ArticlePDF Available

Volunteering by Older Adults and Risk of Mortality: A Meta-Analysis

Authors:

Abstract and Figures

Organizational volunteering has been touted as an effective strategy for older adults to help themselves while helping others. Extending previous reviews, we carried out a meta-analysis of the relation between organizational volunteering by late-middle-aged and older adults (minimum age = 55 years old) and risk of mortality. We focused on unadjusted effect sizes (i.e., bivariate relations), adjusted effect sizes (i.e., controlling for other variables such as health), and interaction effect sizes (e.g., the joint effect of volunteering and religiosity). For unadjusted effect sizes, on average, volunteering reduced mortality risk by 47%, with a 95% confidence interval ranging from 38% to 55%. For adjusted effect sizes, on average, volunteering reduced mortality risk by 24%, with a 95% confidence interval ranging from 16% to 31%. For interaction effect sizes, we found preliminary support that as public religiosity increases, the inverse relation between volunteering and mortality risk becomes stronger. The discussion identifies several unresolved issues and directions for future research. (PsycINFO Database Record (c) 2013 APA, all rights reserved).
Content may be subject to copyright.
Volunteering by Older Adults and Risk of Mortality: A Meta-Analysis
Morris A. Okun and Ellen WanHeung Yeung
Arizona State University
Stephanie Brown
Stony Brook University Medical Center and University of
Michigan
Organizational volunteering has been touted as an effective strategy for older adults to help
themselves while helping others. Extending previous reviews, we carried out a meta-analysis of the
relation between organizational volunteering by late-middle-aged and older adults (minimum age
55 years old) and risk of mortality. We focused on unadjusted effect sizes (i.e., bivariate relations),
adjusted effect sizes (i.e., controlling for other variables such as health), and interaction effect sizes
(e.g., the joint effect of volunteering and religiosity). For unadjusted effect sizes, on average,
volunteering reduced mortality risk by 47%, with a 95% confidence interval ranging from 38% to
55%. For adjusted effect sizes, on average, volunteering reduced mortality risk by 24%, with a 95%
confidence interval ranging from 16% to 31%. For interaction effect sizes, we found preliminary
support that as public religiosity increases, the inverse relation between volunteering and mortality
risk becomes stronger. The discussion identifies several unresolved issues and directions for future
research.
Keywords: volunteering, mortality, older adults, moderation
Using insights generated from evolutionary theory, Brown and
her colleagues (S. L. Brown & Brown, 2006; S. L. Brown, Nesse,
Vinokur, & Smith, 2003) advanced the hypothesis that helping
behavior, whatever its effect on the recipient, promotes the psy-
chological well-being and health of the helper. Providing assis-
tance to another improves relationship satisfaction and enhances
stress regulation (Post, 2007). Brown and her colleagues have
shown that helping behavior among older adults is associated with
accelerated recovery from depressive symptoms that accompany
spousal loss (S. L. Brown, Brown, House, & Smith, 2008) and
reduced mortality risk (S. L. Brown et al., 2003) even among
caregivers (S. L. Brown et al., 2009). An independent research
team, using a sample of more than 1,000 older adults from New
York City, reported similar findings for morbidity (W. M. Brown,
Consedine, & Magai, 2005).
Prosocial behaviors refer to intentional efforts to provide assis-
tance to another individual or communities. Planned prosocial
activities include caregiving, providing support to others, contrib-
uting to other churchgoers, and volunteering. Organizational or
formal volunteering is an unpaid, voluntary activity that involves
“. . . taking actions within an institutional framework that poten-
tially provides some service to one or more other people or to the
community at large” (Piliavin & Siegl, 2007, p. 454). In the current
meta-analysis, we examined the relation between organizational
volunteering and risk of mortality among adults 55 years old and
older.
We chose to focus exclusively on organizational volunteering be-
cause, in contrast to helping familiar others and engaging in social
activities with familiar others in informal social contexts, organiza-
tional volunteering entails helping unfamiliar others in an institutional
context. Theoretically, we expect that social activities recruit different
neural circuitry than helping others and, although helping familiar and
unfamiliar others should recruit similar neural circuitry under some
conditions (S. L. Brown, Brown, & Preston, 2012), we reasoned that
tests of this possibility should await an initial inquiry into whether
there is indeed a reliable association between volunteering and re-
duced mortality risk.
We made the decision to exclude younger adults for both
pragmatic and theoretical reasons. Pragmatically, the variability in
mortality is much lower in younger than older adults, the causes of
death differ for younger and older adults (Mathers, Boerma, & Fat,
2009), and the effects of volunteering on mortality via processes
such as stress regulation are not likely to be observed until later life
(Belloc & Breslow, 1972). From a theoretical perspective, aging is
associated with life transitions that often involve role losses. Con-
sequently, the role of volunteer may be especially important to the
emotional and physical health of older adults (Van Willigen,
2000).
Because volunteering was a measured rather than a manipulated
variable in the sources included in the current meta-analysis, it was
important to take into account variables such as health, social
interaction, and social connection that are positive selection factors
for volunteering (Thoits & Hewitt, 2001). In examining the ad-
justed relation between volunteering and mortality risk, we used as
covariates age, sex, physical health, socioeconomic status, health
behaviors, marital status, religiosity/religious behavior, emotional
health, social connection, social interaction, ethnicity, work status,
cognitive functioning, and leisure activity.
This article was published Online First February 18, 2013.
Morris A. Okun and Ellen WanHeung Yeung, Department of Psychol-
ogy, Arizona State University; Stephanie Brown, Center for Medical
Humanities, Compassionate Care, and Bioethics, Stony Brook University
Medical Center and Institute for Social Research, University of Michigan.
Correspondence concerning this article should be addressed to Morris A.
Okun, Department of Psychology, Arizona State University, Tempe, AZ
85287-1104. E-mail: okun@asu.edu
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Psychology and Aging © 2013 American Psychological Association
2013, Vol. 28, No. 2, 564 –577 0882-7974/13/$12.00 DOI: 10.1037/a0031519
564
Previous Reviews
We identified five reviews of studies examining the relation
between organizational volunteering and mortality risk. In these
reviews, the number of studies of the relation between volunteer-
ing and mortality risk ranged from five to eight. S. L. Brown and
Okun (in press) focused on the bivariate relation between volun-
teering and mortality risk and reported that volunteering reduced
mortality risk. The other reviewers looked at adjusted relations in
which sets of covariates were included in models examining the
association between volunteering and mortality risk. These review-
ers concluded that even when other factors are statistically con-
trolled, individuals who volunteer are more likely to live longer
(Grimm, Spring, & Dietz, 2007; Harris & Thoresen, 2005; Oman,
2007; von Bonsdorff & Rantanen, 2011).
Although previous reviewers have drawn similar conclusions
regarding the inverse relation between volunteering and mortality
risk, several limitations in the methodologies of the individual
studies suggest the need for meta-analytic techniques, which
would enhance our confidence in conclusions drawn from the
results of several studies. Using these techniques, we are able to
provide more precise information on the distribution of effect
sizes, including central tendency, variability, and confidence in-
tervals. Based on theoretical analyses (S. L. Brown & Brown,
2006) and the conclusions of previous reviewers (Grimm et al.,
2007; Oman, 2007), we predicted that there would be (a) a signif-
icant (p 05) inverse unadjusted (bivariate) association between
volunteering and mortality risk and (b) a significant inverse ad-
justed association between volunteering and mortality risk. Inspec-
tion of the numerical estimates of the strength of the association
between volunteering and mortality risk (S. L. Brown & Okun, in
press) revealed that they vary substantially. Therefore, we pre-
dicted that there would be a significant (p .05) amount of
heterogeneity in the unadjusted and adjusted effect sizes.
Harris and Thoresen (2005) concluded that although the pres-
ence of covariates did not fully eliminate the association, it did
substantially reduce the relation between volunteering and mortal-
ity risk. To examine whether part of the association was due to
third-party variables such as health and social interaction, we
tested the hypothesis that there would be a significant (p .05)
decrease in the magnitude of the association between volunteering
and mortality risk when adjusted effect sizes are compared with
unadjusted effect sizes.
An unresolved issue in this literature pertains to the form of the
relation between volunteering and morality risk. Researchers have
reported that the relation between volunteering and mortality is (a)
linear (Oman, Thoresen, & McMahon, 1999); (b) nonlinear, ex-
hibiting a threshold effect (Luoh & Herzog, 2002); and (c) non-
linear, exhibiting a U-shaped effect (Musick, Herzog, & House,
1999). The threshold effect is based on the notion that a certain
minimum amount of volunteering is required for older adults to
obtain the health-related benefits. The curvilinear effect adds to the
threshold effect the notion that volunteering beyond certain levels
creates role strain, which in turn nullifies the health-related bene-
fits of volunteering. Therefore, we examined whether the data
provide more support for depicting the relation between volunteer-
ing and mortality risk as linear or nonlinear.
Another unresolved issue in this literature involves whether
there are individual differences in the benefits that older adults
derive from volunteering. Oman (2007) proposed two alternative
hypotheses regarding how individual difference variables influ-
ence the association between volunteering and risk of mortality.
According to the compensatory hypothesis, as the individual’s
resources (human, social, and cultural capital) decrease, the benefit
of volunteering on mortality risk reduction increases. In contrast,
according to the complementary hypothesis, as the individual’s
resources increase, the benefit of volunteering on mortality risk
reduction increases. The assumption underlying the compensatory
hypothesis is that volunteering provides older adults with in-
creased capital and a role that can offset the loss of other roles. In
this case, volunteering should provide the greatest benefits to those
with the fewest resources. According to the complementary hy-
pothesis, volunteering by older adults taxes their limited reservoir
of coping resources. Thus, the benefits of volunteering should be
greatest for individuals who already have adequate amounts of
funds or capital. To test between these alternative hypotheses, we
carried out two types of analyses. We investigated whether the
relation between volunteering and mortality risk differed in sub-
samples (e.g., older adults with weak and strong social ties). We
also computed volunteering by moderator variable (e.g., volun-
teering by individual difference) interaction effect sizes.
The ability to statistically investigate whether the association
between volunteering and mortality risk differs as a function of
personal, social, situational, and cultural influences is an advantage
of using meta-analysis. Such analyses can shed light on why
studies yield diverse effect sizes and suggest methodological and
substantive boundary conditions on the relation between volun-
teering and mortality risk. Contingent on finding that the effect
sizes were heterogeneous and that a substantial proportion of the
observed variation was not spurious, we sought to identify study-
level moderator variables that might explain this variation, includ-
ing study focus, publication impact factor, country where volun-
teering took place, historical time of the study, age composition of
the sample, and proportion of sample deceased.
Because of biases against publication of null effects, we ex-
pected that effect sizes would be stronger in studies explicitly
testing hypotheses regarding volunteering as a predictor variable
as compared to studies that focused on other predictor variables
and included volunteering as a control variable. Using a similar
rationale, we anticipated that effect sizes would be stronger in
articles published in more as opposed to less prestigious journals.
Because of cohort and/or period effects, the relation between
volunteering and mortality risk may have shifted over historical
time. Consequently, we used year of publication as a moderator
variable. Due to differences in cultural norms regarding helping
others via organizational volunteering, the relation between vol-
unteering and mortality risk may vary between countries. In the
current meta-analysis, we compared effect sizes derived from U.S.
samples with effect sizes derived from Israeli samples. Because
role loss increases with age, we examined whether effect sizes are
stronger in studies with older rather than younger minimum age
requirements. As the death and volunteering rates deviate from
0.50, these variables have less variability, which may lead to
smaller effect sizes. To determine whether variability in the death
and volunteering rates was associated with effect size magnitude,
we used percentage deceased and percentage volunteering as mod-
erator variables.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
565
VOLUNTEERING AND MORTALITY
Narrative and meta-analyses alike are potentially biased by the
tendency for studies yielding statistically significant effects to be
published, whereas studies yielding nonstatistically significant ef-
fects end up in the file draw. In the current meta-analysis, we
examined the potential impact of publication bias using the trim
and fill procedure (Duval & Tweedie, 2000).
Only 32% of the potential interaction effect sizes were retrieved
from the studies. We posited that interaction effect sizes would be
more likely to be missing when the tests of the interaction effects
were not statistically significant. To test this notion, we investi-
gated whether effect sizes were less likely to be reported when the
p values associated with the test of the volunteer by moderator
variable interaction effects were greater than .05 as opposed to less
than .05.
Method
Literature Search Procedure
The processes of searching, selecting, and coding sources were
carried out by the first author, a PhD, and the second author, a
senior graduate student with extensive training in quantitative
methods.
Inclusion criteria. To be included in this meta-analysis, (a)
the source had to be published as a journal article or book chapter
written in English; (b) the source had to report on empirical
research; (c) the study had to include a measure of organizational
volunteering and mortality had to be an outcome variable; (d) the
design had to be prospective, that is, volunteering had to be
assessed prior to a mortality surveillance period; and (e) the unit of
analysis had to be the individual.
Search strategies. We used multiple strategies to compile our
list of studies. We searched the Medline and PsycINFO databases
on November 3, 2011. The search strategy involved pairing vol-
unteer, volunteerism, and volunteering in the document title with
the keywords mortality, death, longevity,orsurvival. For the
Medline database, the command line used for the search was
((volunteer
.ti. and mortality.af. and english.lg.) or (volunteer
.ti.
and death.af. and english.lg.) or (volunteer
.ti. and longevity.af.
and english.lg.) or (volunteer
.ti. and survival.af. and english.lg.)).
This search yielded 253 journal articles. The specific syntax used
in the command line for the search of the PsycINFO database was
(TI(volunteer
)) AND ((cabs(mortality) or cabs(death) or cab
-
s(longevity) or cabs(survival))). This search yielded a total of 38
nonredundant sources. Next, we searched the reference lists of
previous reviews of the relation between volunteering and mortal-
ity (S. L. Brown & Okun, in press; Grimm et al., 2007; Harris &
Thoresen, 2005; Oman, 2007). This strategy netted us additional
two sources.
Based on a preliminary screening of the abstracts of the (291)
sources, 13 sources were retrieved for coding. The low yield
rate was due to the inclusion of the search term volunteers. This
search term captured many studies in which the word volunteers
was used to describe how the sample was drawn, and these
studies had nothing to do with studying the effects of organi-
zational volunteering. Eleven of these articles were included in
the meta-analysis. One article was excluded because the unit of
analysis was the neighborhood rather than the individual
(Blakely et al., 2006), and a second article was excluded be-
cause the focus was on the survival times of terminally ill
patients who did and did not receive support from volunteers
(Herbst-Damm & Kulik, 2005). In an effort to obtain additional
sources, we examined the reference lists of the articles accepted
for inclusion in the meta-analysis. This endeavor yielded an-
other three sources, raising the total number of sources included
in the meta-analysis to 14.
Coding
Coding was guided by a codebook devised by the first author. A
planning sheet was used to facilitate the coding process. The
planning sheet enabled the coder to organize the information in
the source with respect to whether effect sizes were extracted from
the total sample and from subsamples, the measure(s) of volun-
teering, and the types of effect sizes extracted. Nine forms were
developed for coding information regarding the source, the total
sample, subsamples, volunteering measures, mortality measure,
unadjusted effect sizes, adjusted effect sizes, interaction effect
sizes, and interaction tests of statistical significance. The source
form was filled out in its entirety except for the two sources that
were rejected. For all accepted sources, the total sample form was
completed. The remaining forms were used as many times as
needed.
The first and second authors independently coded five of the
accepted sources. The mean number of disagreements per 100
items coded was 4.50 (SD 2.99). A total of 37 disagreements
occurred, including eight disagreements over whether items
needed to be coded (omission disagreements) and 29 regarding the
values of items (commission disagreements). The first set of omis-
sion disagreements pertained to whether five statistical tests of
interaction effects from one source should be coded. This set of
disagreements arose because the authors of this source stated in
their overview that they tested eight interaction effects, but in the
Results section, they reported statistical tests for only three of the
eight interaction effects (i.e., those that were statistically signifi-
cant). To avoid this source of discrepancies, the author added a
sentence to the codebook stating that coders should compare the
overview of the statistical analyses with the entire set of analyses
reported in the Results section. The second set of omission dis-
agreements pertained to whether three interaction effect sizes from
one source should be coded. This set of disagreements arose
because the statistical tests of the interaction effects were carried
out on subsamples and the codebook did not provide guidance on
what to do in this situation. To avoid this source of discrepancies,
the author added a sentence to the codebook stating that two-way
interaction effects should be coded only when the analyses were
carried out on the total sample.
The item generating the most commission disagreements was
“number of types of covariates.” Discrepancies in coding this item
arose for two reasons. First, one coder classified the covariates
based on the labels provided by the authors, whereas the other
coder classified the covariates based on the items used to assess
them. Second, the definitions of the leisure, social connection, and
social interaction covariates were blurred. To eliminate disagree-
ments in coding number of types of covariates, we modified our
definitions of the leisure, social connection, and social interaction
covariates, and coders were instructed to examine the survey items
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
566
OKUN, YEUNG, AND BROWN
provided in the Method section as opposed to using the label
provided by the authors.
The first author coded the remaining nine sources.
Effect Sizes
Effect sizes consisted of estimates of the relation between
volunteering as measured via an item or items on a survey and
mortality assessed after a period of time elapsed, referred to as
the mortality surveillance period. We extracted hazard ratios
(HRs), odds ratios (ORs), and relative risks (RRs). We focused
on three types of effect sizes. Unadjusted effect sizes assessed
the magnitude of the relation between volunteering and mor-
tality risk in the absence of covariates. In contrast, adjusted
effect sizes assessed the magnitude of the relation between
volunteering and mortality risk in the presence of covariates.
Unadjusted and adjusted effect sizes were computed on the total
sample and on independent subsamples within a study (e.g.,
participants in poor health and participants in good health).
Interaction effect sizes assessed the magnitude of the joint
effect of volunteering and a moderator variable on mortality.
These two-way interaction effects consisted of terms formed by
multiplying scores on the volunteer variable by scores on a
moderator variable. For example, Okun, August, Rook, and
Newsom (2010) examined the joint effect of volunteering and
functional health limitations on mortality risk. As indicated
previously, two-way interaction effect sizes were extracted only
from the total sample and not from subsamples. For volunteer
by moderator variable interaction effects, we coded information
reported in the article regarding the p value associated with the
statistical test regardless of whether we were able to extract an
effect size.
Complexities of Data Analysis
We established a common metric for effect sizes by converting
RR estimates and OR estimates of effect sizes to HR estimates of
effect sizes. In the current meta-analysis, the HR was an index of
how often death occurred in a group of volunteers (or more
frequent volunteers) compared with how often death occurred in a
group of nonvolunteers (or less frequent volunteers) over time. An
RR and an OR were converted to HR using one or both of the
following equations (Zhang & Yu, 1998):
RROR / 关共1 r rOR兲兴 and
HR1n1 RRr /1n1 r,
where r is the death rate in the volunteering reference group.
When r was not provided in the source, we generated a predicted
value for r using the proportion of deceased in the total sample as
the predictor. The correlation between death rate in the total
sample and r was .984, and the prediction equation was Y
i
1.073X
i
.014.
Prior to conducting the inferential analyses, the HR effect sizes
were log transformed and weighted by the reciprocal of the con-
ditional variance. For ease of interpretation, we transformed sum-
mary statistics back into HRs prior to presentation. When the
conditional variance of the effect size was not available, we
generated a predicted value using the sample size associated with
the effect size as the predictor. The correlation between the sample
size associated with the effect size and its conditional variance was
.511, and the prediction equation was Y
i
–.00000253X
i
.039.
There are several sources of complexity that need to be taken
into account in conducting meta-analytic analyses on effect sizes
(Borenstein, Hedges, Higgins, & Rothstein, 2009). In the present
meta-analysis, we faced three sources of complexity: (a) effect
sizes published in different journal articles from the same data set,
(b) two or more unadjusted (or adjusted) effect sizes extracted
from the same source, and (c) partitioning the sources of variability
in effect sizes extracted from different studies.
In our meta-analysis, Lum and Lightfoot (2005) and Luoh and
Herzog (2002) both reported analyses conducted using the Asset
and Health Dynamics Among the Oldest Old data set. Similarly,
Harris and Thoresen (2005) and Sabin (1993) both reported anal-
yses conducting using the Longitudinal Study of Aging. Our
decision rule was to delete overlapping effect sizes from the source
that yielded the fewest effect sizes. Consequently, we excluded the
adjusted effect sizes extracted from the total samples from the Lum
and Lightfoot and Sabin studies.
Two or more effect sizes can be extracted from a source when
a researcher creates two or more volunteer-related predictor vari-
ables and examines their relations with mortality risk. For exam-
ple, Okun et al. (2010) reported unadjusted and adjusted effect
sizes in separate analyses of mortality risk in the total sample in
which (a) volunteering was coded as a dummy variable and (b)
volunteer frequency was coded as a continuous variable. In this
case, we extracted two unadjusted effect sizes and two adjusted
effect sizes from the source. Following the recommendation of
Borenstein et al. (2009, pp. 227–230), we created a synthetic effect
size and a synthetic variance for the unadjusted effect sizes and for
the adjusted effect sizes.
Several indices of between-studies variability in effect sizes
have been developed. The oldest and most commonly used index
of heterogeneity is the Q statistic. The Q statistic provides a test of
the null hypothesis that all studies share a common effect size. The
lack of a common effect size may indicate that each study has its
own true population effect size or it may be due to sampling error.
The main drawback of the Q statistic is that it does not partition the
variability observed among studies into random error and “real”
differences in the true effect sizes. To overcome this limitation, we
report two additional statistics related to variability—tau and I
2
.
Tau provides us with an estimation of the standard deviation of the
true effect sizes, which serves to contextualize the meaning of the
estimate of the population effect size. I
2
tells us what proportion of
the observed variance is due to differences in the true effect sizes.
As I
2
increases, the proportion of the observed variance that is real
increases. I
2
ranges from 0% to 100%, and it has been suggested
that 50% and 75% are benchmarks for moderate and high real
variation, respectively (Borenstein et al., 2009, p. 119). Moderate
and high values of I
2
indicate that it is worthwhile for researchers
to search for study characteristics that account for the variation in
effect sizes.
The meta-analytic analyses were conducted using Comprehen-
sive Meta-Analysis, Version 2.0 (Borenstein, Hedges, Higgins, &
Rothstein, 2006). Unless otherwise specified, we employed ran-
dom effect models that take into account the amount of variance
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
567
VOLUNTEERING AND MORTALITY
due to differences between studies as well as differences among
participants within studies.
Results
Study Characteristics
Information describing the 14 studies is provided in Table 1.
The articles were published over a span exceeding 25 years. Four
articles were published prior to 2000, seven articles were published
between 2000 and 2009, and the remaining three articles were
published between 2010 and 2012. Twelve different data sets were
analyzed, with the Longitudinal Study of Aging and the Asset and
Health Dynamics Among the Oldest Old each being analyzed in
two studies. Nine of the studies used U.S. samples, and the
remaining three studies employed Israeli (n 2) and Taiwanese
samples. The total sample sizes, which do not necessarily corre-
spond to the sample sizes associated with the effect sizes, ranged
from 868 to 15,938, with a median of 4,927.50. The minimum age
of the participants ranged from 55 to 75 years old, with a median
of 66.50 years old. The mean and standard deviation for length of
the mortality surveillance period in years were 5.94 and 1.86,
respectively.
Measurement and Coding of Volunteering
Table 2 summarizes the measures and coding of volunteering as
they pertain to the effect sizes extracted from the sources. Four
types of volunteer predictor variables were used by researchers
studying the relation between volunteering and mortality risk,
including (a) comparing nonvolunteers with volunteers, (b) assess-
ing individual differences in number of organizations volunteered
for, (c) individual differences in hours volunteered, and (d) indi-
vidual differences in frequency of volunteering.
Description of Unadjusted and Adjusted Effect Sizes
As can be seen in Table 3, we extracted 25 unadjusted effect
sizes. Twenty-one of the unadjusted effect sizes were derived from
the total samples of nine studies (aggregate N 49,320) and four
of the unadjusted effect sizes were derived from subsamples as-
sociated with two studies. The (unweighted) unadjusted effect
sizes derived from total samples ranged from 0.31 to 0.96, with a
median HR of 0.56. Unadjusted and adjusted HRs were coded such
that values greater than 1.00 indicated that volunteering was as-
sociated with an increased risk of dying, whereas values less than
1.00 indicated that volunteering was associated with a decreased
risk of dying. An HR effect size of 1.00 indicated that volunteering
was unrelated to risk of mortality. The variances associated with
the unadjusted effect sizes derived from total samples ranged from
0.00 to 0.06, with a median of 0.03.
Excluding the adjusted effects from the total samples for the
Lum and Lightfoot (2005) study and the Sabin (1993) study, we
extracted 31 adjusted effect sizes. Twenty-five of the adjusted
effect sizes were derived from the total samples of 11 studies
(aggregate N 49,400), and six of the adjusted effect sizes were
derived from subsamples associated with three studies. The (un-
weighted) adjusted effect sizes derived from total samples ranged
from 0.40 to 1.11, with a median HR of 0.80. The variances
associated with the adjusted effect sizes derived from total samples
ranged from 0.00 to 0.06, with a median of 0.03.
Forest Plot of Unadjusted Effect Sizes
We began our inferential analyses by constructing a forest plot
of the unadjusted effect sizes derived from total samples. In a
forest plot, each study as well as the summary effect are depicted
as a point estimate bounded by a confidence interval. As can be
seen from Table 3, synthetic unadjusted effect sizes and variances
were created for five of the studies: Harris and Thoresen (2005);
Konrath, Fuhrel-Forbis, Lou, and Brown (2012); Musick et al.
(1999); Okun et al. (2010); and Oman et al. (1999).
As can be seen in Figure 1, the confidence intervals for the nine
unadjusted effect sizes were all below 1.00. The weighted mean of
these effect sizes was 0.53, with a 95% confidence interval of 0.45
to 0.62. The p value associated with the weighted mean is less than
.001. Thus, in the absence of control variables, the average effect
size suggests that relative to nonvolunteers, volunteers have a 47%
decrease in the risk of death, with a 95% confidence interval of
38% to 55%.
Table 1
Description of Study Characteristics
Author(s)
Year of
publication Data set Country N
Minimum
age (years)
Ayalon 2008 Israeli Census Bureau Survey Israel 5,055 60
Gruenewald et al. 2007 MacArthur Study of Successful Aging United States 1,030 70
Harris & Thoresen 2005 Longitudinal Study of Aging United States 7,496 70
Hsu 2007 Survey of Health and Living Status of the Elderly Taiwan 2,825 60
Konrath et al. 2012 Wisconsin Longitudinal Study United States 10,317 68
Lee et al. 2011 Health and Retirement Survey United States 6,408 65
Lum & Lightfoot 2005 Asset and Health Dynamics Among the Oldest Old United States 7,322 70
Luoh & Herzog 2002 Asset and Health Dynamics Among the Oldest Old United States 4,860 75
Musick et al. 1986 American’s Changing Lives United States 1,211 65
Okun et al. 2010 Later Life Study of Social Exchanges United States 868 65
Oman et al. 1999 Marin County United States 1,972 55
Rogers 1996 National Health Interview Survey Supplement on Aging United States 15,938 55
Sabin 1993 Longitudinal Study of Aging United States 7,485 70
Shmotkin et al. 2003 Cross-Sectional and Longitudinal Aging Study Israel 1,343 75
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
568
OKUN, YEUNG, AND BROWN
Heterogeneity of the Unadjusted Effect Sizes
The Q test statistic with 8 degrees of freedom was 44.17, p
.001, indicating that the effect sizes are heterogeneous. Tau equals
.22, which given a weighted average effect size of 0.53 means that
the distribution of true effects is likely to include effect sizes
ranging from 0.31 to 0.75. The value of I
2
was 82%, indicating that
a large proportion of the observed variance reflects differences in
the true effect sizes across studies.
Forest Plot of Adjusted Effect Sizes
Prior to analyzing the adjusted effect sizes, it is useful to note
the frequency with which various types of variables were used as
covariates. The percentages that each type of covariate was used
ranged from 18% (leisure) to 100% (age, sex, and physical health).
For the remaining type of covariates, the percentages of use were
socioeconomic status (91%), health behaviors (91%), marital sta-
tus (73%), religiosity/religious behavior (64%), emotional health
(64%), social connection (64%), social interaction (64%), ethnicity
(55%), work status (55%), and cognitive functioning (27%). As
can be seen from Table 3, synthetic adjusted effect sizes and
variances were created for six of the studies: Harris and Thoresen
(2005); Konrath et al. (2012); Musick et al. (1999); Okun et al.
(2010); Oman et al. (1999); and Shmotkin, Blumstein, and Modan
(2003). As can be seen in Figure 2, the confidence intervals for
seven of the 11 adjusted effect sizes derived from total samples
were below 1.00. The weighted mean of these effect sizes was
0.76, with a 95% confidence interval of 0.69 to 0.84. The p value
associated with the weighted mean is less than .001. Thus, in the
presence of control variables, the average effect size indicates that,
relative to nonvolunteers, volunteers have a 24% decrease in the
risk of death, with a 95% confidence interval of 16% to 31%.
Heterogeneity of the Adjusted Effect Sizes
The Q test statistic with 10 degrees of freedom was 24.42, p
.01, indicating that the effect sizes are heterogeneous. Tau equals
.11, which given a weighted average effect size of 0.76 means that
the distribution of true effects is likely to include effect sizes
ranging from 0.54 to 0.98. The value of I
2
was 59%, indicating that
a moderate proportion of the observed variance is real rather than
spurious.
Comparison of Adjusted and Unadjusted Effect Sizes
To examine the reduction in the relation between volunteering
and mortality associated with the introduction of control variables,
we used matched pairs of effect sizes from the nine studies that
yielded both adjusted and unadjusted effect sizes. We created a
difference score for each study by subtracting its unadjusted effect
size from its adjusted effect size. For example, in the Rogers
(1996) study, the adjusted and unadjusted effect sizes were 0.81
Table 2
Measurement and Coding of Volunteering Relevant to Effect Sizes
Author(s) Measure of volunteering Coding of volunteer variables
Ayalon, 2008 Volunteering within an organization Yes vs. no
Okun et al., 2010 How often volunteered in the past month: Response options ranged
from never or almost never (coded 0) to daily (coded 5).
Frequency of volunteering was treated as a
continuous variable and as a binary variable
(never or almost never vs. all other response
options combined).
Harris & Thoresen, 2005 How often did volunteer work: Response options included
never, rarely, sometimes, and frequently.
Three dummy variables were formed (rarely,
sometimes, and frequently) with never as the
reference group.
Hsu, 2007 Did volunteer work Yes vs. no
Lum & Lightfoot, 2005 Hours volunteered in the past year 0 to 99 hr vs. 100 hr or more
Rogers, 1996 Did volunteer work in the community Yes vs. no
Gruenewald et al., 2007 Volunteered in the past year Yes vs. no
Shmotkin et al., 2003 Volunteered with an organization and frequency of volunteering:
Response options included several times a week, several times
a month, less than several times a month, and did not answer
frequency question.
Yes vs. no plus four dummy variables (several times
a week, several times a month, less than several
times a month, and did not answer frequency
question) with nonvolunteer as the reference
group
Musick et al., 1986 Volunteered in the past year for religious, school, political,
senior citizen, and “other” organizations and hours
volunteered.
Two sets of dummy variables (1 organization and
2 organizations vs. nonvolunteer and less than
40 hr and 40 hr or more vs. nonvolunteers)
Oman et al., 1999 Number of organizations involved with as a volunteer and hours
volunteered per week
Two sets of dummy variables (1 organization and
2 organizations vs. nonvolunteer and less than 4
hr and 4 hr or more vs. nonvolunteers)
Lee et al., 2011 Spent time doing volunteer work for religious, educational,
health-related, or other charitable organization
Yes vs. no
Luoh & Herzog, 2002 Hours volunteered in the past year 0 to 99 hr vs. 100 hr or more
Sabin, 1993 Did volunteer work in the past 12 months Yes vs. no
Konrath et al., 2012 Volunteered in the past 10 years, regularity of volunteering in
the past 10 years (0 not at all to 3 volunteered regularly
the whole time), and hours volunteered per month during the
past year
Yes vs. no for volunteering in past 10 years,
continuous measures of regularity of volunteering,
and hours volunteered
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
569
VOLUNTEERING AND MORTALITY
and 0.50, respectively. Thus, the difference score for this study
was 0.31. Using the formula provided by Borenstein et al. (2009,
p. 228), we calculated the variance of the differences in effect sizes
for each study.
Forest Plot of the Difference Between the Adjusted
and Unadjusted Effect Sizes
As can be seen in Figure 3, the difference between the adjusted
and unadjusted effect sizes ranged from 0.07 to 0.32. The weighted
mean of the difference between the adjusted and unadjusted effect
sizes derived from total samples was 0.20, with a 95% confidence
interval of 0.16 to 0.25. The p value associated with the weighted
mean is less than .001. Thus, the HR, on average, increases by 0.20
when the adjusted and unadjusted effect sizes are directly com-
pared across the set of nine studies. Keeping in mind that for effect
sizes below 1.00 larger values indicate smaller effects, the mag-
nitude of the relation between volunteering and mortality risk is
significantly (p .001) reduced by the inclusion of covariates.
Heterogeneity of the Difference Between the Adjusted
and Unadjusted Effect Sizes
With 8 degrees of freedom, the Q test statistic was 54,693.49,
p .001, indicating that the differences between the adjusted and
unadjusted effect sizes are heterogeneous. Tau equals .34. The
value of I
2
was 99.9%, indicating that virtually all of the observed
variance is real rather than spurious.
What Is the Form of the Association Between
Volunteering and Mortality Risk?
To ascertain whether the association between volunteering and
mortality risk is linear or nonlinear, we examined whether volun-
teer predictor variable was related to unadjusted and adjusted
effect sizes. Because these studies focused on within-study differ-
ences in effect sizes as a function of volunteer predictor variable,
we used fixed effects models. Table 4 summarizes the results of
these analyses. Of the 10 comparisons, only two were statistically
significant. On the one hand, in the Musick et al. (1999) study, the
adjusted effect size for volunteering for one organization (0.60)
was stronger than the adjusted effect size for volunteering for two
or more organizations (1.11), suggesting a curvilinear relation. On
the other hand, in the Oman et al. (1999) study, the unadjusted
effect size for volunteering for one organization (0.74) was weaker
than the unadjusted effect size for volunteering for 2 or more
organizations (0.37), suggesting a linear relation. The results of
these analyses do not provide a warrant for drawing a firm con-
clusion regarding whether the volunteering–mortality risk associ-
ation is linear or curvilinear.
Do Adjusted Effect Sizes Vary With Moderator
Variables?
We examined the relation between two categorical moderator
variables—study focus and country—and effect sizes using the Q
B
statistic. We focused on adjusted effect sizes from total samples
(N 11) because these estimates should be more precise than
estimates based on unadjusted effect sizes. Nine of the studies
focused on the volunteering–mortality risk association and the
remaining two studies did not. Also, nine of the studies were
conducted in the United States, whereas the remaining two studies
were conducted in Israel. Neither test was significant (lowest p
.39).
We examined the relation between five quantitative moderator
variables and adjusted effect sizes using metaregression with a
mixed-effects model estimated using the method of moments. The
five moderator variables were (a) journal impact factor (M 2.48,
SD 0.81), (b) year of publication (Mdn 2005, SD 5.37), (c)
minimum age of sample (M 65.73 years, SD 6.92, (d)
percentage of sample deceased (M 24.50, SD 15.70), and (e)
percentage of sample volunteering (M 24.56, SD 13.02).
None of the regression coefficients were statistically significant
(ps .05).
Table 3
Unadjusted and Adjusted Effect Sizes and Variances
Author(s)
Unadjusted effect
sizes
(and variances)
Adjusted effect
sizes
(and variances)
Ayalon, 2008 0.50 (0.03) 0.77 (0.04)
Gruenewald et al., 2007 0.72 (0.05)
Harris & Thoresen, 2005 0.59 (0.04) 1.01 (0.04)
0.58 (0.01) 0.71 (0.00)
0.47 (0.01) 0.81 (0.01)
Hsu, 2007 0.81
a
(0.03)
2.28
a
(0.03)
Konrath et al., 2012 0.53 (0.03) 0.63 (0.05)
0.74 (0.01) 0.97 (0.00)
0.96 (0.00) 0.84 (0.01)
0.92
a
(0.07)
0.37
a
(0.04)
Lee et al., 2011 0.41 (0.01) 0.68 (0.01)
0.41
a
(0.01)
0.65
a
(0.01)
0.75
a
(0.02)
0.91
a
(0.02)
Lum & Lightfoot, 2005 0.67
b
(0.00)
Luoh & Herzog, 2002 0.31 (0.03) 0.40 (0.03)
Musick et al., 1986 0.40 (0.03) 0.60 (0.03)
0.65 (0.03) 1.11 (0.03)
0.46 (0.03) 0.70 (0.03)
0.58 (0.03) 0.93 (0.03)
Okun et al., 2010 0.86 (0.00) 1.00 (0.00)
0.56 (0.03) 0.82 (0.03)
Oman et al., 1999 0.58 (0.02) 0.80 (0.02)
0.74 (0.02) 0.94 (0.02)
0.37 (0.06) 0.56 (0.06)
0.69 (0.03)
0.49 (0.04)
Rogers, 1996 0.50 (0.02) 0.81 (0.00)
Sabin, 1993 0.58
c
(0.02)
0.53
a
(0.03)
0.80
a
(0.03)
Shmotkin et al., 2003 0.67 (0.02)
0.62 (0.03)
0.60 (0.03)
0.86 (0.03)
0.96 (0.03)
a
Extracted from subsample.
b
Excluded from analysis because of overlap
with adjusted effect size extracted from Luoh and Herzog (2002). Excluded
from analysis because of overlap with adjusted effect size extracted from
Harris and Thoresen (2005).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
570
OKUN, YEUNG, AND BROWN
Comparison of Effect Sizes Derived From Independent
Subsamples Within Studies
For independent subsamples within studies, we used the Q test
with fixed effects to compare five pairs of effect sizes. As can be
seen in Table 5, three comparisons were statistically significant
(p .01). In the Lee, Steinman, and Tan (2011) study, the
unadjusted relation between volunteering and mortality risk was
stronger among nondrivers/limited drivers than regular drivers. In
the Hsu (2007) study, the adjusted relation between volunteering
and mortality risk was inverse among Taiwanese men (0.81) but
positive and stronger among Taiwanese women (2.28). In the
Konrath et al. (2012) study, the unadjusted relation between vol-
unteering and mortality risk was stronger among participants who
were primarily motivated to volunteer by concerns for others as
opposed to concerns for self.
Volunteering by Moderator Variable Interaction
Effect Sizes
In five of the studies, researchers tested for volunteering by mod-
erator variable interaction effects. In tests of interaction effects, co-
variates were included in the model, along with the main effects of the
volunteering and moderator variable and one or more interaction
terms. Table 6 provides a summary of the measurement and coding of
the moderator variables. Of the 34 tests of volunteering by moderator
variable interaction effects, four tests involved measures of leisure,
four tests involved measures of religiosity, seven tests used measures
of health, eight tests used measures of social connection, nine tests
employed measures of social interaction, and sex was the moderator
variable in the remaining two tests.
As can been seen in Table 7, we extracted 11 volunteering by
moderator interaction effect sizes from four of the five studies
(aggregate N 5,226). Interaction effect sizes were coded such
that HR values greater than 1.00 indicated that the relation
between volunteering and mortality risk increased as the mod-
erator variable decreased, whereas HR values less than 1.00
indicated that the relation between volunteering and mortality
risk increased as the moderator variable increased. An HR
interaction effect size of 1.00 indicated that the relation be-
tween volunteering and mortality risk did not vary across levels
of the moderator variable. Interaction effect sizes greater than
1.00 are consistent with the compensation hypothesis, whereas
interaction effect sizes less than 1.00 are consistent with the
complementary hypothesis.
The (unweighted) interaction effect sizes ranged from 0.37 to
2.44. Seven of the volunteering by moderator variable interaction
effect sizes were below 1.00, and the remaining four were above
1.00. The variances associated with the volunteering by moderator
Figure 1. Unadjusted effect sizes. Note. For effect sizes less than 1.00, smaller values indicate larger, beneficial
effects of volunteering on mortality risk.
Figure 2. Adjusted effect sizes. Note. For effect sizes less than 1.00, smaller values indicate larger, beneficial
effects of volunteering on mortality risk.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
571
VOLUNTEERING AND MORTALITY
variable effect sizes ranged from 0.00 to 0.13, with a median of
0.03. Because sufficient information to extract effect sizes was
reported for only 11 of the 34 volunteering by moderator variable
interaction effects and because the moderator variables were di-
verse, we did not carry out inferential meta-analytic statistical
techniques on the summary effect size.
Confidence intervals were generated for each of the 11 in-
teraction effect sizes. As indicated in Table 7, the 11 confidence
intervals were significant (p .05). For all three volunteering
by religiosity interaction effect sizes and for all three volun-
teering by social connection interaction effect sizes, the entire
confidence intervals were below 1.00, which supports the
complementary hypothesis. Similarly, for the social interaction
effect size, the entire confidence interval was below 1.00.
In contrast, for the two volunteering by health interaction
effect sizes and for the volunteering by social interaction effect
size, all three confidence intervals were entirely above 1.00,
supporting the compensatory hypothesis. For the two effect
sizes involving leisure as the moderator variable, one confi-
dence interval was entirely above 1.00, whereas the other
confidence interval was entirely below 1.00. Finally, regardless
of whether we were able to extract a volunteering by moderator
variable interaction effect size, we extracted information pro-
vided in the source regarding the p value associated with the
statistical test of the volunteering by moderator variable inter-
action effect. When researchers did not report exact p values,
we included information from the source regarding whether the
p values were less than a specified value or greater than a
specified value. Of the 34 p values, 10 were less than .05. Only
two of the p values less than .05 were associated with tests of
Figure 3. Difference between adjusted and unadjusted effect sizes. Note. As the difference scores increase,
adjusting for covariates reduces the volunteering–mortality association to a greater degree.
Table 4
Relation Between Volunteer Predictor Variable and Effect Sizes
Author(s) Volunteer predictor Effect size df Q
Harris & Thoresen, 2005 Rarely 0.59 (U) 2 2.95
Sometimes 0.58 (U)
Frequently 0.47 (U)
Harris & Thoresen, 2005 Rarely 1.01 (A) 2 3.97
Sometimes 0.71 (A)
Frequently 0.81 (A)
Shmotkin et al., 2003 Several times a month 0.60 (A) 2 3.46
Several times a month 0.86 (A)
Several times a week 0.96 (A)
Musick et al., 1986 1 organization 0.40 (U) 1 3.39
2 organizations 0.65 (U)
Musick et al., 1986 1 organization 0.60 (A) 1 5.44
2 organizations 1.11 (A)
Musick et al., 1986 40 hr per week 0.46 (U) 1 0.77
40 hr per week 0.58 (U)
Musick et al., 1986 40 hr per week 0.70 (A) 1 1.16
40 hr per week 0.93 (A)
Oman et al., 1999 1 organization 0.74 (U) 1 6.21
2 organizations 0.37 (U)
Oman et al., 1999 1 organization 0.94 (A) 1 3.36
2 organizations 0.56 (A)
Oman et al., 1999 1–3 hr per week 0.69 (U) 1 1.79
4 hr per week 0.49 (U)
Note. (A) adjusted; (U) unadjusted.
p .05.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
572
OKUN, YEUNG, AND BROWN
interaction effects that did not yield effect sizes. In the Harris
and Thoresen (2005) study, religiosity and one of the social
interaction variables amplified the inverse relation between
volunteering and mortality risk.
The Robustness of Conclusions Regarding Sets of
Effect Sizes
We examined the robustness of the conclusions drawn regarding
the unadjusted effect sizes and the adjusted effect sizes using the
Duval and Tweedie (2000) trim and fill procedure, which provides
an estimate of the unbiased mean effect size. For the interaction
effect sizes, we used the Fisher’s exact test (Agresti, 1992)to
determine whether researchers were less likely to report effect
sizes when the p values associated with the statistical test were
greater than .05.
Unadjusted and adjusted effect sizes. Application of the
trim and fill procedure separately for the unadjusted effect sizes
and the adjusted effect sizes revealed that the means of the distri-
butions of effect sizes did not change, indicating the absence of
publication bias.
Interaction effect sizes. The Fisher’s exact test revealed that
interaction effect sizes were significantly (p .001) more likely to
be reported when the p values associated with the statistical tests
were less than .05 (80%) as compared with when the p values
associated with the statistical tests were greater than .05 (12.5%).
This association indicates that researchers are biased toward re-
porting interaction effect sizes that achieve conventional levels of
statistical significance. Thus, the volunteering by moderator vari-
Table 5
Comparison of Effect Sizes From Within-Study
Independent Samples
Author(s) Independent sample
Effect
size df Q
Lee et al., 2011 Nondrivers/limited drivers 0.41 (U) 1 11.78
ⴱⴱⴱ
Regular drivers 0.75 (U)
Lee et al., 2011 Nondrivers/limited drivers 0.65 (A) 1 2.85
Regular drivers 0.91 (A)
Sabin, 1993 Poor health 0.80 (A) 1 2.98
Good health 0.53 (A)
Hsu, 2007 Women 2.28 (A) 1 15.46
ⴱⴱⴱ
Men 0.81 (A)
Konrath et al.,
2012
Motivated by concern for
others
0.37 (U) 1 7.76
ⴱⴱ
Motivated by concern for
self
0.92 (U)
Note. (A) adjusted; (U) unadjusted.
ⴱⴱ
p .01.
ⴱⴱⴱ
p .001.
Table 6
Measurement and Coding of Moderator Variables
Author(s) Measure of moderator Coding of moderator
Harris & Thoresen, 2005 Attended sporting or other event Yes vs. no
Attended religious services Yes vs. no
Sex Male vs. female
Living alone Yes vs. no
Living with spouse Yes vs. no
Visited senior center Yes vs. no
Visited with friends/neighbors Yes vs. no
Visited with family Yes vs. no
Musick et al., 1986 Living alone Yes vs. no
Frequency of talking with friends, neighbors or relatives, and
frequency of getting together with friends and relatives
Talking on scale from 1 (never)to6(more than once
a day) and getting together on scale from 1 (never)
to6(once a week)
Okun et al., 2010 Functional health limitations Limitations on scale from 0 (not at all)to3(very
difficult)
Number of health conditions Number of health conditions out of 12
Self-related health Self-related health from 0 (poor)to1(excellent)
Oman et al., 1999 Number of leisure activities out of eight 1 3 or more leisure activities; 0 2 or fewer
leisure activities
Attended religious services weekly Yes vs. no
Attended religious services at all Yes vs. no
Attended other religious group activities monthly Yes vs. no
Sex Male vs. female
Feels close to 3 or more friends, feels close to 3 or more
relatives, and sees 3 close friends or relatives
0to3
Living with others Yes vs. no
Get out of house everyday Yes vs. no
Participate in organizational group activities Yes vs. no
Shmotkin et al., 2003 Frequency of physical activities including walking,
gardening, and any sport.
1 not at all to 4 3 or more times a week
Have a hobby Yes vs. no
Frequency of passive activities (e.g., watching TV), talking
with family and friends, going out to do something, and
playing cards or another game
0 never to 3 every day
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
573
VOLUNTEERING AND MORTALITY
able interaction effect sizes included in the current meta-analysis
overestimate the magnitude of the joint effect of volunteering and
moderator variables on mortality risk.
Discussion
This is the first meta-analysis of the volunteering–mortality
association and as such provides strong evidence in favor of the
growing consensus that helping others yields health benefits for
the helper. Across 11 studies, volunteerism appeared to reduce
mortality risk by almost half in unadjusted models when variables
that likely mediate the effect are not first removed from the
analysis. When the more conservative test is applied, one that
controls for covariates such as age, sex, ethnicity, socioeconomic
status, work status, marital status, religiosity, emotional health,
health behaviors, social connection, social interaction, and physi-
cal health, the adjusted effect size remains substantial, predicting a
25% reduction in the risk of death. Furthermore, we detected no
evidence of publication bias for our estimates of the means of the
unadjusted and adjusted effect sizes.
However, researchers did exhibit a bias toward reporting inter-
action effect sizes only when the tests of the interaction effects
reached conventional levels of statistical significance. Thus, it is
premature to draw conclusions regarding the merits of the com-
plementary and compensatory hypotheses. Keeping this caveat in
mind, our analyses revealed that religious involvement appears to
amplify the association between volunteering and mortality risk.
Consistent with the complementary hypothesis, the greater re-
sources derived from religious involvement enhance the health-
related benefits of volunteering. In fact, all four estimates of the
volunteering by public religiosity interaction effect on mortality
risk (see Table 7) were significant (highest p .05). In trying to
understand this effect further, we think it is reasonable to consider
whether religiosity confers cultural capital or reflects more altru-
istic values.
Wilson and Musick (1997) identified altruistic values as a
resource that contributes to cultural capital. Konrath et al. (2012)
found that volunteering reduced mortality risk only among older
adults motivated primarily by a concern for others rather than a
concern for oneself. Furthermore, Pargament (1997) posited that
involvement in public religious activities is associated with stron-
ger motivation to engage in actions that benefit humanity. Thus,
publically religious older adults may benefit more from volunteer-
ing in terms of reduction of mortality risk than their nonpublically
religious peers because they are more motivated to volunteer by
Table 7
Interaction Effect Sizes and p Values
Author(s) Moderator variable N Effect size (and variance) Hypothesis supported p
a
Harris & Thoresen, 2005 Leisure 7,496 .05
Religiosity 7,496 Complementary .05
Sex 7,496 .05
Social connection 7,496 .05
Social connection 7,496 .05
Social interaction 7,496 .05
Social interaction 7,496 Complementary .05
Social interaction 7,496 .05
Musick et al., 1986 Social connection 1,211 0.51 (0.03)
b
Complementary .10
Social connection 1,211 0.55 (0.03)
b
Complementary .10
Social connection 1,211 .05
Social connection 1,211 .05
Social interaction 1,211 1.88 (0.03)
c
Compensation .05
Social interaction 1,211 .05
Social interaction 1,211 .05
Social interaction 1,211 .05
Okun et al., 2010 Health 868 1.05 (0.00)
c
Compensation .06
Health 868 2.44 (0.10)
c
Compensation .01
Health 868 .12
Health 868 .12
Health 868 .12
Health 868 .12
Oman et al., 1999 Leisure 1,973 1.75 (0.07)
c
Compensation .05
Religiosity 1,973 0.40 (0.10)
b
Complementary .01
Religiosity 1,973 0.37 (0.13)
b
Complementary .01
Religiosity 1,973 0.66 (0.04)
b
Complementary .05
Sex 1,973 .15
Social connection 1,973 0.69 (0.03)
b
Complementary .05
Social connection 1,973 .05
Social interaction 1,973 .05
Social interaction 1,973 .05
Shmotkin et al., 2003 Health 1,174 .05
Leisure 1,174 .05
Leisure 1,174 0.56 (0.03)
b
Complementary .04
a
p refers to the probability associated with a test of statistical significance for an interaction effect that was provided in the article by the authors.
b
Entire
confidence interval is below 1.00.
c
Entire confidence interval is above 1.00.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
574
OKUN, YEUNG, AND BROWN
other-oriented motives, which have been theorized to be beneficial
for physical health (S. L. Brown et al., 2012).
Our results also reveal substantial differences among studies,
reflected in the heterogeneity of the unadjusted and adjusted effect
sizes. In other words, the volunteering–mortality risk effect sizes
reflect systematic differences among the studies and not just sam-
pling error. Rather than conceiving the effect sizes extracted from
the studies as variation around a single (common) population
effect size, each study should be viewed as having its own popu-
lation effect size that gives rise to a distribution of population
effect sizes.
Unfortunately, we were unable to account for variation in effect
size magnitude. One possibility is that the analyses of predictors of
effect size magnitude were statistically underpowered. Thus, we
cannot draw firm conclusions from our metaregressions, including
the form of the relation between amount of volunteering and
mortality risk. Additional research is warranted that tests the
notion that high levels of volunteering are detrimental to the health
of the volunteers (Musick et al., 1999).
Limitations and Guidelines for Future Studies
Our review has several limitations. First, it was limited by the
small, published literature and a relatively few “file draw” studies
with null findings would negate the associations that we observed
between volunteering and mortality risk. Second, the studies used
nonexperimental designs. Although researchers, on average, con-
trolled for more than nine types of covariates, these efforts do not
permit us to draw conclusions regarding the causal impact of
volunteering on mortality (von Bonsdorff & Rantanen, 2011). The
third limitation of studies examining the volunteering–mortality
relation has been the lack of standardization of volunteer predictor
variables. For example, the lack of consistency in assessing and
coding frequency of volunteering and hours volunteered makes it
more difficult to establish the form of the relation between volun-
teering and mortality risk. Finally, some researchers did not (a)
report both unadjusted and adjusted effect sizes, (b) test for inter-
action effects when reporting results for independent subsamples,
and (c) provide effect sizes when testing interaction effects.
To rectify these limitations, in future studies of the relation
between volunteering and mortality risk, researchers should (a)
employ experimental designs, (b) assess frequency of volunteering
and hours volunteered using clearly specified anchors (i.e., once a
month rather than occasionally) and collect objective as well as
self-report data, (c) report unadjusted as well as adjusted effects
from nonexperimental studies, (d) test interaction terms instead of
carrying out separate tests of the effects of volunteering within
each subsample, and (e) provide confidence intervals and HRs (or
ORs) for all main and joint effects that are tested.
Agenda for Future Research
We hope that the current meta-analysis inspires a new genera-
tion of research on the volunteering–mortality risk association that
(a) explores individual differences in volunteer-related variables,
(b) unpacks and tests causal mechanisms, and (c) expands the
range of prosocial behaviors.
Individual differences in volunteer-related variables. The
limited research conducted on who benefits the most from volun-
teering in terms of mortality risk reduction has largely ignored
volunteer-related variables. Konrath et al. (2012) found that vol-
unteers who were primarily motivated by self-oriented reasons did
not live longer than nonvolunteers. We believe that additional
volunteer-related individual differences variables should be exam-
ined as moderator variables. More specifically, the positive impact
of volunteering on health outcomes may vary with variables such
as volunteer work autonomy, efficacy, and mattering. In other
words, the health-related benefits of volunteering may be negated
when volunteers do not derive a sense of control, competence, and
making a difference from their unpaid work.
Unpacking and testing causal mechanisms. To make prog-
ress in unraveling the mystery of how volunteering reduces mor-
tality risk, it will be important to take advantage of integrative,
theoretical models that have been advanced in related research.
Specifically, a “caregiving system” model that is grounded in
evolutionary biology, neuroscience, social psychology, and attach-
ment theory has recently been advanced that proposes mechanisms
that link prosocial behavior with mortality risk (S. L. Brown et al.,
2012; S. L. Brown & Preston, 2012). This framework integrates
animal models of parenting (Numan, 2006) with human neuroim-
aging studies of parental responses to specify the triggers of
prosocial behavior and conditions that favor beneficial versus
harmful effects of prosocial behavior. This model suggests that
perceptions of another’s need in combination with the ability to
meet the need trigger the motivation to help, which in turn acti-
vates neural circuits related to parenting that release hormones
such as oxytocin and progesterone, both of which regulate stress
and down-regulate inflammation. Critically, situational (recipient,
interpersonal, organizational, and cultural) characteristics such as
authenticity of need or interdependence with the recipient are
hypothesized to interact with personal resources to generate either
intrinsic motives to help (i.e., mediated by hypothalamic pro-
cesses; Numan, 2006) or to generate extrinsic motives to help that
bypass other-regarding emotions and lengthen potential exposure
to harmful levels of chronic stress and inflammation.
Thus, the caregiving system model suggests that volunteering
can be mediated by neural circuitry that activates natural tenden-
cies we all have to be caring toward others. Whether volunteering
will produce health benefits is thought to depend not entirely on
whether resources exist to give, but also the extent to which the
signals for need are authentic—that is, they occur in the context of
a trusting relationship, trusted organization, or cultural norms that
minimize the possible threat of exploitation.
Expanding the range of prosocial behaviors. Finally, we
advocate that investigators examine the health consequences of
other types of prosocial behavior in addition to organizational
volunteering. The majority of studies that demonstrate morality
benefits associated with prosocial behavior were not initially de-
signed to examine the health effects of providing support, but were
instead designed to investigate the health effects of receiving
support (S. L. Brown et al., 2003). As such, systematic investiga-
tions into the health consequences of helping others are rare. In the
absence of these efforts, researchers have reanalyzed existing data,
which is supportive but does not lend itself to meta-analytic
techniques, tests of mechanisms, tests of causal relationships, or
tests of boundary conditions. Given the strong evolutionary bio-
logical theoretical underpinnings of integrative approaches to
prosocial and caregiving behavior (S. L. Brown et al., 2012),
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
575
VOLUNTEERING AND MORTALITY
attempts to formally and systematically examine the health con-
sequences of helping behaviors within close relationships are
likely to be informative and relevant and may suggest important
caveats for translating basic research on volunteerism to health
policy.
Implications for Public Health
The baby boomers pose a major challenge, and innovative
changes are required to sustain our system of public health, in-
cluding finding ways to keep them as healthy as possible and to
integrate them into the fabric of our communities (Knickman &
Snell, 2002). Volunteering has been described as a win–win ac-
tivity because of the benefits derived by both the recipients
(Wheeler, Gorey, & Greenblatt, 1998) and the providers (Post &
Neimark, 2007).
The results of this meta-analysis suggest that it is no longer a
question of whether volunteering is predictive of reduced mortality
risk; rather, our results suggest that the volunteering–mortality
association is reliable, and that the magnitude of the relationship is
sizable. The findings of the current study are bolstered by research
using a true experimental design to investigate the health-related
benefits of volunteering. In a comparison of Experience Corps
program volunteers with wait-listed controls, Fried et al. (2004)
showed that whereas the control group exhibited decreases in
strength, the volunteering group exhibited increases in strength. In
a more recent study using cognitively at-risk volunteers in the
Experience Corps program, Carlson et al. (2009) demonstrated that
relative to participants in the control group, participants in the
intervention group who received training in general literacy sup-
port, library support, and conflict resolution exhibited more cog-
nitive activity in the left prefrontal cortex and anterior cingulate
cortex (related to empathy) during a selective attention task.
At the same time that the health-related benefits of volunteering
have been documented, forecasts suggest that there will be a severe
shortage of volunteers (Gottlieb & Gillespie, 2008). Given these
circumstances, strategies should be identified to encourage older
adults to volunteer. For example, online volunteering activities are
expanding the range of opportunities available to healthy older
adults as well as to older adults with functional limitations (Cra-
vens, 2003), although there may be boundary conditions on whom
and under what circumstances volunteering has a salutary effect on
health. In addition to interventions that focus on volunteering, it is
also possible to leverage its benefits by incorporating volunteering
into psychosocial interventions with other foci. For example, in-
terventions that target family members who are caregivers can
provide opportunities for them to serve as peer mentors for novice
caregivers (Pillemer, Suitor, Landreneau, Henderson, & Brang-
man, 2000).
Ultimately, the possibility that volunteering reduces mortality
risk is exciting and a mystery. Our results suggest that it is now
permissible, desirable, and even necessary for researchers to begin
to delve into this mystery. What we discover may do more than
inform health policy and volunteerism. The complex and intricate
systems of the body that account for the volunteering-mortality
association suggest that what we discover may tell us something
bigger about disease, the aging process itself, and/or how behavior,
perception, and motivation are instantiated in the body.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
Agresti, A. (1992). A survey of exact inference for contingency tables.
Statistical Science, 7, 131–153. doi:10.1214/ss/1177011454
Ayalon, L. (2008). Volunteering as a predictor of all-cause mortality:
What aspects of volunteering really matter? International Psychogeri-
atrics, 20, 1–14. doi:10.1017/S1041610208007096
Belloc, N. B., & Breslow, L. (1972). Relationship of physical health status
and health practices. Preventive Medicine, 1, 409 421. doi:10.1016/
0091-7435(72)90014-X
Blakely, T. R., Atkinson, J., Ivory, V., Collings, S., Wilton, J., & Howden-
Chapman, P. (2006). No association of neighborhood volunteerism with
mortality in New Zealand: A national multilevel cohort study. Interna-
tional Journal of Epidemiology, 35, 981–989. doi:10.1093/ije/dy1088
Borenstein, M., Hedge, L. V., Higgins, J. P. T., & Rothstein, H. R. (2006).
Comprehensive meta-analysis (Version 2.0) [Computer software].
Englewood, NJ: Biostat.
Borenstein, M., Hedge, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009).
Introduction to meta-analysis. West Sussex, UK: Wiley.
Brown, S. L., & Brown, R. M. (2006). Selective investment theory:
Recasting the functional significance of close relationships. Psycholog-
ical Inquiry, 17, 1–29. doi:10.1207/s15327965pli1701_01
Brown, S. L., Brown, R. M., House, J., & Smith, D. (2008). Coping with
spousal loss: The potential buffering effects of self-reported helping
behavior. Personality & Social Psychology Bulletin, 34, 849 861. doi:
10.1177/0146167208314972
Brown, S. L., Brown, R. M., & Preston, S. (2012). The human caregiving
system: A neuroscience model of compassionate motivation and behav-
ior. In S. Brown, R. Brown, & L. Penner (Eds.), Moving beyond
self-interest: Perspectives from evolutionary biology, neuroscience, and
the social sciences (pp. 75– 88). New York, NY: Oxford University
Press.
Brown, S. L., Nesse, R. M., Vinokur, A., & Smith, D. (2003). Providing
social support may be more beneficial than receiving it: Results from a
prospective study of mortality. Psychological Science, 14, 320 –327.
doi:10.1111/1467-9280.14461
Brown, S., & Okun, M. A. (in press). Using the caregiver system model to
explain the resilience-related benefits that older adults derive from
volunteering. In M. Kent, M. Davis, & J. Reich (Eds.), The resilience
handbook: Approaches to stress and trauma. New York, NY: Routledge.
Brown, S. L., & Preston, S. (2012). Mechanisms linking prosocial behavior
to physical health. Manuscript submitted for publication.
Brown, S. L., Smith, D. M., Schulz, R., Kabeto, M., Ubel, P., Yee, J.,
. . . Langa, K. (2009). Caregiving and decreased mortality in a national
sample of older adults. Psychological Science, 20, 488 494. doi:
10.1111/j.1467-9280.2009.02323.x
Brown, W. M., Consedine, N. S., & Magai, C. (2005). Altruism relates to
health in an ethnically diverse sample of older adults. The Journals of
Gerontology, Series B: Psychological Sciences and Social Sciences, 60,
143–152. doi:10.1093/geronb/60.3.P143
Carlson, M. C., Erickson, K. I., Kramer, A. F., Voss, M. W., Bolea, N.,
Mielke, M.,...Fried, L. P. (2009). Evidence of neuroplasticity in at-risk
older adults: The Experience Corps program. The Journals of Geron-
tology, Series A: Biological Sciences and Medical Sciences, 64, 1275–
1282. doi:10.1093/Gerona/glp117
Cravens, J. (2003). Online mentoring: Program and suggested practices as
of February 2001. Journal of Technology in Human Services, 21, 85–
109. doi:10.1300/J017v21n01_05
Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel-plot–
based method of testing and adjusting for publication bias in meta-
analysis. Biometrics, 56, 455– 463. doi:10.1111/j.0006-341X.2000
.00455.x
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
576
OKUN, YEUNG, AND BROWN
Fried, L. P., Carlson, M. C., Freedman, M., Frick, K. D., Glass, T. A., Hill,
J.,...Zeger, S. (2004). A social model for health promotion for an aging
population: Initial evidence on the Experience Corps model. Journal of
Urban Health, 81, 64 –78. doi:10.1093/jurban/jth094
Gottlieb, B., & Gillespie, A. A. (2008). Volunteerism, health, and civic
engagement among older adults. Canadian Journal on Aging, 27, 399
406. doi:10.3138/cja.27.4.399
Grimm, R., Spring, K., & Dietz, N. (2007). The health benefits of volun-
teering: A review of recent research. New York, NY: Corporation for
National and Community Service.
Gruenewald, T. L., Karlamangla, A. S., Greendale, G. A., Singer, B. H.,
& Seeman, T. E. (2007). Feelings of usefulness to others, disability, and
mortality in older adults: The MacArthur Study of Successful Aging.
The Journals of Gerontology, Series B: Psychological Sciences and
Social Sciences, 62, P28 –P37. doi:10.1093/geronb/62.1.P28
Harris, A. H., & Thoresen, C. E. (2005). Volunteering is associated with
delayed mortality in older people: Analysis of the Longitudinal Study of
Aging. Journal of Health Psychology, 10, 739 –752. doi:10.1177/
1359105305057310
Herbst-Damm, K. L., & Kulik, J. A. (2005). Volunteer support, marital
status, and the survival times of terminally ill patients. Health Psychol-
ogy, 24, 225–229. doi:10.1037/0278-6133.24.2.225
Hsu, H. C. (2007). Does social participation by the elderly reduce mor-
tality and cognitive impairment? Aging & Mental Health, 11, 699 –707.
doi:10.1080/13607860701366335
Knickman, J. R., & Snell, E. K. (2002). The 2030 problem: Caring for
aging baby boomers. Health Services Research, 37, 849884. doi:
10.1034/j.1600-0560.2002.56.x
Konrath, S., Fuhrel-Forbis, A., Lou, A., & Brown, S. (2012). Motives for
volunteering are associated with mortality risk in older adults. Health
Psychology, 31, 87–96. doi:10.1037/a0025226
Lee, S. J., Steinman, M. A., & Tan, E. J. (2011). Volunteering, driving
status, and mortality in U.S. retirees. Journal of the American Geriatrics
Society, 59, 274 –280. doi:10.1111/j.1532-5415.2010.03265.x
Lum, T. Y., & Lightfoot, E. (2005). The effects of volunteering on the
physical and mental health of older people, Research on Aging, 27,
31–55. doi:10.1177/0164027504271349
Luoh, M. C., & Herzog, A. R. (2002). Individual consequences of
volunteer and paid work in old age: Health and mortality. Journal of
Health and Social Behavior, 43, 490 –509.
Mathers, C. D., Boerma, T., & Fat, D. M. (2009). Global and regional
causes of death. British Medical Bulletin, 92, 7–32. doi:10.1093/bmb/
ldp028
Musick, M. A., Herzog, A. R., & House, J. S. (1999). Volunteering and
mortality among older adults: Findings from a national sample. The
Journals of Gerontology, Series B: Psychological Sciences and Social
Sciences, 54, S173–S180. doi:10.1093/geronb/54B.3.S173
Numan, M. (2006). Hypothalamic neural circuits regulating maternal re-
sponsiveness toward infants. Behavioral Cognitive Neuroscience Re-
view, 5, 163–190. doi:10.1177/1534582306288790
Okun, M. A., August, K. J., Rook, K. S., & Newsom, J. T. (2010). Does
volunteering moderate the relation between functional limitations and
mortality? Social Science & Medicine, 71, 1662–1668. doi:10.1016/j
.socscimed.2010.07.034
Oman, D. (2007). Does volunteering foster physical health and longevity?
In S. G. Post (Ed.), Altruism and health: Perspectives from empirical
research (pp. 15–32). New York, NY: Oxford University Press.
Oman, D., Thoresen, C., & McMahon, K. (1999). Volunteerism and
mortality among the community-dwelling elderly. Journal of Health
Psychology, 4, 301–316. doi:10.1177/135910539900400301
Pargament, K. I. (1997). The psychology of religion and coping: Theory,
research and practice. New York, NY: Guilford Press.
Piliavin, J. A., & Siegl, E. (2007). Health benefits of volunteering in the
Wisconsin Longitudinal Study. Journal of Health and Social Behavior,
48, 450 464. doi:10.1177/002214650704800408
Pillemer, K., Suitor, J. J., Landreneau, L. T., Henderson, C. R., & Brang-
man, S. (2000). Peer support for Alzheimer’s caregivers: Lessons from
an intervention study. In K. Pillemer, J. J. Suitor, L. T. Landreneau, C. R.
Henderson, & S. Brangman (Eds.), Social integration in the second half
of life (pp. 265–286). Baltimore, MD: John Hopkins University Press.
Post, S. G. (Ed.). (2007). Altruism and health: Perspectives from empirical
research. New York, NY: Oxford University Press.
Post, S. G., & Neimark, J. (2007). Why good things happen to good people.
New York, NY: Broadway Books.
Rogers, R. (1996). The effects of family composition, health, and social
support linkages on mortality. Journal of Health and Social Behavior,
37, 328 –329.
Sabin, E. P. (1993). Social relationships and mortality among the elderly. Journal
of Applied Gerontology, 12, 44 60. doi:10.1177/073346489301200105
Shmotkin, D., Blumstein, T., & Modan, B. (2003). Beyond keeping
active: Concomitants of being a volunteer in old-old age. Psychology
and Aging, 18, 602– 607.
Thoits, P. A., & Hewitt, L. N. (2001). Volunteer work and well-being.
Journal of Health and Social Behavior, 42, 115–131.
Van Willigen, M. (2000). Differential benefits of volunteering across the
life course. The Journals of Gerontology, Series B: Psychological Sci-
ences and Social Sciences, 55, S308 –S318. doi:10.1093/geronb/55.5
.S308
von Bonsdorff, M. B., & Rantanen, T. (2011). Benefits of formal voluntary
work among older people: A review. Aging Clinical and Experimental
Research, 23, 162–169. doi:10.3275/7200
Wheeler, J. A., Gorey, K. M., & Greenblatt, B. (1998). The beneficial
effects of volunteering of older volunteers and the people they serve: A
meta-analysis. International Journal of Aging and Human Development,
47, 69 –79. doi:10.2190/VUMP-XCMF-FQYU-V0JH
Wilson, J., & Musick, M. (1997). Who cares? Toward an integrated theory
of volunteer work. American Sociological Review, 62, 694–713. doi:
10.2307/2657355
Zhang, J., & Yu, K. F. (1998). What’s the relative risk? A method of
correcting the odds ratio in cohort studies of common outcomes. Journal
of the American Medical Association, 280, 1690 –1691. doi:10.1001/
jama.280.19.1690
Received September 19, 2012
Revision received November 19, 2012
Accepted November 21, 2012
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
577
VOLUNTEERING AND MORTALITY
... Observational studies documented numerous associated benefits of volunteering for older adults, including better mental and physical health, better physical and cognitive functioning, lower pain, higher life satisfaction, and reduced/delayed mortality [25][26][27]. Observational studies also documented changes in several dimensions of social relationships, including enhanced social support [28,29], increased social engagement [26,30,31], opportunities to meet new people [32,33], and increased number of social ties [33,34]. However, relatively few studies used randomized controlled trials (RCTs) to examine benefits of volunteering to rule out confounding and reverse causality. ...
Article
Full-text available
Objectives Evidence-based strategies to reduce loneliness in later life are needed because loneliness impacts all domains of health, functioning, and quality of life. Volunteering is a promising strategy, as a large literature of observational studies documents associations between volunteering and better health and well-being. However, relatively few studies have used randomized controlled trials (RCTs) to examine benefits of volunteering, and none have examined loneliness. The primary objective of the Helping Older People Engage (HOPE) study is to examine the social-emotional benefits of a social volunteering program for lonely older adults. This manuscript describes the rationale and design of the trial. Methods We are randomly assigning adults aged 60 or older (up to 300) who report loneliness to 12 months of either AmeriCorps Seniors volunteering program or an active control (self-guided life review). Co-primary outcomes are assessed via self-report—loneliness (UCLA Loneliness Scale) and quality of life (WHOQOL-Bref). Enrollment was completed in May 2022 and follow-up assessments will continue through May 2023, with completion of primary outcomes soon thereafter. Conclusions Since older adults who report loneliness are less likely to actively seek out volunteering opportunities, if results support efficacy of volunteering for reducing loneliness, dissemination and scaling up efforts may involve connecting primary care patients who are lonely with AmeriCorps Seniors through aging services agencies. This RCT is registered at clinicaltrials.gov (NCT03343483).
... The value placed on being productive, such as that seen by Denise and Cathy, is consistent with existing literature which highlights the value of continued engagement in later life in an occupation of sorts whether volunteering or lifelong learning (Etienne and Jackson, 2011; Jenkinson et al., 2013;Galenkamp et al., 2016). The bene ts of participating in such activities are well documented, even when outcomes are controlled for initial health status (Okun et al., 2013). This was acknowledged by Angela when she described her volunteering at the hospital as 'sel sh,' as she took up the work for the psychosocial bene ts to herself. ...
Preprint
Full-text available
Background and Objectives More people are living alone across the life course: in later life this can have implications for practical and psychosocial support. The Covid pandemic emphasised the importance of this when the UK government restricted movement outside of households to limit the spread of disease. This had important ramifications regarding social contact and practical support. The objectives of this study were to explore the experience of older women living alone during this time, with a focus on health and wellbeing. Research Design and Methods This study used an Interpretative Phenomenological approach. Semi-structured interviews were undertaken with seven women (aged 65+), living alone in the UK. Interviews were carried out between May and October 2020. Interpretative Phenomenological Analysis was used to analyse the transcripts. Results Findings show that life course events shaped how living alone was experienced in later life. Convergences and divergences in lived experience were identified. Three superordinate themes emerged from the Interpretative Phenomenological Analysis: Productivity, Ownership, and Interconnectedness. Discussion and Implications Findings highlight the importance of life course events in shaping the experience of later life. They also provide a better understanding of the lived experience of living alone as an older woman, increasing knowledge of this group and how living alone can affect health and wellbeing. Implications for research and practice are discussed, such as the importance of recognising the specific support needs for this group in later life, and the need for further knowledge about groups whose needs are not met by standard practice.
... In a study examining whether awe would influence altruistic behavior after looking up at a grove of towering eucalyptus trees (awe condition) or looking at a tall building (neutral condition), after the awe induction, participants were more helpful toward the experimenter (i.e., picked up more pens) and reported feeling less entitled compared with the neutral condition (Piff et al., 2015). Relevant empirical evidence has consistently found that such prosocial relationality elevates well-being (e.g., Dunn et al., 2020) and might even boost life expectancy (e.g., Okun et al., 2013). ...
Article
Full-text available
How do experiences in nature or in spiritual contemplation or in being moved by music or with psychedelics promote mental and physical health? Our proposal in this article is awe. To make this argument, we first review recent advances in the scientific study of awe, an emotion often considered ineffable and beyond measurement. Awe engages five processes—shifts in neurophysiology, a diminished focus on the self, increased prosocial relationality, greater social integration, and a heightened sense of meaning—that benefit well-being. We then apply this model to illuminate how experiences of awe that arise in nature, spirituality, music, collective movement, and psychedelics strengthen the mind and body.
... Prosocial behaviors, such as displays of empathy, providing assistance to others, and engaging in community service, not only enhance the individual's social status, but also promote distinct health advantages 1,3 . Individuals that voluntarily engage in higher rates of prosociality benefit from greater physical health, lower disease rates, higher quality of life, and longer average lifespans [4][5][6][7][8] . Psychologically, prosociality induces feelings of happiness, which in turn increase the motivation to engage in further acts of prosocialty [9][10][11][12] , thereby creating a positive behavioral health loop for the actor 13 . ...
Article
Full-text available
The existing literature largely focuses on health risks and other pharmacodynamics of using cannabis, with fewer investigations of other normative psychological effects from consumption among otherwise healthy people. We measured several basic constructs of social psychology corresponding to the concept of prosociality among 146 healthy young adults between 18 and 25 years (M = 18.9, SD = 1.4) with varying detectable levels of tetrahydrocannabinol (THC) in their urine, controlling for participant’s sex, age, ethnicity, and childhood socio-economic status. Compared to THC-free individuals, cannabis users scored higher than non-users on validated measures of Prosocial Behaviors (d = .34, p = .04), the Empathy Quotient (d = .36, p < .01), Moral Harmlessness (d = .76, p < .01) and Moral Fairness (d = .49, p < .01), but exhibited a lower sense of Ingroup Loyalty (d = .33, p = .04). Relative to THC-free, same-sex individuals, female cannabis users scored significantly higher on measurements of Aggression (ds = .65 and .57, ps < .05) and male users scored higher on the Agreeableness dimension of personality (d = .91, p < .01).. Linear associations were found between the recency of last cannabis usage and the Prosocial Behaviors, Empathy Quotient, Moral Harmlessness, Moral Fairness and Agreeableness personality scores (rs from − .24 to .38, ps < .05). The findings suggest cannabis usage is associated with an increased sense of prosociality and prioritization of humanitarian behaviors that declines with time following cannabis consumption. Further research should focus on heterogeneity in the effects of cannabis consumption across users.
... Apart from a potentially reduced mortality, our systematic review did not identify much evidence for protective effects of informal caregiving in terms of reduced chronic morbidity, although we did not exclude these outcomes. Beyond the focus of our present work based on stringent methodological quality assessments, however, further research exists suggesting that the provision of emotional and practical support to others may result in improved mental and/or physical health for the provider of such support [61][62][63]. Informal caregiving can lead to a high level of self-esteem and a positive change in the sense of mastery among female informal caregivers when caring for a non-resident care recipient [64,65]. In addition, it also seems to have a stress-buffering effect, which leads to lower mortality [17]. ...
Article
Full-text available
A systematic overview of mental and physical disorders of informal caregivers based on population-based studies with good methodological quality is lacking. Therefore, our aim was to systematically summarize mortality, incidence, and prevalence estimates of chronic diseases in informal caregivers compared to non-caregivers. Following PRISMA recommendations, we searched major healthcare databases (CINAHL, MEDLINE and Web of Science) systematically for relevant studies published in the last 10 years (without language restrictions) (PROSPERO registration number: CRD42020200314). We included only observational cross-sectional and cohort studies with low risk of bias (risk scores 0–2 out of max 8) that reported the prevalence, incidence, odds ratio (OR), hazard ratio (HR), mean- or sum-scores for health-related outcomes in informal caregivers and non-caregivers. For a thorough methodological quality assessment, we used a validated checklist. The synthesis of the results was conducted by grouping outcomes. We included 22 studies, which came predominately from the USA and Europe. Informal caregivers had a significantly lower mortality than non-caregivers. Regarding chronic morbidity outcomes, the results from a large longitudinal German health-insurance evaluation showed increased and statistically significant incidences of severe stress, adjustment disorders, depression, diseases of the spine and pain conditions among informal caregivers compared to non-caregivers. In cross-sectional evaluations, informal caregiving seemed to be associated with a higher occurrence of depression and of anxiety (ranging from 4 to 51% and 2 to 38%, respectively), pain, hypertension, diabetes and reduced quality of life. Results from our systematic review suggest that informal caregiving may be associated with several mental and physical disorders. However, these results need to be interpreted with caution, as the cross-sectional studies cannot determine temporal relationships. The lower mortality rates compared to non-caregivers may be due to a healthy-carer bias in longitudinal observational studies; however, these and other potential benefits of informal caregiving deserve further attention by researchers.
Article
The extant literature on volunteering has focused primarily on the many benefits of volunteering for older adults. However, the question rarely investigated is whether these benefits dissipate when older adults retire from their volunteering. Given the U.S. policy context wherein volunteering is promoted as a solution to the problems of aging, this research investigates the association between the loss of one’s volunteering role through retirement and well-being. Utilizing three waves of the U.S.-based National Social Life, Health, and Aging Project (NSHAP) (2005–2016) and a fixed-effects modeling approach, we find that the well-being of older adults, measured as self-reported health, happiness, and depressive risk, is negatively associated with volunteer retirement. Our study contributes to the literature on well-being and volunteering for older adults and is the first study focusing on this critical transition point in the life of older volunteers. In addition, policymakers and organizations must broaden their focus to include not only the recruitment and retention of older adult volunteers but also the transition out of volunteering that many of them will eventually face.
Article
Using 2017 Survey of Health, Aging and Retirement in Europe (SHARE) data, this study presents background characteristics and psycho-social correlates of volunteering among European older adults. Variables included presence/frequency of volunteer/charity work, socio-demographic variables, and a well-being index. ANOVA and regression analysis demonstrated significant differences in volunteer/charity activities across EU (European Union) nations; volunteering was more common in Nordic-Western European countries than in Middle Eastern and Southern Europe. Younger, female, married, and retired individuals were more likely to volunteer. Controlling socio-demographic variables, previous-year volunteers were more likely to score high on self-rated health, quality of life, life satisfaction, and score low on depression and loneliness. Volunteer engagement was significantly and positively associated across well-being indices. The study reiterates and updates previous SHARE-based and other studies. Policy provisions and institutional support should create volunteer opportunities that rebuild social roles and functions, and create challenges, for older adults.
Article
Building on theoretical and empirical literatures showing that choices not only reflect but also create preferences, we develop a two-stage compound task mechanism to promote pro-sociality. The first stage involves incentivizing participants to complete a compound task consisting of a targeted pro-social activity—volunteering– and a complement activity—writing about volunteering. The second stage involves incentivizing participants to repeatedly complete only the writing about volunteering. We conduct a field experiment and show that, conditional on completing first-stage volunteering + writing, intrinsic interest in volunteering is promoted even when people fail to complete the second-stage writing about volunteering. By contrast, participants assigned in the second stage either to volunteering, or to volunteering + writing, but who failed to complete these tasks, did not develop intrinsic interest in volunteering. These results are consistent with the theory underlying our two-stage compound-task mechanism.
Article
In the past 20 years, older adults’ civic participation has received considerable attention. Current literature shows that rates of voting and volunteering have been consistently lower among African Americans and Latinx older adults compared to White older adults. However, little research has explored civic participation in the context of historical structures of inequality that influence how Black and Latinx populations participate in civic life. I draw from an intersectional life course perspective and phenomenological methods to examine experiences of civic participation through participants’ lens. Findings draw our attention to how race/racism and age/ageism shape how, where, and with whom participants participate. Findings demonstrate how civic participation is embedded within systems of inequality that inform individual behavior as well as available opportunities for engagement. These findings call attention to the need to re-conceptualize and support civic participation that centers the experiences of historically ethnoracially oppressed populations.
Article
Full-text available
Unlabelled: There is a vast literature on the health benefits associated with volunteering for volunteers. Such health advantages are likely to vary across groups of volunteers with different characteristics. The current paper aims to examine the health advantages of volunteering for European volunteers and explore heterogeneity in the association between volunteering and health. We carry out a mega-analysis on microdata from six panel surveys, covering 952,026 observations from 267,212 respondents in 22 European countries. We provide open access to the code we developed for data harmonization. We use ordinary least squares, fixed effects, first difference, and fixed effect quantile regressions to estimate how volunteering activities and changes therein are related to self-rated health for different groups. Our results indicate a small but consistently positive association between changes in volunteering and changes in health within individuals. This association is stronger for older adults. For respondents 60 years and older, within-person changes in volunteering are significantly related to changes in self-rated health. Additionally, the health advantage of volunteering is larger for respondents in worse health. The advantage is largest at the lowest decile and gradually declines along the health distribution. The magnitude of the association at the first decile is about twice the magnitude of the association at the ninth decile. These results suggest that volunteering may be more beneficial for the health of specific groups in society. With small health advantages from year to year, volunteering may protect older and less healthy adults from health decline in the long run. Supplementary information: The online version contains supplementary material available at 10.1007/s10433-022-00691-5.
Chapter
Full-text available
In this essay we describe the essential features of a neurobiological system whose purpose is to provide the motivation needed to bestow resources on others-the &"caregiving system.&" After presenting a brief review of the evolutionary theoretical background, we describe how insights from selective investment theory and animal models of maternal care can be used to identify caregiving neural circuitry that may be involved in human helping behavior. At a minimum, we suggest that caregiving neural circuitry should be responsive to need in others, manage motivational conflict, and be selectively attuned to cues that there is a low risk of exploitation. We conclude with some implications of this model, including challenges it poses to views of human motivation that emphasize self-interest.
Chapter
This chapter provides an overview of software Comprehensive Meta‐Analysis (CMA) and shows how to use it to implement the ideas. The same approach could be used with any other program as well. The chapter also provides a sense for the look‐and‐feel of the program. CMA features a spreadsheet view and a menu‐driven interface. As such, it allows a researcher to enter data and perform a simple analysis in a matter of minutes. At the same time, it offers a wide array of advanced features, including the ability to compare the effect size in subgroups of studies, to run meta‐regression, to estimate the potential impact of publication bias, and to produce high‐resolution plots. The program is designed to work with studies that compare an outcome in two groups or that estimate an outcome in one group.
Chapter
This chapter offers a scientific perspective and reviews available evidence on how volunteering affects health and longevity. It focuses on formal volunteer work - performed through a school, hospital, library, or environmental, political, or other organization. Formal volunteer work stands in contrast to more casual or unorganized helping activities, often termed 'informal helping', such as giving directions to a stranger or serving as a caregiver for a family member or a neighbour. The chapter's primary focus is on physical health outcomes, although it also cites evidence linking volunteering with improved mental health and subjective well-being. First, it describes the mechanisms by which volunteering might affect physical health, as well as moderating factors that might strengthen or weaken these influences. Next, it reviews empirical evidence suggesting that volunteering may indeed provide physical and mental health benefits. It concludes by discussing some practical implications and needs for further research.
Article
Logistic regression is used frequently in cohort studies and clinical trials. When the incidence of an outcome of interest is common in the study population (>10%), the adjusted odds ratio derived from the logistic regression can no longer approximate the risk ratio. The more frequent the outcome, the more the odds ratio overestimates the risk ratio when it is more than 1 or underestimates it when it is less than 1. We propose a simple method to approximate a risk ratio from the adjusted odds ratio and derive an estimate of an association or treatment effect that better represents the true relative risk.
Article
IntroductionIndividual studiesThe summary effectHeterogeneity of effect sizesSummary points
Article
We construct an integrated theory of formal and informal volunteer work based on the premises that volunteer work is (1) productive work that requires human capital, (2) collective behavior that requires social capital, and (3) ethically guided work that requires cultural capital. Using education, income, and functional health to measure human capital, number of children in the household and informal social interaction to measure social capital, and religiosity to measure cultural capital, we estimate a model in which formal volunteering and informal helping are reciprocally related but connected in different ways to different forms of capital. Using two-wave data from the Americans' Changing Lives panel study, we find that formal volunteering is positively related to human capital, number of children in the household, informal social interaction, and religiosity. Informal helping, such as helping a neighbor, is primarily determined by gender, age, and health. Estimation of reciprocal effects reveals that formal volunteering has a positive effect on helping, but helping does not affect formal volunteering.
Article
A growing literature documents the positive effect of social relationships on health, in general, and in reducing mortality, in particular. Much remains to be learned about which relationships have this effect, particularly among the elderly. This research, a secondary analysis of a national sample of community-dwelling elderly aged 70 and over, addresses these questions. Data come from the Longitudinal Study on Aging. Five scales constructed from relationship questions asked in the survey are related to respondent mortality over a 4-year follow-up using multiple logistic regression . The research confirms that certain relationships appear to reduce subsequent mortality: going to church/temple, volunteering, seeing friends or neighbors, and talking with them on the phone (socioexpressive relationships). The effect can be seen among healthy and more infirm elders. Helping relationships and household and kin relationships do not seem to reduce the risk of elder moriality.
Book
An Introduction to the Psychology of Religion and Coping. Part I: A Perspective on Religion. The Sacred and the Search for Significance. Religious Pathways and Religious Destinations. Part II: A Perspective on Coping. An Introduction to the Concept of Coping. The Flow of Coping. Part III: The Religion and Coping Connection. When People Turn to Religion. When They Turn Away. The Many Faces of Religion in Coping. Religion and the Mechanisms of Coping - The Transformation of Significance. Part IV: Evaluative and Practical Implications. Does it Work? Religion and the Outcomes of Coping. When Religion Fails - Problems of Integration in the Process of Coping. Putting Religion into Practice.