by gling these possibilities is an important direction for
identifying plausible health and economic mechanisms future research.

123
Qual Life Res (2009) 18:179–189 187

Although these racial/ethnic differences in well-being Our health variables were strongly and consistently
and its correlates are consistent with cultural moderation associated with life satisfaction in both samples, but which
hypotheses [4], we had no direct culture measures and it is is more important to well-being, subjective or objective
perilous to conflate ethnicity with culture [31]. In addition, health measures [37]? On the one hand, we found that
our use of one ‘‘Hispanic’’ category likely masks variation reported chronic disease diagnoses were weakly related to
in well-being across Hispanic national origin groups, var- life satisfaction relative to other health reports. However,
iation which has been documented for SES markers and the putatively less objective disability and self-rated health
health risk [32]. Even though we statistically controlled for reports tap substantive health dimensions. For example,
differences in education, income, employment status, and self-reported disabilities [40] and global health appraisals
wealth, racial/ethnic imbalances in SES may persist [33]. [41] predict mortality, and the latter health-related quality
Thus, disparities may reflect residual SES confounding of life judgment exhibits discriminant validity from nega-
rather than inadequate well-being models or unmeasured tive moods [42]. Chronic diseases were, in fact, associated
well-being correlates. with poorer well-being in our data, but that association was
The robust association of life satisfaction with social eliminated or substantially reduced when controlling for
relationships provides strong support for theories empha- disability and subjective health appraisals (data not shown).
sizing social ties [13–15]. The independent association of Thus, the well-being impact of ill health is better captured
both functional (emotional support) and structural (num- by reported disability and global self-rated health.
ber of social contacts) social relationships replicates This study has a number of strengths. We presented
previous work [10] and is consistent with studies showing timely, population-based life satisfaction estimates for
the broader health relevance of these two dimensions [34, Whites, Hispanics, and Blacks using two different repre-
35]. Moreover, these aspects of the social environment are sentative samples of U.S. adults. These surveys utilized
potentially modifiable, revealing intervention targets (as rigorous sampling procedures and multiple assessment
well as resources to protect) when considering clinical or methods, i.e., both telephone and in-home computer-
policy interventions [3]. The potency of social relation- assisted interviews. The combined sample size in this
ship predictors supports theoretical models of SES that report is commensurate with the sum of participants in all
emphasize active participation in society and social U.S. life satisfaction literature to date, and the Hispanic and
engagement as the ultimate resources conferred by high Black participants in our study each exceeded the total
social status [36]. Relative to income or education, the sample for 33 years of aggregated data from the U.S.
social relationship markers in this study are more proxi- General Social Survey [9]. This study, thus, directly and
mal indicators of this participatory capacity, and, thus, decisively addresses population well-being [5]. In addition,
from this perspective, one would expect a stronger life we incorporated multiple markers of SES, health, and
satisfaction association for variables more directly social relationships, and provided the first population-based
assessing this resource. This interpretation is consistent analysis of well-being among Hispanics, a group that is
with longitudinal studies showing that social contacts are estimated to comprise 15.5% of the population by 2010 [6].
substantially more important to life satisfaction than We also partitioned the importance, in explained variance
increases in income [16]. terms, of marital status relative to emotional support, and
Prior reviews emphasizing marriage as a key life satis- provided evidence prioritizing the relative importance of
faction predictor have not considered the joint effects of theoretically important life satisfaction predictors within
other types of social ties [37, 38]. Our analyses confirm the and across racial/ethnic groups. These analyses help con-
importance of marital status, yet, suggest a subordinate role textualize the relative influence of life satisfaction
for this condition. Relative to being married, reporting determinants and provide a common metric for comparison
emotional support accounted for roughly eight times more with other studies.
variance in life satisfaction, a pattern seen in research using Despite these strengths, a number of cautions are appro-
depressed mood as an outcome [39]. The weak bivariate priate. We measured life satisfaction with a single item
association between marital status and emotional support assessed at one point in time. Including more sophisticated
undermines the argument that being married confers social life satisfaction assessments obtained on multiple occasions
support, whereas the positive correlation between marital is desirable. There are still unrepresented minority popula-
status and education, income, and home ownership suggest tions that deserve research attention, and space constraints
alternative mechanisms for the higher well-being enjoyed precluded the examination of happiness and sadness,
by those who are married. Continued attention to the joint dimensions that capture somewhat independent aspects of
contributions of multiple well-being indicators will clarify well-being [43]. Contextual factors, such as neighborhood
their relative importance and the potential mechanisms quality, receive less research attention but also predict well-
through which they influence well-being. being [44], as do moods [45], personality [46], and major life

123
188 Qual Life Res (2009) 18:179–189

events [11]. Most important, our cross-sectional design Journal of the American Medical Association, 299, 2081–2083.
cannot determine whether the observed associations are doi:10.1001/jama.299.17.2081.
4. Diener, E., & Seligman, M. E. P. (2004). Beyond money: Toward
causal. In fact, existing evidence reveals a fascinating con- an economy of well-being. Psychological Science in the Public
coction of reciprocal influence among the variables Interest, 5, 1–31.
considered here. For example, social integration predicts and 5. Diener, E. (2000). Subjective well-being. The science of happi-
results from well-being [16]; slightly lower life satisfaction ness and a proposal for a national index. The American
Psychologist, 55, 34–43. doi:10.1037/0003-066X.55.1.34.
has been observed among people who will divorce [47]; 6. U.S. Census Bureau. (2008). Hispanics in the United States.
people in poor health are more likely to become unemployed Available online at: http://www.census.gov/population/www/soc
[48]; and continuous employment can preserve psychologi- demo/hispanic/hispanic_pop_presentation.html.
cal health, even among those with initially poorer health (at 7. Whitehead, M. (1992). The concepts and principles of equity and
health. International Journal of Health Services, 22, 429–445.
least among men [49]). Only by evaluating the prevalence of doi:10.2190/986L-LHQ6-2VTE-YRRN.
various health and socioeconomic conditions in combination 8. Easterlin, R. A. (2003). Explaining happiness. Proceedings of the
with the effect sizes for each possible bidirectional associa- National Academy of Sciences of the United States of America,
tion can we begin to address the relative importance of these 100, 11176–11183. doi:10.1073/pnas.1633144100.
9. Yang, Y. (2008). Social inequalities in happiness in the United
pathways and their suitability for modification [11]. States, 1972 to 2004: An age-period-cohort analysis. American
Sociological Review, 73, 204–226.
10. Taylor, R. J., Chatters, L. M., Hardison, C. B., & Riley, A.
(2001). Informal social support networks and subjective well-
Conclusions being among African Americans. The Journal of Black Psy-
chology, 27, 439–463. doi:10.1177/0095798401027004004.
These data provide an inclusive picture of life satisfaction 11. Ballas, D., & Dorling, D. (2007). Measuring the impact of major
in the U.S., revealing reliable racial/ethnic disparities and life events upon happiness. International Journal of Epidemiol-
ogy, 36, 1244–1252. doi:10.1093/ije/dym182.
suggesting order for the relative importance of a number of 12. Bishop, A. J., Martin, P., & Poon, L. (2006). Happiness and
life satisfaction correlates. Having emotional support, a congruence in older adulthood: A structural model of life satis-
job, good health, no disabilities, diverse social networks, faction. Aging & Mental Health, 10, 445–453.
and a spouse are all strong and independent predictors of 13. Baumeister, R. F., & Leary, M. R. (1995). The need to belong:
Desire for interpersonal attachments as a fundamental human
being satisfied with life. These patterns are particularly motivation. Psychological Bulletin, 117, 497–529. doi:10.1037/
compelling given their consistency across two nationally 0033-2909.117.3.497.
representative samples and racial/ethnic groups. Our study 14. Oakes, J. M., & Rossi, P. H. (2003). The measurement of SES in
informs health policy decision-making and provides a health research: current practice and steps toward a new
approach. Social Science & Medicine, 56, 769–784. doi:10.1016/
foundation for exploring psychological factors that may be S0277-9536(02)00073-4.
selected or impaired by conditions such as disability, 15. Cohen, S. (2004). Social relationships and health. The American
unemployment, or social isolation [50]. We hope that our Psychologist, 59, 676–684. doi:10.1037/0003-066X.59.8.676.
work clarifies the relative importance of well-being deter- 16. Powdthavee, N. (2008). Putting a price tag on friends, relatives
and neighbours: Using surveys of life satisfaction to value social
minants [3] in order to improve individual and population relationships. Journal of Socio-Economics, 37, 1459–1480. doi:
health in the U.S. and elsewhere. 10.1016/j.socec.2007.04.004.
17. Uchino, B. N. (2004). Pathways linking social support to health
Acknowledgments The views expressed in this paper are the outcomes. In B. N. Uchino (Ed.), Social support and physical
authors and neither those of the National Center for Health Statistics health: Understanding the health consequences of our relation-
(NCHS) nor the Centers for Disease Control and Prevention (CDC). ships (pp. 109–144). New Haven, CT: Yale University Press.
We are grateful to the NCHS, the CDC, and the survey participants 18. National Center for Health Statistics (2008). National Health
for making this study possible. Interview Survey (description). Available online at: http://www.
cdc.gov/nchs/about/major/nhis/hisdesc.htm.
19. National Center for Health Statistics. (2003). National Health
Interview Survey. Data file documentation, National Health
Interview Survey, 2001 (machine readable data file and docu-
References mentation). Hyattsville, Maryland: National Center for Health
Statistics, Centers for Disease Control and Prevention.
1. World Health Organization. (2005). Promoting mental health: 20. National Center for Health Statistics. (2003). 2001 National
Concepts, emerging evidence, practice. World Health Organiza- Health Interview Survey (NHIS). Public use data release. NHIS
tion, Department of Mental Health and Substance Abuse in Survey description. Hyattsville, Maryland: National Center for
collaboration with the Victorian Health Promotion Foundation Health Statistics, Centers for Disease Control and Prevention.
and the University of Melbourne. 21. Centers for Disease Control and Prevention. (2008). Health risks
2. U.S. Department of Health and Human Services. (2000). Healthy in the United States: Behavioral risk factor surveillance system
People 2010: Understanding and improving health (2nd ed.). (At a glance). Available online at: http://www.cdc.gov/nccdphp/
Washington, DC: US Government Printing Office. publications/aag/pdf/brfss.pdf.
3. Kindig, D. A., Asada, Y., & Booske, B. (2008). A population 22. Centers for Disease Control, Prevention. (2008). 2007 Behavioral
health framework for setting national and state health goals. Risk Factor Surveillance System (BRFSS). Summary data quality

123
Qual Life Res (2009) 18:179–189 189

report. Atlanta, Georgia: U.S. Department of Health and Human 36. Marmot, M. (2004). The status syndrome: How social standing
Services, Centers for Disease Control and Prevention. affects our health and longevity. New York: Times Books.
23. Office of Management and Budget. (1997). Recommendations 37. Seligman, M. E. P. (2002). Authentic happiness: Using the new
from the Interagency Committee for the Review of the Racial and positive psychology to realize your potential for lasting fulfill-
Ethnic Standards to the Office of Management and Budget ment. New York: Free Press.
Concerning Changes to the Standards for the Classification of 38. Myers, D. G. (2000). The funds, friends, and faith of happy
Federal Data on Race and Ethnicity. Federal Register, 62, people. The American Psychologist, 55, 56–67. doi:10.1037/
36873–36946. 0003-066X.55.1.56.
24. National Research Council. (2004). Understanding racial and 39. Ross, C. E. (1995). Reconceptualizing marital status as a con-
ethnic differences in health in late life: A research agenda. tinuum of social attachment. Journal of Marriage and the Family,
Washington, DC: The National Academies Press. 57, 129–140. doi:10.2307/353822.
25. Korn, E. L., & Graubard, B. I. (1999). Analysis of health surveys. 40. Reuben, D. B., Rubenstein, L. V., Hirsch, S. H., & Hays, R. D.
New York: Wiley. (1992). Value of functional status as a predictor of mortality:
26. Diener, E., Sandvik, E., Seidlitz, L., & Diener, M. (1993). The results of a prospective study. The American Journal of Medicine,
relationship between income and subjective well-being: Relative 93, 663–669. doi:10.1016/0002-9343(92)90200-U.
or absolute? Social Indicators Research, 28, 195–223. doi: 41. Idler, E. L., & Benyamini, Y. (1997). Self-rated health and
10.1007/BF01079018. mortality: A review of twenty-seven community studies. Journal
27. Myers, D. G., & Diener, E. (1995). Who is happy? Psychological of Health and Social Behavior, 38, 21–37. doi:10.2307/2955359.
Science, 6, 10–19. doi:10.1111/j.1467-9280.1995.tb00298.x. 42. Barger, S. D., Burke, S. M., & Limbert, M. J. (2007). Do induced
28. Thomas, M. E., & Holmes, B. J. (1992). Determinants of satis- moods really influence health perceptions? Health Psychology,
faction for Blacks and Whites. The Sociological Quarterly, 33, 26, 85–95. doi:10.1037/0278-6133.26.1.85.
459–472. doi:10.1111/j.1533-8525.1992.tb00385.x. 43. Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999).
29. Thomas, M. E., & Hughes, M. (1986). The continuing signifi- Subjective well-being: Three decades of progress. Psychological
cance of race: A study of race, class, and quality of life in Bulletin, 125, 276–302. doi:10.1037/0033-2909.125.2.276.
America, 1972–1985. American Sociological Review, 51, 830– 44. Subramanian, S. V., Kim, D., & Kawachi, I. (2005). Covariation
841. doi:10.2307/2095370. in the socioeconomic determinants of self rated health and hap-
30. Dressler, W. W., & Bindon, J. R. (2000). The health conse- piness: A multivariate multilevel analysis of individuals and
quences of cultural consonance: Cultural dimensions of lifestyle, communities in the USA. Journal of Epidemiology and Com-
social support, and arterial blood pressure in an African American munity Health, 59, 664–669. doi:10.1136/jech.2004.025742.
community. American Anthropologist, 102, 244–260. doi: 45. Schwarz, N., & Clore, G. L. (1983). Mood, misattribution, and
10.1525/aa.2000.102.2.244. judgments of well-being: Informative and directive functions of
31. Hunt, L. M., Schneider, S., & Comer, B. (2004). Should affective states. Journal of Personality and Social Psychology,
‘‘acculturation’’ be a variable in health research? A critical review 45, 513–523. doi:10.1037/0022-3514.45.3.513.
of research on US Hispanics. Social Science & Medicine, 59, 46. Steel, P., Schmidt, J., & Shultz, J. (2008). Refining the relationship
973–986. doi:10.1016/j.socscimed.2003.12.009. between personality and subjective well-being. Psychological
32. Barger, S. D., & Gallo, L. C. (2008). Ability of ethnic self- Bulletin, 134, 138–161. doi:10.1037/0033-2909.134.1.138.
Identification to partition modifiable health risk among US resi- 47. Lucas, R. E. (2005). Time does not heal all wounds: A longitu-
dents of Mexican ancestry. American Journal of Public Health, dinal study of reaction and adaptation to divorce. Psychological
98, 1971–1978. doi:10.2105/AJPH.2007.122754. Science, 16, 945–950. doi:10.1111/j.1467-9280.2005.01642.x.
33. Kaufman, J. S., Cooper, R. S., & McGee, D. L. (1997). Socio- 48. Jusot, F., Khlat, M., Rochereau, T., & Serme, C. (2008). Job loss
economic status and health in Blacks and Whites: The problem of from poor health, smoking and obesity: A national prospective
residual confounding and the resiliency of race. Epidemiology survey in France. Journal of Epidemiology and Community
(Cambridge, Mass.), 8, 621–628. doi:10.1097/00001648-1997 Health, 62, 332–337. doi:10.1136/jech.2007.060772.
10000-00002. 49. Cable, N., Sacker, A., & Bartley, M. (2008). The effect of
34. Eng, P. M., Rimm, E. B., Fitzmaurice, G., & Kawachi, I. (2002). employment on psychological health in mid-adulthood: findings
Social ties and change in social ties in relation to subsequent total from the 1970 British Cohort Study. Journal of Epidemiology and
and cause-specific mortality and coronary heart disease incidence Community Health, 62, e10. doi:10.1136/jech.2007.063776.
in men. American Journal of Epidemiology, 155, 700–709. doi: 50. Ryff, C. D., Keyes, C. L. M., & Hughes, D. L. (2004). Psycho-
10.1093/aje/155.8.700. logical well-being in MIDUS: Profiles of ethnic/racial diversity
35. Krumholz, H. M., Butler, J., Miller, J., Vaccarino, V., Williams, and life-course uniformity. In O. G. Brim, C. D. Ryff, & R. C.
C. S., Mendes de Leon, C. F., et al. (1998). Prognostic importance Kessler (Eds.), How healthy are we? A national study of well-
of emotional support for elderly patients hospitalized with heart being at midlife (pp. 398–424). Chicago: The University of
failure. Circulation, 97, 958–964. Chicago Press.

123