https://orcid.org/0000-0002-6359-6409
Abstract
Friendships are essential in the face of social network changes in later life and friendships may be important for reducing depression risk. Social participation through volunteering is also associated with fewer depressive symptoms. What is less well-understood is whether friendships serve as a pathway in the link between volunteering and depression.
We used panel data from the Health and Retirement Study (2010, 2014, 2018). Negative binomial regression within the SEM modeling framework was employed to analyze the association between volunteering and friendship, focusing on the indirect effect of friendships for understanding the volunteering and depressive symptoms relationship.
Volunteer hours were positively associated with friendship (1–99 hr: β = 0.17, p < .001, 100–199 hr: β = 0.15, p < .001, 200 hr and more: β = 0.23, p < .001) and negatively associated with number of depressive symptoms (1–99 hr: β = −0.07, p = .06, 100–199 hr: β = −0.14, p < .001, 200 hr and more: β = −0.17, p < .001). Friendship mediated the relationship between volunteer hours and depressive symptoms (indirect effects; 1–99 hr: β = −0.01, 95% confidence interval [CI] = [−0.02, −0.00], p = .03), 100–199 hr: β = −0.01, 95% CI = [−0.02, −0.00], p = .03), 200 hr and more: β = −0.02, 95% CI = [−0.03, −0.00], p = .03).
Our findings underscored the role of volunteering in generating and maintaining friendships, as well as for friendships as a pathway between volunteer hours and depressive symptoms. Providing opportunities to maintain and grow friendships in later life may be a possible intervention strategy for older adults at risk of depression.
Friendship is conceptualized as a voluntary and equitable relationship between individuals that consists of mutual trust and respect, emotional closeness, shared interests, and satisfying collective experiences (Blieszner, 2014). Friendship represents an important component of the convoy model of social relations in later life but this type of relationship is not as well studied as other relationships, such as those with spouses, adult children, and even grandchildren (Antonucci et al., 2014; Blieszner et al., 2019). As such, Blieszner and colleagues (2019) call on scholars to generate a better understanding of the unique role of friendship in the well-being of older adults.
Further, volunteering is a social activity that enables older adults to make and build friendship in later life (Anderson et al., 2014; Bruggencate et al., 2018). Most of the extant literature explores how individuals’ social networks are associated with volunteering activities (Haski-Leventhal et al., 2018). However, few studies examine how volunteering functions as an avenue to expand their social network through friendship, potentially helping to cope with social losses (i.e., death of a family member or friend) that are common in later life (Guiney et al., 2021; Jang et al., 2018).
Research on social participation, including volunteering for formal organizations, is expanding. Morrow-Howell (2010) issued a challenge to researchers in her seminal review of the field to learn more about the mechanisms that explain the volunteering and health connection. Friendship may be one such mechanism that links volunteering behaviors to mental health. We argue that volunteering presents older adults with opportunities to make friends, who may provide actual social support, when needed, and these friendships contribute to increased perceived social support (see Anderson et al., 2014). Friendship is generally beneficial to older adults’ health. In later life, friendship is associated with higher life satisfaction and positive affect (Bruine de Bruin et al., 2020; Tomini et al., 2016), reduced feelings of loneliness (Dahlberg et al., 2022), better cognitive functioning (Peng et al., 2022) and better oral health (Tsakos et al., 2013), and lower risk of loneliness and mortality (Aida et al., 2018; Dahlberg et al., 2022). In addition, friendship alleviates depressive symptoms and reduces the risk of depression (Bookwala et al., 2014; Han et al., 2019). Hence, friendship likely mediates the association between volunteering and depression.
Although few studies evaluate (a) whether volunteering helps build friendships and (b) whether friendships mediate the association between volunteering and depression, the relevant scientific literature on the link between volunteering and social relationships and the association between friendship and depression provides a foundation on which to build the current study. Studies on this topic contain methodological limitations, such as the use of cross-sectional research designs, small sample sizes, and limited measures of friendship characteristics. These limitations lower our confidence in the robustness of the reported relationships, indicating the need for more research (Creaven et al., 2018; Pilkington et al., 2012). Overcoming these limitations will deepen our understanding of the role of friendship in the association between volunteering and depression.
This paper contributes to the scientific literature by examining the associations among volunteering, friendship, and depressive symptoms among older adults. Specifically, using three waves of nationally representative data from the Health and Retirement Study, this study addresses the following research questions: (a) Is volunteering associated with depressive symptoms? (b) Is volunteering associated with friendship? 3) Is friendship related to depressive symptoms? and 4) Does friendship mediate the association between volunteering and depressive symptoms?
Volunteering and Friendship
Formal volunteering is the unpaid act of helping other persons, organizations, and communities by giving time and energy freely without obligation or substantial remuneration (Jongenelis et al., 2020). Volunteering is a way for older adults to stay active, providing the opportunity to make new friends while staying socially engaged (Bruggencate et al., 2018; Burr et al., 2021). This outcome is supported in existing studies, where participation in volunteering programs is associated with increased social ties, fostering high quality social interactions (Fried et al., 2004; Guiney et al., 2021; Jang et al., 2018; Rook & Sorkin, 2003). A study of the Foster Grandparent Program found that their program’s participants formed new friendships, which often turned into close social ties (Rook & Sorkin, 2003). Jang and colleagues (2018) reported that being a new volunteer was linked to higher social contact frequency with friends after participating in the Experience Corps (N = 354), an innovative program where older community members volunteer to provide tutoring in public elementary schools. Further, older participants of the same program reported that they had more people they could ask for help after attending the program (Fried et al., 2004). Another study of 91 older New Zealanders between the ages of 65–75 also found that volunteering was linked to more social activities with friends (Guiney et al., 2021). Overall, these studies suggest that volunteering fosters friendships and social interactions.
However, these studies employ cross-sectional designs and rely on small, non-representative samples. Thus, these results are not generalizable beyond the samples employed and it is unclear if the positive effects of volunteering are associated with friendship over time. Further, these studies only investigated volunteer status, which does not allow us to know if amount of volunteering, as represented in volunteer hours, is informative. Our study contributes to the scientific literature by investigating whether the number of hours volunteered is associated with friendship characteristics with a nationally representative sample of older adults observed at three time points.
Volunteering and Depression
Research shows that volunteering is beneficial for older adults’ mental health, including depression, where volunteering is associated with fewer depressive symptoms (Burr et al., 2021). A meta-analysis conducted by Jenkinson and colleagues (2013) concludes that volunteering has beneficial effects on reducing depression risk. Using two waves of data from the Health and Retirement Study, Huo et al. (2021) report that volunteering for 100 hr or more is linked to fewer depressive symptoms among older adults. Volunteering is advantageous to depression because volunteering activities create new roles for older adults that can increase their sense of meaning and purpose in life, which then leads to better well-being (Burr et al., 2021; Li & Ferraro, 2005). However, it is uncertain from the existing literature what are the plausible pathways that link volunteering to depression risk among older adults. Therefore, this study contributes to the literature by examining whether friendship is a potential mechanism.
Friendship and Depression
Studies show that friendship is linked to lower depression risk. For instance, older adults with friends in which to confide have fewer depressive symptoms (Bookwala, 2017). Engaging in more social interactions with friends is also associated with fewer depressive symptoms among older adults and couples from the Health and Retirement Study (Han et al., 2019; Teo et al., 2015). Further, the support provided by friends is linked to fewer depressive symptoms among older adults transitioning into retirement (Kail & Carr, 2020). Thus, friendship is a plausible mechanism between volunteering and depression because volunteering increases the frequency of social interactions with friends, who provide companionship and emotional support, as well as instrumental support.
Friendship as a Mediator for the Association Between Volunteering and Depression
Despite a limited number of research studies, we posit that friendship has a plausible mediating effect on the association between volunteering and depression, whereby volunteering increases social inclusion and social support from friends, which in turn is linked to a reduction in depression risk. Studying older adults from the Transitions in Later Life Study (N = 562), Pilkington et al. (2012) demonstrate that older volunteers reported higher levels of positive affect, which was mediated by social support received from their friends. Employing data from the European Social Survey,Creaven et al. (2018) report that volunteering was not related to depression after controlling for social connectedness (measured with a latent variable identified by number of social ties, contact frequency, and support received from different social ties including relatives, friends, and work colleagues). These findings provide some support for the expectation that friendship serves as a mediator in the association between volunteering and emotional well-being.
While these two studies have their merits, they also have shortcomings that reduce our confidence about the potential role that friendship may play in the link between volunteering and depression. Both studies are based on a cross-sectional design, which does not allow us to address simultaneity bias because volunteering activity and depressive symptoms are measured at the same observation point. Further, Creaven et al. (2018) did not differentiate friendship relationships from other social relationships, and Pilkington et al. (2012) used a small sample of older adults with a relatively low level of statistical power for uncovering empirical relationships. Thus, more research is needed to increase our confidence in findings of these studies. No study has examined the mediating effect of friendship on the association between volunteering and depressive symptoms using panel data from a large, nationally representative sample of older adults.
Conceptual Framework
Our study is informed by The Focus Theory of Social Interaction. This theory states that social positions and activities, such as volunteering, function as focus events that bring individuals together, promote social interaction, and establish connections when individuals participate in these focus events (Feld, 1981; Häuberer & Häuberer, 2014; Paik & Navarre-Jackson, 2011). Thus, older adults are likely to foster new friendships and maintain existing friendships through volunteering. We also argue that friendship is a likely pathway for understanding the association between volunteering and depression. As discussed above, volunteering creates environments for individuals to develop social relations with others. Besides providing meaning and purpose in life, volunteering activities often enhance actual and perceived social support and provide more opportunities for social participation (Jongenelis et al., 2021). Social support and social participation are fostered through the development of new friendships and by strengthening existing friendships, promoting emotional well-being and reducing the risk of depression (Hooker et al., 2018). Therefore, friendship is likely to mediate the association between volunteering and depression.
The Current Study
This study examines the associations among friendship, volunteering, and depressive symptoms by employing the 2010, 2014, and 2018 waves of the Health and Retirement Study. First, we examine if there is an association between volunteering at time 1 (2010) and friendship at time 2 (2014), and whether there is an association between volunteering at time 1 (2010) and number of depressive symptoms at time 3 (2018). Next, we investigate whether friendship measured at time 2 (2014) mediates the association between volunteering at time 1 (2010) and depressive symptoms at time 3 (2018). We offer four hypotheses: (a) volunteering more hours is related to fewer depressive symptoms; (b) older adults who spend more time volunteering are more likely to report more supportive friendship characteristics; (c) supportive friendship characteristics are related to fewer depressive symptoms; and (d) supportive friendship characteristics mediate the association between volunteer hours and depressive symptoms, whereby older adults who volunteer are more likely to interact with and be supported by their friends, and older adults with more supportive friendship characteristics report fewer depressive symptoms.
Design and Methods
Data Source
This study used data from the Health and Retirement Study (HRS), a nationally representative panel survey of adults aged 50 and over in the United States. The survey also included a battery of psychosocial indicators reported in the Leave-Behind Questionnaire (LBQ). The LBQ is given to a random sample of approximately one-half of the core participants every four years. Further, we took some variables from the RAND HRS longitudinal data file (e.g., household wealth), where variables are cleaned and processed to account for missing values (Bugliari et al., 2018).
To test our hypotheses, the mediation indicator (friendship) is observed (2014) temporally between the measurement of volunteer hours (2010) and the measurement of depressive symptoms (2018). The 2014 wave of the HRS is the “anchor” point for our analyses (N = 18,747). To be included in the study, participants had to answer the LBQ questionnaire in 2014 (N = 7,435). The LBQ included 744 participants without valid information on the friendship questions (N = 6,691). Study sample participants were aged 50 years and above and provided valid information on volunteering (N = 6,187) and depressive symptoms (N = 4,533). Finally, participants were excluded if they did not live in the community or had proxy respondents in any of the three waves. We applied a full information maximum likelihood method to handle missing data for the covariates (2014), maximizing the number of participants for the study. The final study sample included 4,532 participants.
Measures
Depressive symptoms
Depressive symptoms in 2018 were self-reported using the eight-item Center for Epidemiological Studies Depression scale (Radloff, 1977; Turvey et al., 1999). Participants reported whether in the past week they lost interest, enjoyed life, felt happy, felt tired, lost appetite, felt down on themselves, had trouble falling asleep, and had thoughts about death (1 = yes, 0 = no). The two positive items were reverse coded and a summary score of these eight items was created, with higher scores indicating a higher number of depressive symptoms (range = 0–8; α = 0.78).
Volunteering
Participants indicated whether they volunteered in 2010 for any religious, educational, health-related, or other charitable organization in the past 12 months. If they volunteered, participants indicated the amount of time spent volunteering, 1–50 hr, 51–99 hr, 100–199 hr, or 200 or more hours. Based on these options, we created a four-category variable for the number of hours volunteered (0 = no hours volunteered [reference group], 1 = 1–99 hr, 2 = 100–199 hr, and 3 = 200 or more hours).
Friendship characteristics
We generated a latent variable containing three dimensions of friendship from the 2014 wave, the first year the full battery of friendship items was available. These items included number of close friends, positive friendship quality, and social contact frequency with friends. Negative friendship quality was not included in the latent variable because negative friendship quality did not contribute meaningfully to the latent variable and it is considered an independent dimension of friend-based social relationships (Zahodne et al., 2019).
Number of close friends.—
Participants self-reported the number of friends with which they had a close relationship. We top-coded the distribution at the 95th percentile to address skewness, where the maximum number of close friends was 12 (range = 0–12).
Social contact frequency with friends.—
Participants rated how often they either meet up with friends, speak on the phone, write or email their friends, or communicate via social media with their friends, such as through Skype, Facebook, or other social media platforms on a 6-point Likert scale, ranging from 1 = three or more times a week to 6 = less than once a year or never. These items were reversed coded and a mean score across the four items was created (range = 1–6; α = 0.60), with higher scores representing more social contact frequency with friends.
Positive friendship quality.—
Positive friendship quality was measured by the degree to which participants reported that “friends understand the way you feel about things” “you rely on friends if you have serious problems?” and, “you open up to friends if you need to talk about your worries.” The responses to these items were reversed coded so that 1 indicated “not at all” and 4 indicated “a lot.” The final score was the average rating across the three items (range =1–4; α = 0.83). A higher score indicated higher positive friendship quality.
Covariates
Participants’ sociodemographic, wealth, disability, and social relationship characteristics were taken from the 2014 wave. Participants’ age (52–97) and education (0–17) were measured in years. Gender (1 = female, 0 = male) and worked for pay (1 = yes, 0 = no), along with race-ethnic group status (1 = non-Hispanic White [reference group], 2 = non-Hispanic Black, 3 = non-Hispanic other race group, 4 = Hispanic, any race) and marital status (1 = married [reference group], 2 = divorced/separated, 3 = widowed, 4 = never married) were coded as a set of dichotomous variables. Household wealth (assets less debts) was transformed by using an inverse hyperbolic sine transformation to account for skewness and negative values (Friedline et al., 2015). We also included the number of activities of daily living limitations (dressing, bathing, toileting, eating, walking, and getting in and out of bed; range = 0–6), and memory functioning (a summary score of participants’ immediate and delayed word recall scores; range = 0–20).
We included participants’ positive relationship quality with their spouses and children. These questions were based on wording similar to the question for positive friendship quality. Further, participants’ negative relationship quality with their spouses and children were assessed with four items: “your spouse/children make too many demands on you,” “they criticize you,” “they let you down when you are counting on them” and, “they get on your nerves?” All four items were reversed coded so that 1 = not at all and 4 = a lot. The final score was the average rating across the four items (range = 1–4). Higher scores indicated higher negative relationship quality. Cronbach’s α statistics for positive relationship quality with spouses and children were 0.81 and 0.82, respectively, and for negative relationship quality with spouses and children, Cronbach α statistics were 0.78 and 0.79, respectively.
Analytic Strategy
First, we conducted descriptive analyses for the variables listed above. Next, we established a latent friendship variable using confirmatory factor analysis (i.e., measurement model) including three indicators: number of close friends, friend social contact frequency, and positive friendship quality. The latent construct allowed the factor loadings for each item to be freely estimated, which helps account for random measurement error that might exist in the observed indicators. Last, because number of depressive symptoms was a count variable, we employed negative binomial regression modeling within the structural equation modeling framework to analyze the mediating effect of friendship on the association between volunteer hours and number of depressive symptoms (i.e., structural model; see Rijnhart et al., 2021). Two analytic steps were conducted: we first estimated the association between volunteering and friendship and volunteering and depressive symptoms; next, we estimated the indirect effect of friendship.
Model fit indices were identified to determine whether the latent variable was successfully established, including the Tucker–Lewis index (TLI) and comparative fit index (CFI) greater than 0.95, the root mean square error of approximation (RMSEA) lower than 0.07, and the standardized root mean square residual (SRMR) lower than 0.05 (Muthén & Muthén, 2017). The structural model fit was assessed with the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). We reported standardized regression coefficients of the model estimates. All analyses were conducted with Mplus (v. 8).
Results
Sample Characteristics
The descriptive characteristics of the study sample are presented in Table 1. The average number of depressive symptoms reported in 2018 was 1.29 (SD = 1.87). About 46.27% of the participants engaged in volunteering activities in the past year. Among the participants, 26.81%, 10.57%, and 8.89% volunteered 1–99 hr, 100–199 hr, or more than 200 hr, respectively. Participants’ mean scores for number of close friends, friend social contact frequency, and positive friendship quality were 4.02 (SD = 3.08), 3.35 (SD = 1.04), and 3.06 (SD = 0.73), respectively.
Descriptive Characteristics of the HRS Study Sample
| Variables | M (SD) |
|---|---|
| Depressive symptoms (2018)a (range: 0–8) | 1.29 (1.87) |
| Depressive symptoms based on volunteer hours | |
| 0 hr | 0.22 (0.42) |
| 1–99 hr | 0.15 (0.36) |
| 100–199 hr | 0.10 (0.30) |
| 200 hr or above | 0.09 (0.29) |
| Participants who volunteered (2010) (%) | 46.27 |
| Volunteer hours (2010) (%) | |
| 0 hr (reference group) | 53.73 |
| 1–99 hr | 26.81 |
| 100–199 hr | 10.57 |
| 200 hr or above | 8.89 |
| Friendship (2014) | |
| Number of close friends (range: 0–12) | 4.02 (3.08) |
| Contact frequency with friendsb (range: 0–6) | 3.35 (1.04) |
| Friendship quality with friendsc (range: 1–4) | 3.06 (0.73) |
| Covariates (2014) | |
| Age (range: 52–97 years) | 67.97 (9.18) |
| Female (%) | 60.68 |
| Education (range: 0–17 years) | 13.16 (2.95) |
| Married (reference group; %) | 60.66 |
| Divorced or separated (%) | 16.06 |
| Widowed (%) | 18.29 |
| Never married (%) | 4.99 |
| Non-Hispanic White (reference group; %) | 69.25 |
| Non-Hispanic Black (%) | 16.82 |
| Non-Hispanic Other Race (%) | 2.80 |
| Hispanic (%) | 11.10 |
| Household Wealth (range: −497,840–23,805,000) | 195,900.00 (Median) (984,498.77) |
| Household Wealth (logged) (range: −13.81–17.68) | 11.12 (5.80) |
| ADL limitationsd (range: 0–6) | 0.28 (0.84) |
| Memory functioning levele (range: 0–20) | 10.15 (3.16) |
| Worked for pay (%) | 37.42 |
| Positive relationship with spousef (range: 0–4) | 3.51 (0.61) |
| Negative relationship with spousef (range: 0–4) | 1.91 (0.65) |
| Positive relationship with childrenf (range: 0–4) | 3.26 (0.73) |
| Negative relationship with childrenf (range: 0–4) | 1.69 (0.64) |
| Variables | M (SD) |
|---|---|
| Depressive symptoms (2018)a (range: 0–8) | 1.29 (1.87) |
| Depressive symptoms based on volunteer hours | |
| 0 hr | 0.22 (0.42) |
| 1–99 hr | 0.15 (0.36) |
| 100–199 hr | 0.10 (0.30) |
| 200 hr or above | 0.09 (0.29) |
| Participants who volunteered (2010) (%) | 46.27 |
| Volunteer hours (2010) (%) | |
| 0 hr (reference group) | 53.73 |
| 1–99 hr | 26.81 |
| 100–199 hr | 10.57 |
| 200 hr or above | 8.89 |
| Friendship (2014) | |
| Number of close friends (range: 0–12) | 4.02 (3.08) |
| Contact frequency with friendsb (range: 0–6) | 3.35 (1.04) |
| Friendship quality with friendsc (range: 1–4) | 3.06 (0.73) |
| Covariates (2014) | |
| Age (range: 52–97 years) | 67.97 (9.18) |
| Female (%) | 60.68 |
| Education (range: 0–17 years) | 13.16 (2.95) |
| Married (reference group; %) | 60.66 |
| Divorced or separated (%) | 16.06 |
| Widowed (%) | 18.29 |
| Never married (%) | 4.99 |
| Non-Hispanic White (reference group; %) | 69.25 |
| Non-Hispanic Black (%) | 16.82 |
| Non-Hispanic Other Race (%) | 2.80 |
| Hispanic (%) | 11.10 |
| Household Wealth (range: −497,840–23,805,000) | 195,900.00 (Median) (984,498.77) |
| Household Wealth (logged) (range: −13.81–17.68) | 11.12 (5.80) |
| ADL limitationsd (range: 0–6) | 0.28 (0.84) |
| Memory functioning levele (range: 0–20) | 10.15 (3.16) |
| Worked for pay (%) | 37.42 |
| Positive relationship with spousef (range: 0–4) | 3.51 (0.61) |
| Negative relationship with spousef (range: 0–4) | 1.91 (0.65) |
| Positive relationship with childrenf (range: 0–4) | 3.26 (0.73) |
| Negative relationship with childrenf (range: 0–4) | 1.69 (0.64) |
Notes: N = 4,532. HRS = Health and Retirement Study.
aEight-item Center for Epidemiological Studies Depression scale.
bRated from 1 (less than once a year or never) to 6 (three or more times a week).
cRated from 1 (not at all) to 4 (a lot).
dADL = activities of daily living.
eMeasured as the sum of immediate and delayed word recall.
fRated from 1 (not at all) to 4 (a lot).
Descriptive Characteristics of the HRS Study Sample
| Variables | M (SD) |
|---|---|
| Depressive symptoms (2018)a (range: 0–8) | 1.29 (1.87) |
| Depressive symptoms based on volunteer hours | |
| 0 hr | 0.22 (0.42) |
| 1–99 hr | 0.15 (0.36) |
| 100–199 hr | 0.10 (0.30) |
| 200 hr or above | 0.09 (0.29) |
| Participants who volunteered (2010) (%) | 46.27 |
| Volunteer hours (2010) (%) | |
| 0 hr (reference group) | 53.73 |
| 1–99 hr | 26.81 |
| 100–199 hr | 10.57 |
| 200 hr or above | 8.89 |
| Friendship (2014) | |
| Number of close friends (range: 0–12) | 4.02 (3.08) |
| Contact frequency with friendsb (range: 0–6) | 3.35 (1.04) |
| Friendship quality with friendsc (range: 1–4) | 3.06 (0.73) |
| Covariates (2014) | |
| Age (range: 52–97 years) | 67.97 (9.18) |
| Female (%) | 60.68 |
| Education (range: 0–17 years) | 13.16 (2.95) |
| Married (reference group; %) | 60.66 |
| Divorced or separated (%) | 16.06 |
| Widowed (%) | 18.29 |
| Never married (%) | 4.99 |
| Non-Hispanic White (reference group; %) | 69.25 |
| Non-Hispanic Black (%) | 16.82 |
| Non-Hispanic Other Race (%) | 2.80 |
| Hispanic (%) | 11.10 |
| Household Wealth (range: −497,840–23,805,000) | 195,900.00 (Median) (984,498.77) |
| Household Wealth (logged) (range: −13.81–17.68) | 11.12 (5.80) |
| ADL limitationsd (range: 0–6) | 0.28 (0.84) |
| Memory functioning levele (range: 0–20) | 10.15 (3.16) |
| Worked for pay (%) | 37.42 |
| Positive relationship with spousef (range: 0–4) | 3.51 (0.61) |
| Negative relationship with spousef (range: 0–4) | 1.91 (0.65) |
| Positive relationship with childrenf (range: 0–4) | 3.26 (0.73) |
| Negative relationship with childrenf (range: 0–4) | 1.69 (0.64) |
| Variables | M (SD) |
|---|---|
| Depressive symptoms (2018)a (range: 0–8) | 1.29 (1.87) |
| Depressive symptoms based on volunteer hours | |
| 0 hr | 0.22 (0.42) |
| 1–99 hr | 0.15 (0.36) |
| 100–199 hr | 0.10 (0.30) |
| 200 hr or above | 0.09 (0.29) |
| Participants who volunteered (2010) (%) | 46.27 |
| Volunteer hours (2010) (%) | |
| 0 hr (reference group) | 53.73 |
| 1–99 hr | 26.81 |
| 100–199 hr | 10.57 |
| 200 hr or above | 8.89 |
| Friendship (2014) | |
| Number of close friends (range: 0–12) | 4.02 (3.08) |
| Contact frequency with friendsb (range: 0–6) | 3.35 (1.04) |
| Friendship quality with friendsc (range: 1–4) | 3.06 (0.73) |
| Covariates (2014) | |
| Age (range: 52–97 years) | 67.97 (9.18) |
| Female (%) | 60.68 |
| Education (range: 0–17 years) | 13.16 (2.95) |
| Married (reference group; %) | 60.66 |
| Divorced or separated (%) | 16.06 |
| Widowed (%) | 18.29 |
| Never married (%) | 4.99 |
| Non-Hispanic White (reference group; %) | 69.25 |
| Non-Hispanic Black (%) | 16.82 |
| Non-Hispanic Other Race (%) | 2.80 |
| Hispanic (%) | 11.10 |
| Household Wealth (range: −497,840–23,805,000) | 195,900.00 (Median) (984,498.77) |
| Household Wealth (logged) (range: −13.81–17.68) | 11.12 (5.80) |
| ADL limitationsd (range: 0–6) | 0.28 (0.84) |
| Memory functioning levele (range: 0–20) | 10.15 (3.16) |
| Worked for pay (%) | 37.42 |
| Positive relationship with spousef (range: 0–4) | 3.51 (0.61) |
| Negative relationship with spousef (range: 0–4) | 1.91 (0.65) |
| Positive relationship with childrenf (range: 0–4) | 3.26 (0.73) |
| Negative relationship with childrenf (range: 0–4) | 1.69 (0.64) |
Notes: N = 4,532. HRS = Health and Retirement Study.
aEight-item Center for Epidemiological Studies Depression scale.
bRated from 1 (less than once a year or never) to 6 (three or more times a week).
cRated from 1 (not at all) to 4 (a lot).
dADL = activities of daily living.
eMeasured as the sum of immediate and delayed word recall.
fRated from 1 (not at all) to 4 (a lot).
Measurement Model for Friendship
The estimates for the measurement model for the latent friendship variable are presented in Table 2. The model fit statistics showed that the measurement model fit the data well: χ 2(3) = 1,021.23 (p < .001), RMSEA = 0.00 (90% confidence interval [CI] = [0.00, 0.00], CFI = 1.00, TLI = 1.00, and SRMR = 0.00. The factor loadings for the three items were acceptable and statistically significant at p < .001, and the standardized estimates for the factor loadings were 0.60 (friend social contact frequency), 0.60 (positive friendship quality), and 0.44 (number of close friends).
Measurement Model for Friendship Support
| Unstandardized estimate | SE | Standardized estimate | SE | |
|---|---|---|---|---|
| Friendship indicators | ||||
| Number of close friends | 1.00 | 0.00 | 0.44*** | 0.02 |
| Contact frequency with friends | 0.46*** | 0.03 | 0.60*** | 0.02 |
| Friendship quality with friends | 0.32*** | 0.02 | 0.60*** | 0.02 |
| Model fit | ||||
| χ 2(3) | 1,021.23*** | |||
| CFI | 1.00 | |||
| TLI | 1.00 | |||
| RMSEA | 0.00 | |||
| SRMR | 0.00 |
| Unstandardized estimate | SE | Standardized estimate | SE | |
|---|---|---|---|---|
| Friendship indicators | ||||
| Number of close friends | 1.00 | 0.00 | 0.44*** | 0.02 |
| Contact frequency with friends | 0.46*** | 0.03 | 0.60*** | 0.02 |
| Friendship quality with friends | 0.32*** | 0.02 | 0.60*** | 0.02 |
| Model fit | ||||
| χ 2(3) | 1,021.23*** | |||
| CFI | 1.00 | |||
| TLI | 1.00 | |||
| RMSEA | 0.00 | |||
| SRMR | 0.00 |
Notes: N = 4,532. CFI = Comparative fit index; RMSEA = root mean square error of approximation; SE = standard errors; SRMR = standardized root mean square residual; TLI = Tucker–Lewis index.
***p < .001.
Measurement Model for Friendship Support
| Unstandardized estimate | SE | Standardized estimate | SE | |
|---|---|---|---|---|
| Friendship indicators | ||||
| Number of close friends | 1.00 | 0.00 | 0.44*** | 0.02 |
| Contact frequency with friends | 0.46*** | 0.03 | 0.60*** | 0.02 |
| Friendship quality with friends | 0.32*** | 0.02 | 0.60*** | 0.02 |
| Model fit | ||||
| χ 2(3) | 1,021.23*** | |||
| CFI | 1.00 | |||
| TLI | 1.00 | |||
| RMSEA | 0.00 | |||
| SRMR | 0.00 |
| Unstandardized estimate | SE | Standardized estimate | SE | |
|---|---|---|---|---|
| Friendship indicators | ||||
| Number of close friends | 1.00 | 0.00 | 0.44*** | 0.02 |
| Contact frequency with friends | 0.46*** | 0.03 | 0.60*** | 0.02 |
| Friendship quality with friends | 0.32*** | 0.02 | 0.60*** | 0.02 |
| Model fit | ||||
| χ 2(3) | 1,021.23*** | |||
| CFI | 1.00 | |||
| TLI | 1.00 | |||
| RMSEA | 0.00 | |||
| SRMR | 0.00 |
Notes: N = 4,532. CFI = Comparative fit index; RMSEA = root mean square error of approximation; SE = standard errors; SRMR = standardized root mean square residual; TLI = Tucker–Lewis index.
***p < .001.
Structural Models of Volunteering, Friendship, and Depression
The results for the negative binomial regression models estimating the association between volunteering and friendship are presented in Table 3. For the unadjusted model, we found that older adults who reported volunteering for three different amounts of time in the previous 12 months reported higher levels of friendship support than non-volunteers (Model A: 1–99 hr: β = 0.23, p < .001; 100–199 hr: β = 0.20, p < .001; 200 hr and more: β = 0.28, p < .001). In the model adjusted for covariates, we also found a statistically significant relationship between volunteer hours and friendship (Model B: 1–99 hr: β = 0.17, p < .001, 100–199 hr: β = 0.15, p < .001; 200 hr and more: β = 0.23, p < .001).
Structural Equation Model Results for Volunteer Hours, Friendship Support, and Number of Depressive Symptoms
| Friendship | Depressive symptoms | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Model A | Model B | Model C | Model D | ||||||
| Variables | β | SE p | β | SE p | β | SE p | β | SE p | |
| Volunteered 1–99 hoursa | 0.23 0.20 0.28 | 0.02 | 0.17 0.152 0.23 −0.14 0.20 0.11 0.07 0.09 0.06 0.05 0.00 0.11 0.04 −0.04 0.06 −0.01 0.11 −0.00 0.17 −0.01 | 0.02 | −0.59 −0.73 −0.77 | 0.09 | −0.07 −0.14 −0.17 −0.04 0.16 −0.19 0.12 0.11 0.14 −0.09 −0.01 −0.03 −0.13 0.30 −0.22 −0.23 −0.19 0.21 −0.15 0.24 | 0.04 | |
| Volunteered 100–199 hoursa | 0.02 | 0.02 | 0.08 | 0.04 | |||||
| Volunteered 200 hours or abovea | 0.02 | 0.02 | 0.07 | 0.04 | |||||
| Age | 0.03 | 0.05 | |||||||
| Female | 0.02 | 0.04 | |||||||
| Education | 0.02 | 0.04 | |||||||
| Divorced or separatedb | 0.02 | 0.04 | |||||||
| Widowedb | 0.02 | 0.04 | |||||||
| Never marriedb | 0.02 | 0.04 | |||||||
| Non-Hispanic Blackc | 0.03 | 0.05 | |||||||
| Non-Hispanic Otherc | 0.02 | 0.04 | |||||||
| Hispanicc | 0.03 | 0.06 | |||||||
| Household wealthd | 0.02 | 0.04 | |||||||
| ADL limitationse | 0.02 | 0.03 | |||||||
| Memory functioning level | 0.02 | 0.04 | |||||||
| Worked for pay | 0.02 | 0.04 | |||||||
| Positive relationship with spouse | 0.03 | 0.06 | |||||||
| Negative relationship with spouse | 0.03 | 0.06 | |||||||
| Positive relationship with children | 0.03 | 0.05 | |||||||
| Negative relationship with children | 0.03 | 0.05 | |||||||
| Model fit | |||||||||
| −2 log-likelihood | 52,402.11 | 208,507.44 | 22,636.70 | 178,541.74 | |||||
| AIC | 52,444.11 | 209,025.45 | 22,754.70 | 179,045.74 | |||||
| BIC | 52,578.91 | 210,687.95 | 22,754.57 | 180,663.31 |
| Friendship | Depressive symptoms | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Model A | Model B | Model C | Model D | ||||||
| Variables | β | SE p | β | SE p | β | SE p | β | SE p | |
| Volunteered 1–99 hoursa | 0.23 0.20 0.28 | 0.02 | 0.17 0.152 0.23 −0.14 0.20 0.11 0.07 0.09 0.06 0.05 0.00 0.11 0.04 −0.04 0.06 −0.01 0.11 −0.00 0.17 −0.01 | 0.02 | −0.59 −0.73 −0.77 | 0.09 | −0.07 −0.14 −0.17 −0.04 0.16 −0.19 0.12 0.11 0.14 −0.09 −0.01 −0.03 −0.13 0.30 −0.22 −0.23 −0.19 0.21 −0.15 0.24 | 0.04 | |
| Volunteered 100–199 hoursa | 0.02 | 0.02 | 0.08 | 0.04 | |||||
| Volunteered 200 hours or abovea | 0.02 | 0.02 | 0.07 | 0.04 | |||||
| Age | 0.03 | 0.05 | |||||||
| Female | 0.02 | 0.04 | |||||||
| Education | 0.02 | 0.04 | |||||||
| Divorced or separatedb | 0.02 | 0.04 | |||||||
| Widowedb | 0.02 | 0.04 | |||||||
| Never marriedb | 0.02 | 0.04 | |||||||
| Non-Hispanic Blackc | 0.03 | 0.05 | |||||||
| Non-Hispanic Otherc | 0.02 | 0.04 | |||||||
| Hispanicc | 0.03 | 0.06 | |||||||
| Household wealthd | 0.02 | 0.04 | |||||||
| ADL limitationse | 0.02 | 0.03 | |||||||
| Memory functioning level | 0.02 | 0.04 | |||||||
| Worked for pay | 0.02 | 0.04 | |||||||
| Positive relationship with spouse | 0.03 | 0.06 | |||||||
| Negative relationship with spouse | 0.03 | 0.06 | |||||||
| Positive relationship with children | 0.03 | 0.05 | |||||||
| Negative relationship with children | 0.03 | 0.05 | |||||||
| Model fit | |||||||||
| −2 log-likelihood | 52,402.11 | 208,507.44 | 22,636.70 | 178,541.74 | |||||
| AIC | 52,444.11 | 209,025.45 | 22,754.70 | 179,045.74 | |||||
| BIC | 52,578.91 | 210,687.95 | 22,754.57 | 180,663.31 |
Notes: N = 4,532. SE = standard errors. Standardized coefficients are reported.
aReference group = No volunteered hours.
bReference group = married.
cReference group = Non-Hispanic White.
dLog-transformed.
eADL = activities of daily living; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion.
*p < .05,
**p < .01,
***p < .001.
Structural Equation Model Results for Volunteer Hours, Friendship Support, and Number of Depressive Symptoms
| Friendship | Depressive symptoms | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Model A | Model B | Model C | Model D | ||||||
| Variables | β | SE p | β | SE p | β | SE p | β | SE p | |
| Volunteered 1–99 hoursa | 0.23 0.20 0.28 | 0.02 | 0.17 0.152 0.23 −0.14 0.20 0.11 0.07 0.09 0.06 0.05 0.00 0.11 0.04 −0.04 0.06 −0.01 0.11 −0.00 0.17 −0.01 | 0.02 | −0.59 −0.73 −0.77 | 0.09 | −0.07 −0.14 −0.17 −0.04 0.16 −0.19 0.12 0.11 0.14 −0.09 −0.01 −0.03 −0.13 0.30 −0.22 −0.23 −0.19 0.21 −0.15 0.24 | 0.04 | |
| Volunteered 100–199 hoursa | 0.02 | 0.02 | 0.08 | 0.04 | |||||
| Volunteered 200 hours or abovea | 0.02 | 0.02 | 0.07 | 0.04 | |||||
| Age | 0.03 | 0.05 | |||||||
| Female | 0.02 | 0.04 | |||||||
| Education | 0.02 | 0.04 | |||||||
| Divorced or separatedb | 0.02 | 0.04 | |||||||
| Widowedb | 0.02 | 0.04 | |||||||
| Never marriedb | 0.02 | 0.04 | |||||||
| Non-Hispanic Blackc | 0.03 | 0.05 | |||||||
| Non-Hispanic Otherc | 0.02 | 0.04 | |||||||
| Hispanicc | 0.03 | 0.06 | |||||||
| Household wealthd | 0.02 | 0.04 | |||||||
| ADL limitationse | 0.02 | 0.03 | |||||||
| Memory functioning level | 0.02 | 0.04 | |||||||
| Worked for pay | 0.02 | 0.04 | |||||||
| Positive relationship with spouse | 0.03 | 0.06 | |||||||
| Negative relationship with spouse | 0.03 | 0.06 | |||||||
| Positive relationship with children | 0.03 | 0.05 | |||||||
| Negative relationship with children | 0.03 | 0.05 | |||||||
| Model fit | |||||||||
| −2 log-likelihood | 52,402.11 | 208,507.44 | 22,636.70 | 178,541.74 | |||||
| AIC | 52,444.11 | 209,025.45 | 22,754.70 | 179,045.74 | |||||
| BIC | 52,578.91 | 210,687.95 | 22,754.57 | 180,663.31 |
| Friendship | Depressive symptoms | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Model A | Model B | Model C | Model D | ||||||
| Variables | β | SE p | β | SE p | β | SE p | β | SE p | |
| Volunteered 1–99 hoursa | 0.23 0.20 0.28 | 0.02 | 0.17 0.152 0.23 −0.14 0.20 0.11 0.07 0.09 0.06 0.05 0.00 0.11 0.04 −0.04 0.06 −0.01 0.11 −0.00 0.17 −0.01 | 0.02 | −0.59 −0.73 −0.77 | 0.09 | −0.07 −0.14 −0.17 −0.04 0.16 −0.19 0.12 0.11 0.14 −0.09 −0.01 −0.03 −0.13 0.30 −0.22 −0.23 −0.19 0.21 −0.15 0.24 | 0.04 | |
| Volunteered 100–199 hoursa | 0.02 | 0.02 | 0.08 | 0.04 | |||||
| Volunteered 200 hours or abovea | 0.02 | 0.02 | 0.07 | 0.04 | |||||
| Age | 0.03 | 0.05 | |||||||
| Female | 0.02 | 0.04 | |||||||
| Education | 0.02 | 0.04 | |||||||
| Divorced or separatedb | 0.02 | 0.04 | |||||||
| Widowedb | 0.02 | 0.04 | |||||||
| Never marriedb | 0.02 | 0.04 | |||||||
| Non-Hispanic Blackc | 0.03 | 0.05 | |||||||
| Non-Hispanic Otherc | 0.02 | 0.04 | |||||||
| Hispanicc | 0.03 | 0.06 | |||||||
| Household wealthd | 0.02 | 0.04 | |||||||
| ADL limitationse | 0.02 | 0.03 | |||||||
| Memory functioning level | 0.02 | 0.04 | |||||||
| Worked for pay | 0.02 | 0.04 | |||||||
| Positive relationship with spouse | 0.03 | 0.06 | |||||||
| Negative relationship with spouse | 0.03 | 0.06 | |||||||
| Positive relationship with children | 0.03 | 0.05 | |||||||
| Negative relationship with children | 0.03 | 0.05 | |||||||
| Model fit | |||||||||
| −2 log-likelihood | 52,402.11 | 208,507.44 | 22,636.70 | 178,541.74 | |||||
| AIC | 52,444.11 | 209,025.45 | 22,754.70 | 179,045.74 | |||||
| BIC | 52,578.91 | 210,687.95 | 22,754.57 | 180,663.31 |
Notes: N = 4,532. SE = standard errors. Standardized coefficients are reported.
aReference group = No volunteered hours.
bReference group = married.
cReference group = Non-Hispanic White.
dLog-transformed.
eADL = activities of daily living; AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion.
*p < .05,
**p < .01,
***p < .001.
Next, we examined whether the number of hours volunteered was associated with number of depressive symptoms. In the unadjusted model, older adults who volunteered at each level reported fewer depressive symptoms four years later compared to non-volunteers (Model C: 1–99 hr: β = −0.59, p < .001, 100–199 hr: β = −0.73, p < .001, 200 hr and more: β = −0.77, p < .001). In the adjusted model, participants who volunteered 100–199 hr and those who volunteered 200 hr or more reported fewer depressive symptoms compared to non-volunteer participants, but no significant difference was found between non-volunteers and those who volunteered 1–99 hr (Model D: 1–99 hr: β = −0.07, p = .06, 100–199 hr: β = −0.14, p < .001, 200 hr and more: β = −0.17, p < .001).
Finally, we examined the mediating effects of friendship characteristics for the association between hours volunteered and depressive symptoms (Figure 1; Log-likelihood = 222,287.20, AIC = 222,817.20, and BIC = 224,518.21). The results showed that compared to non-volunteers’ hours volunteered was significantly associated with friendship (1–99 hr: β = 0.23, p < .001, 100–199 hr: β = 0.20, p < .001, 200 hr and more: β = 0.28, p < .001), and friendship characteristics was significantly associated with number of depressive symptoms (β = −0.12, p = .03). Estimates of the total, direct, and indirect effects for the relationships between volunteer hours, friendship, and number of depressive symptoms are presented in Table 4. Friendship had a significant indirect effect on the association between hours volunteered and depressive symptoms (1–99 hr: β = −0.01, 95% CI = [−0.02, −0.00], p = .03), 100–199 hr: β = −0.01, 95% CI = [−0.02, −0.00], p = .03), 200 hr and more: β = −0.02, 95% CI = [−0.03, −0.00], p = .03. The direct effect of hours volunteered and depressive symptoms are also significant, except for those who volunteered 1–99 hr (1–99 hr: β = −0.03, p = .21, 100–199 hr: β = −0.06, p = .004, 200 hr and more: β = −0.07, p = .002). The total effect of the association between hours volunteered and number of depressive symptoms was also significant (1–99 hr: β = −0.04, p = .46, 100–199 hr: β = −0.07, p < .001, 200 hr and more: β = −0.08, p < .001). The proportions of the total effect showing that friendship mediated hours volunteered and depressive symptoms were 25.0% (1–99 hr), 14.3% (100–199 hr), and 25.0% (200 hr or above). Overall, friendship was a pathway for the association between hours volunteered and depressive symptoms.
Direct, Indirect, and Total Effects of Volunteer Hours on Depressive Symptoms
| Standardized estimate | SE | 95% CI | p | |
|---|---|---|---|---|
| 1–99 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.03 | 0.02 | [−0.06, 0.01] | .21 |
| Indirect effect (via Friendship in 2014) | −0.01 −0.04 | 0.01 | [−0.02, −0.00] | .03 |
| Total effect | 0.02 | [−0.07, −0.01] | .05 | |
| 100–199 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.06 −0.01 −0.07 | 0.02 | [−0.10, 0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.02, −0.00] | .03 | |
| Total effect | 0.02 | [−0.11, −0.04] | <.001 | |
| 200 or more hours (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.07 −0.02 −0.08 | 0.02 | [−0.10, −0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.03, −0.00] | .03 | |
| Total effect | 0.02 | [−0.12, −0.05] | <.001 |
| Standardized estimate | SE | 95% CI | p | |
|---|---|---|---|---|
| 1–99 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.03 | 0.02 | [−0.06, 0.01] | .21 |
| Indirect effect (via Friendship in 2014) | −0.01 −0.04 | 0.01 | [−0.02, −0.00] | .03 |
| Total effect | 0.02 | [−0.07, −0.01] | .05 | |
| 100–199 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.06 −0.01 −0.07 | 0.02 | [−0.10, 0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.02, −0.00] | .03 | |
| Total effect | 0.02 | [−0.11, −0.04] | <.001 | |
| 200 or more hours (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.07 −0.02 −0.08 | 0.02 | [−0.10, −0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.03, −0.00] | .03 | |
| Total effect | 0.02 | [−0.12, −0.05] | <.001 |
Notes: N = 4,532. CI = confidence intervals; SE = standard errors. Standardized coefficients are reported.
*p < .05,
**p < .01,
***p < .001.
Direct, Indirect, and Total Effects of Volunteer Hours on Depressive Symptoms
| Standardized estimate | SE | 95% CI | p | |
|---|---|---|---|---|
| 1–99 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.03 | 0.02 | [−0.06, 0.01] | .21 |
| Indirect effect (via Friendship in 2014) | −0.01 −0.04 | 0.01 | [−0.02, −0.00] | .03 |
| Total effect | 0.02 | [−0.07, −0.01] | .05 | |
| 100–199 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.06 −0.01 −0.07 | 0.02 | [−0.10, 0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.02, −0.00] | .03 | |
| Total effect | 0.02 | [−0.11, −0.04] | <.001 | |
| 200 or more hours (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.07 −0.02 −0.08 | 0.02 | [−0.10, −0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.03, −0.00] | .03 | |
| Total effect | 0.02 | [−0.12, −0.05] | <.001 |
| Standardized estimate | SE | 95% CI | p | |
|---|---|---|---|---|
| 1–99 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.03 | 0.02 | [−0.06, 0.01] | .21 |
| Indirect effect (via Friendship in 2014) | −0.01 −0.04 | 0.01 | [−0.02, −0.00] | .03 |
| Total effect | 0.02 | [−0.07, −0.01] | .05 | |
| 100–199 hr (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.06 −0.01 −0.07 | 0.02 | [−0.10, 0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.02, −0.00] | .03 | |
| Total effect | 0.02 | [−0.11, −0.04] | <.001 | |
| 200 or more hours (2010) → Depressive symptoms (2018) | ||||
| Direct | −0.07 −0.02 −0.08 | 0.02 | [−0.10, −0.03] | .00 |
| Indirect effect (via Friendship in 2014) | 0.01 | [−0.03, −0.00] | .03 | |
| Total effect | 0.02 | [−0.12, −0.05] | <.001 |
Notes: N = 4,532. CI = confidence intervals; SE = standard errors. Standardized coefficients are reported.
*p < .05,
**p < .01,
***p < .001.
The indirect effect of friendship in the association between volunteering and depressive symptoms. Dotted and bold lines indicate insignificant and significant pathways, respectively. Covariates were controlled for outcome variable and mediator, including age, gender, education, marital status, race-ethnic status, household wealth (logged), ADL limitations, working status, memory functioning level, positive and negative relationship quality with spouse and children. ADL = activities of daily living.
Discussion
The association between volunteering and mental health among older adults, especially depression, has been the subject of a considerable amount of research (Burr et al., 2021). Research shows that depression is related to poor quality of life among older adults because they experience loss of appetite, sleep deprivation, and less ability to concentrate (Hussenoeder et al., 2021). Over time, depressive symptoms have negative health consequences, including higher risk for physical frailty, cognitive decline, and mortality (van den Berg et al., 2021; Borges et al., 2021; Olaya et al., 2019). However, not much is known about the benefits of friendship and volunteering for number depressive symptoms, especially whether friendship links volunteering and depressive symptoms (Filges et al., 2020). Hence, we examined whether friendship was a possible mechanism that links volunteering and depressive symptoms. Our study results uncovered the potential benefits of volunteering as related to older adults’ friendship support, and in turn, how friendship support played a role in the link between volunteer hours and number of depressive symptoms for older adults.
Volunteering and Depressive Symptoms
The results supported our first hypothesis by showing that number of hours volunteered was negatively associated with number of depressive symptoms but only among those who volunteered 100 or more hours (about 2 hr per week). This finding was consistent with existing studies that showed volunteer hours were beneficial for those who volunteered 100 or more hours in terms of depressive symptoms (Huo et al., 2021; Webster et al., 2021). These findings were suggestive of a possible dose-effect for formal volunteering behavior in terms of depression risk among older adults, although further research is needed to confirm this possibility. A possible explanation for the dose-effect of formal volunteering and depression risk could be that volunteering for 100 or more hours is the amount that allows older adults’ access to social resources that may boost their sense of meaning in life, self-esteem, and self-worth, which then reduces their depressive mood (Webster et al., 2021). Overall, our results were consistent with prior literature that volunteering status is beneficial to depression and a minimum number of hours volunteered may be needed for the salutary effect of volunteering to emerge (Huo et al., 2021).
Volunteering and Friendship
Our second hypothesis that volunteering more hours was related to better friendships was supported. Our study results were in line with the Focus Theory of Social Interaction and consistent with past research showing that volunteering is associated with better friendship (Guiney et al., 2021; Jang et al., 2018; Rook & Sorkin, 2003). Volunteering is considered the “focus” event that brought older adults together with others and provided them with the opportunity to establish new friendships or strengthen existing ones (Feld, 1981; Häuberer & Häuberer, 2014; Paik & Navarre-Jackson, 2011). Given the limited research available, more effort should be aimed at examining which types of volunteering generate more friends and higher quality friendships.
Friendship and Depression
Further, the finding that better friendships were associated with fewer depressive symptoms supported our third hypothesis and this was in line with the extant literature showing that friendships offered protection against possible depression as measured by depressive symptoms among older adults (Bookwala, 2017; Han et al., 2019; Kail & Carr, 2020; Teo et al., 2015). Friendship may alleviate depression risk because having friends, staying connected with friends, and receiving support from friends increase meaning in life and promote a sense of belonging among older adults (Blieszner et al., 2019; Bruggencate et al., 2018).
The Mediating Role of Friendship
Finally, the results also supported our fourth research hypothesis that friendship would mediate the relationship between volunteering and depressive symptoms. This finding was generally consistent with research reported by Creaven et al. (2018) and Pilkington et al. (2012). As older adults age, many lose social ties due to poor health, limited mobility, geographic relocation, or the death of significant others (Nicolaisen & Thorsen, 2017). These losses have the potential to increase the risk of being lonely and socially isolated. However, the process of making friends and staying connected socially through volunteering allowed older adults to meet their social needs, especially in the context of diminishing social ties, which reduces the risk of experiencing depression (Bruggencate et al., 2018). Thus, our study demonstrated friendship is a likely pathway that links volunteering and depression.
Overall, our longitudinal study design provided evidence of the salubrious effect of volunteering for reducing risk of depressive symptoms and also for the enduring effect of friendship for reducing the risk of poor emotional well-being. Our findings suggested that one intervention strategy for health care providers treating middle-aged persons for depressive symptoms may be to encourage them to seek out volunteer opportunities after retirement. As middle-aged adults move into the retirement life stage they often experience role loss, making them susceptible to depression (Han, 2021). Volunteering provides a social activity that may be protective against the negative consequences of retirement (Filges et al., 2020). Further, friendships that emanate from some forms of volunteering may help older adults facilitate a healthy transition into retirement (Kail and Carr, 2020). One example of the benefits of volunteering is found in research on the Experience Corps, where older elementary school volunteers have benefitted along several health dimensions (Fried et al., 2004). Hence, finding ways to encourage middle-aged and older adults to volunteer, if they wish and are able to engage in this form of prosocial behavior, could be a strategy to help them continue to build new friendships and maintain old ones, staying socially engaged to compensate for the loss of social ties and role loss, and thus reduce depression risk.
Limitations and Future Directions
The study had limitations. First, given the observational nature of the research design, reciprocal relationships among volunteering, friendship, and depressive symptoms cannot be ruled out. Older adults with fewer depressive symptoms may be more likely to develop and maintain friendships and to be volunteers. Future studies should examine alternative models of the relationships between volunteering, friendship, and depression. For example, depression may be a predictor of friendship outcomes, and friendship may be a predictor of volunteering behavior. This will help gain a better understanding of the dynamics of the relationships among these three variables. Finally, future studies using longitudinal experimental research designs with multiple time intervals and appropriate sample size may help to unpack these complex associations and lend confidence in our findings.
Second, we examined the benefits of the number of volunteer hours (with time thresholds determined by the HRS survey questions) for friendship support and fewer depressive symptoms. Future work may explore volunteer hours measured continuously to better detect thresholds, possible dose-effects, and nonlinear effects. Additionally, the HRS questionnaire asked participants to indicate whether they volunteered in the past 12 months. This item was followed by a question on number of hours of volunteering. Because of the relatively long-time frame in the question, recall bias is possible, as participants may not accurately report their volunteered hours (Kjellsson et al., 2014). This is an inherent limitation of the HRS. Future research should use a volunteer hours variable with shorter recall periods (e.g., 1 month, 6 months) and results from such studies should be compared to the results reported here. Further, future research should determine which types of formal volunteering help older adults develop better friendship support and reduce depressive symptoms, an option that is not possible with the HRS. Volunteering roles that include a social component increase opportunities for making friends, such as volunteering for public schools, religious organizations, and civic groups. These types of organizations provide well-defined roles that allow individuals to feel appreciated and to experience a greater sense of self-achievement, reducing the likelihood of depression (Li & Ferraro, 2005; Matthews & Nazroo, 2021). Thus, future studies about friendship and depression should include an evaluation of specific types of volunteering work.
Third, further research on informal volunteering and friendship is also warranted. Informal volunteering is conceptualized as providing support to persons who do not live with the older adults (Burr et al., 2021). Informal volunteering takes place in unstructured environments and is often intermittent (Dean, 2021). As another form of productive aging, future research should evaluate whether friendship mediates the relationship between informal volunteering and emotional well-being.
Conclusion
Using a large national sample from the HRS, our study contributed evidence to the literature on volunteering and depression in later life and explored one underlying mechanism, friendship, through which volunteering is associated with fewer depressive symptoms. Our findings suggested that volunteering was significantly associated with friendship and depression, friendship was also significantly associated with depression, and friendship mediated the relationship between volunteering and depressive symptoms, demonstrating that friendship is one possible pathway. The findings provided implications for practice and interventions, for example, programs aiming at reducing barriers for people who wish to participate in volunteering activities may help them to be more socially engaged, increasing social interactions with others, and providing support for coping with psychological disorders and stress. Further, for those who are socially isolated or who feel lonely due to constricted friendship networks, communities and organizations may help older adults develop and maintain better friendships through creating volunteer opportunities.
Funding
None declared.
Conflict of Interest
None declared.
Author Contributions
E. Lim planned, conducted the data analysis, and wrote the paper. C. Peng helped with the planning of data analysis and contributed to the revising of the paper. J. A. Burr contributed to the planning and revising of the manuscript.
