Abstract
This study aims to further our understanding of formal volunteering as a protective mechanism for health in the context of housing relocation and to explore race, gender, and education as moderators. A quasi-experimental design evaluated the effects of volunteering on older adults’ health (self-report health, number of instrumental activities of daily living [IADLs], and depressive symptoms) among individuals who relocated but did not volunteer at Time 1 (N = 682) in the Health and Retirement Study (2008–2010). Propensity score weighting examined health differences at Time 2 between 166 volunteers (treated) and 516 nonvolunteers (controlled). Interaction terms tested moderation. Individuals who moved and engaged in volunteering reported higher levels of self-rated health and fewer IADL difficulties compared to the control group. Race moderated the relationship between volunteering and depressive symptoms, while gender moderated the relationship between volunteering and self-assessed health. Formal volunteering protects different dimensions of health after relocation. Volunteering was particularly beneficial for females and older Whites.
The aging of the population presents many challenges and opportunities for researchers, policy makers, and practitioners to identify risks and protective factors for health. Gerontologists have posited that formal volunteering can bolster health, particularly during life transitions in which there is a health risk (Greenfield & Marks, 2004; Morrow-Howell, Hinterlong, Rozario & Tang, 2003). This study contributes to the current body of research by employing a longitudinal quasi-experimental design to panel data, the Health and Retirement Study (HRS), to examine formal volunteering as a protective mechanism for health in the context of relocation. This study also explores whether and how key social determinants of health (race, gender, and education) may influence the relationship. To our knowledge, this intersection of relocation, productive aging, and health remains uncharted in the scientific literature, yet holds much promise for research, policy, and practice.
Theoretical and Empirical Background
The relationship of older adults to their homes and community has been shown to contribute to their aging processes (see Iwarsson et al., 2007, for description of ENABLE-AGE). However, older adults may relocate despite preferences to remain in their homes; relocation occurs throughout the life span for some older adults and has been shown to impact several aspects of health. There is a compelling body of research with population-based longitudinal data showing that relocation can compromise mental health, physical health, and self-rated health status (Bradley & Van Willigen, 2010; Hong, Hasche, & Bowland, 2009; Walker, Curry, & Hogstel, 2007). There are many reasons why older adults move or do not move; the most common ones are related to one’s health, cognitive, and financial statuses (Torres-Gil & Harland, 2012; Wiseman, 1980). Some relocate due to death of a spouse, preference to live in a different climate and/or community, or changes in health status (Litwak & Longino, 1987; Wiseman, 1980; Perry, Andersen & Kaplan, 2014). One such move, an amenity move, often requires numerous resources including health to plan and execute one’s moving plans and financial resources to undertake a voluntary move (Litwak & Longino, 1987). Changes in health status may lead to relocation near one’s kin or to institutional care when kin or other informal supports are not sufficient (Haas & Serow, 1993; Litwark & Longino, 1987). In addition to the triggers why older adults relocate, scholars are developing important schemas for understanding how these decisions are made (see Granbom et al.’s [2014] important work on residential reasoning) and how we can understand the ways older adults actively and proactively cope with their “incongruent residential environment” (Golant, 2011, 2014). Koss and Ekerdt’s (2017) qualitative study highlights the importance of peers in the timing and geography of later moves or even if they planned to move, as older adults imagine a future that involves changes in health status. Pope and Kang (2010) also found differences between older adults who move proactively and those for whom their moves are reactive. The latter are typically older, have lower incomes, and lack insurance. Thus, Pope and Kang suggest that planned moves may have a better outcome for physical and emotional well-being.
Evidence stemming from the ecological framework (Lawton & Simon, 1968) has shown that relocation can disrupt the aging process at multiple levels and dimensions, which in turn influence physical and mental health. Functional, symbolic, and emotional attachments to home, neighborhood, and community (Wiles, Leibing, Guberman, Reeve, & Ruth, 2011); the dispersal and/or storage of material possessions; and changes in social networks can elevate biopsychosocial mechanisms (Lutgendorf et al., 2001).
Recent research on triggers to relocation considers the importance of changing and interconnected family structures including divorce and interaction with grandchildren as a trigger to relocation rather than previous work-framing moves toward family in terms of older adults being care recipients (Winke, 2017). In addition to kin, the lack of quality friendships has been shown to have higher depressive symptoms and lower perceptions of safety (Stafford, Mcmunn, & De Vogli, 2011).
Beyond the importance of family and friends, the inclusion of a productive aging framework as a relocation trigger is an important step to understanding the complex reasons that older adults relocate. This approach builds on existing scholarship on the importance of social participation to health and well-being. In fact, opportunities for social reciprocity, such as volunteering, have been identified as a key theme in promoting age-friendly communities (Emlet & Moceri, 2012). Notably is the variability among older adults, as research has shown a relationship between types of housing and neighborhoods and social participation (Perry, Ruggiano, Shtompel, & Hassevoort, 2015; Perry, Anderson, & Kaplan, 2014). Comparing with their counterparts, older adults living in less disadvantaged residences and neighborhoods have lower social participation and decreased overall well-being (Tomaszewski & Barnes, 2008; Tomaszewski, 2013). With regard to relocation, volunteering has been identified as a stabilizing factor that makes it more likely that older adults will age in place (Shen & Perry, 2014, 2016), and it may also be a protective factor to health when relocation does occur. Identifying protective factors to health in the context of relocation can enhance the quality and length of life and potentially help delay the onset of morbidity and mortality.
Formal volunteering as a protective factor in the face of risk
Formal volunteering helps to shape health behavior through social and physical activity, psychological engagement, and cognitive stimulation (Matz-Costa, Carr, McNamara, & James, 2015). Longitudinal studies have linked formal volunteering to improvements in self-rated health (Morrow-Howell, Hinterlong, Rozario, & Tang, 2003), lower levels of functional dependence and depressive symptomatology (Morrow-Howell et al., 2003), elevated levels of life satisfaction (Van Willigen, 2000), and lower risk levels of mortality (Lum & Lightfoot, 2005; Musick, Herzog, & House, 1999). Experimental and quasi-experimental research suggests older volunteers in a high-intensity intergenerational program experience decreased frailty and falls; fewer depressive symptoms; enhanced memory, strength, balance, walking speed, cortical plasticity, and executive functioning; increased social and psychological engagement; and improved mobility and instrumental activities of daily living (IADLs; Carlson et al., 2015; Fried et al., 2013; Hong & Morrow-Howell, 2010).
Greenfield and Marks (2004) speculated that formal volunteering could be a potential health protector in the context of risk, particularly with major role-identity absences (e.g., retirement) and losses (e.g., spouse). Utilizing a cross-sectional population data set, Greenfield and Marks’ (2004) analyses supported the theory that major role loss is associated with poorer psychological health; they also found that formal volunteering moderated the relationship, with volunteers reporting higher levels of psychological well-being. In this study, we extend the framework to explore the impacts of relocation. This theoretical framework and the empirical findings lead to the following hypotheses:
Heterogeneity of aging experiences
Race, gender, and education are important social determinants of health and reflect the heterogeneous pathways and experiences of aging. Health inequities by these key social determinants are a result of disparate access to education, certain occupations, health care, and safe neighborhoods across the life span and are factors of the cumulative (dis)advantage concept (Gonzales, Matz-Costa, & Morrow-Howell, 2015; Shen & Perry, 2016).
With regard to race, Blacks are more likely to be unmarried (Pugliesi & Shook, 1998), have a smaller social network (Ajrouch, Antonucci, & Janevic, 2001), and have fewer opportunities to formally volunteer in later life (Gonzales et al., 2015) when compared to their White counterparts. While older Whites tend to volunteer at higher rates, older Blacks volunteer more annual hours than their White counterparts (Musick & Wilson, 2008; Tang, Copeland, & Wexler, 2012), offering evidence of a stronger social bond to formal organizations such as educational institutions, places of worship, hospitals, and charitable organizations. McBride, Gonzales, Morrow-Howell, and McCrary (2011) and Tan et al. (2016) document how certain institutional arrangements, such as providing a stipend, can help to expand access to and the benefits of volunteering among older populations that have been historically excluded racial and ethnic minorities. McBride et al. (2011) and Tang et al. (2012) found that older racial and ethnic minorities reported greater health and social benefits from volunteering when compared to older Whites.
Gender norms also impact lifetime trajectories in the formal labor force, family, and the civil sector. Women have larger social networks and higher rates of formal volunteer engagement before and after retirement than men (Musick & Wilson, 2008), leading some to speculate that gender may moderate the relationship between volunteer status and health outcomes (Greenfield & Marks, 2004; Shen & Perry, 2016; McIntosh, & Danigelis, 1995).
Finally, education is a key component of the social gradient. When viewed from a life-course perspective, more education often predicts better access to well-paid occupations with benefits, health care, job security, locus of control, and healthy life styles, all of which shape and improve health as individuals age (Centers for Disease Control, 2012; Schafer, Wilkinson, & Ferraro, 2013). Research that examines whether and how health may be protected by formal volunteering along key social determinants can help to shape targeted social policies and practices to address health disparities. Given these observations:
Method
Data and Sample
Data were drawn from the 2008 and 2010 HRS. Since 1992, HRS surveys people over the age of 50, as well as their spouses or partners, and follows up the same respondents biennially (Servais, 2010). Using a multistage area probability sample design, HRS oversamples African American, Hispanic, and Floridian respondents and adds new cohorts every 6 years to ensure its representativeness of the United States.
To explore whether volunteering can buffer health after relocation, the present study started with a subsample of respondents who participated in both the 2008 and 2010 HRS surveys. Inclusion criteria were individuals aged 51 and older who did not volunteer in 2008 and who relocated between 2008 and 2010. Relocation status was defined as “no move since 2008,” “moved within area” (39.59% of sample) if respondents had changed residence but not geographic area by 2010, and “moved out of area” (60.41% of sample); those in the latter two groups were considered to have relocated and were included in our sample. The final sample consisted of 682 older adults who had data from all the study variables.
Measures
Dependent variables
Four dependent health variables measured in 2010 were examined: self-rated health status, ADLs, IADLs, and depressive symptoms.
Self-rated health
It was measured with “Would you say your health is excellent, very good, good, fair, or poor?” and dichotomized, where “good, very good, excellent” was coded 1, and “poor and fair” was coded 0.
ADLs
The number of ADL tasks an older person reported getting help with, not conducting, or having difficulty performing was assessed. Six ADL tasks included dressing, bathing, eating, toileting, walking, and getting in and out of bed.
IADLs
IADLs were more complex tasks that required a certain amount of physical, mental, and organizational skills. The four IADL tasks were preparing a hot meal, shopping for groceries, making a telephone call, and taking medications.
Depressive symptoms
These were assessed with the 8-item version of the Center for Epidemiologic Studies–Depression Scale (CES-D). The items measured three subscales: depressed mood (depressed, lonely, and sad), psychomotor retardation (everything an effort, restless sleep, and couldn’t get going), and lack of well-being (reverse-coded happy, enjoyed life). A respondent’s CES-D score was the sum of the 6 negative items and the 2 reverse-coded positive items. CES-D score ranged from 0 to 8, with the higher score indicating more depressive symptoms.
Key predictor and moderating variables
Formal volunteering
It was assessed with the question “Have you spent any time in the past 12 months doing volunteer work for religious, educational, health-related or other charitable organizations?” (yes = 1). In 2010, there were 166 volunteers and 516 nonvolunteers.
Race
Race was transformed into a dichotomous variable using race and ethnicity categories, where White was coded 1 and non-White was 0 (Blacks, Hispanic, and Other).
Gender
Gender was a binary variable, with male coded 1 and female 0.
Education
It referred to the years of formal education completed, ranging between 0 and 17 years.
Propensity score weighting
A literature review helped us to identify 22 key variables from the 2008 HRS to predict the probability of volunteering in 2010 using a generalized boosted model (GBM; Guo & Fraser, 2015; McNamara & Gonzales, 2011; Greenfield & Marks, 2004; Hong & Morrow-Howell, 2010; Musick & Wilson, 2008; Gonzales, Matz-Costa, & Morrow-Howell, 2015). Demographic, health, economic, and social variables that were expected to generate the satisfactory propensity score were chosen.
Demographic predictors were age, gender, and race.
Health predictors included self-rated health, number of ADL limitations, number of IADL limitations, depressive symptoms, number of chronic health conditions (diabetes, heart condition, stroke, lung disease, cancer, arthritis, psychiatric problems, urine control, and legal blindness or very poor eyesight), having cognition limitations (identified as having a cognitive problem, coded 1, if he or she failed to correctly answer 50% or more of the questions in the telephone interview for cognitive status), self-rated memory status (reverse coded, where 5 = excellent and 1 = poor), and life satisfaction (reverse coded, where 5 = completely satisfied and 1 = not at all satisfied).
Economic predictors included education (0–17 completed years), annual household income in the past 12 months with five categories (US$0–20,000 = 1, US$20,001–40,000 = 2, US$40,001–60,000 = 3, US$60,001–100,000 = 4, and US$100,001+ = 5), total household wealth was collected as a continuous variable and transformed using the inverse hyperbolic sine function (STATA 14 command, ASINH), a preferable method to handle skewness, retain zero and negative values, and avoid stacking and disproportionately misrepresentation (Friedline, Masa, & Chowa, 2015).
Social predictors included marital status (married or not), relatives living nearby, friends living nearby, number of social visits per month, parental caregiving (older adults spent a total of 100 or more hours in last 2 years helping a parent with ADLs or IADLs), spousal caregiving (older adults helped the spouse with care needs in the last month), employment (worked for pay), and importance of religion (“very important,” “somewhat important,” and “not too important”).
Analytical Strategies
Following the three steps in propensity score weighting (Guo & Fraser, 2015, p. 246), the GBM was first conducted to estimate the propensity score using a comprehensive set of sample observed covariates at the baseline (demographic, health, economic, and social). Observations were dropped using listwise deletion before the GBM, resulting a sample of 682 with 166 volunteers “treated” and 516 nonvolunteers “controlled”.
Second, using the estimated propensity score, two types of weights were calculated, one to estimate the average treatment effect (ATE) and the other to estimate the average treatment effect for the treated (ATT). ATE is defined as the average effect, at the population level, of moving an entire population from untreated to treated. In the context of this study, ATE refers to the average causal effect of volunteering among older adults who relocated and can be expressed as “What difference in health would we expect to observe if all older adults who relocated in our study had volunteered, compared to all of them who had not volunteered?” (Lanza, Moore, & Butera, 2013). The second, ATT is defined as the average effect of treatment on the subjects who received the treatment (Austin, 2011). Here, in this study, our treatment, volunteering is a social engagement measured by a self-reported question (Guo & Fraser, 2015).
ATE is the most commonly invoked treatment effect and focus of many investigations in economics and statistics (Heckman, 2005; Heckman & Vytlacil, 2005). ATT is a conventional parameter that is commonly paired with ATE (Heckman, 2005) and may be of greater interest in estimating the effectiveness of a social engagement (Austin, 2011; Lanza et al., 2013). We created both ATE and ATT weights using the propensity score and applied both weights in the outcome analyses.
For a model-predicted probability, P, the weight is 1/P for a treated participant and [1/(1 − P)] for a control participant when estimating ATE; the weight is 1 for a treated participant and [P/(1 − P)] for a control participant when estimating ATT (Guo & Fraser, 2015). The weights were used in multivariate analysis, which is analogous to multivariate modeling that incorporates sampling weights (Guo & Fraser, 2015). A differential amount of information from each participant depending on the participant’s conditional probability of receiving treatment (i.e., engaging in volunteering) was taken into consideration through the weighting.
To ensure any covariate imbalance was corrected after applying the propensity score estimated by the GBM, weighted simple regression models were conducted. Using the dichotomous treatment variable as the single independent variable and the 22 observed covariates as dependent variables, nine weighted simple regressions (for models with a continuous covariate as the dependent variable) and 13 weighted simple logistic regressions (for models with a dichotomous covariate as the dependent variable) were performed. Results suggest that the propensity score model satisfactorily removed imbalance on the comprehensive set of covariates, and the propensity selection imbalance was corrected to the best of our knowledge. While there is a possibility of selection bias due to omitted covariates (see Discussion on reasons for relocation below), efforts were made to best correct the selection imbalance using a literature-informed, theory-based set of covariates as recommended by Guo and Fraser (2015).
Lastly, to examine the effects of the “treatment” (volunteering) on health outcomes in 2010, regression models were performed with ATE and ATT weights. Covariates (number of health conditions, total household income, total household wealth, age, education, gender, and race) from the 2010 HRS were included to adjust parameter estimates to test the effect of volunteering on health among relocated older individuals. Weighted logistic regression models were conducted for the outcome variable health status; weighted linear regression models were conducted for the outcome variables ADLs, log-transformed IADLs, and log-transformed depressive symptoms. The health-related measures at baseline were controlled for in all weighted regression models. Interaction terms of volunteering and race, gender, and education were added to the weighted regression models separately.
Results
We found that in the baseline year (2008) on average, respondents were 72 years of age and had reported a high school education. Approximately two thirds were female (64%), almost 6 of the 10 were not married, and most were White (82%). More than half of the respondents rated health status as “good, very good, or excellent” (56%) and had an average of 1.11 ADL, 0.48 IADL limitations, and 2.16 depressive symptoms (Table 1). Approximately 24% of older adults engaged in volunteering after relocation.
Sample Characteristics at Baseline (2008).
Note. SD = standard deviation; ADL = activities of daily living; IADL = instrumental activities of daily living.
Table 2 reports the results of weighted regression models. Controlling for important covariates (age, education, gender, race, household income, household wealth, number of chronic health conditions, and baseline health), older volunteers reported higher levels of self-rated health and lower levels of ADLs and IADLs when compared to the nonvolunteers (control group). Older volunteers were 1.92 times (p = .006; ATE weighted) or 1.85 times (p = .01; ATT weighted) more likely to report higher self-assessed health. Volunteering also significantly reduced the number of ADL (b = −.49, p < .000, ATE weighted; b = −.44, p < .000, ATT weighted) and IADL limitations (b = −.14, p = .001, ATE weighted; b = −.11, p = .003, ATT weighted). There was no statistically significant difference in depressive symptoms between volunteers and nonvolunteers.
Weighted Regression Models.
Note. N = 682. ATE = average treatment effect; ATT = average treatment effect for the treated; DV = dependent variable.
*p < .05. **p < .01. ***p < .001.
Moderating effects of race, gender, and education were further explored by adding interaction terms (volunteering and race, volunteering and gender, and volunteering and education) to the weighted regression models (Table 2). Gender moderated the relationship between volunteering and self-assessed health in the ATT weighted regression. Post hoc analyses reveal that female volunteers reported higher levels of self-assessed health when compared to male volunteers (adjusted probabilities of .67 and .53, respectively). Importantly, race moderated the relationship between volunteering and depressive symptoms: White volunteers reported significantly fewer depressive symptoms when compared to the other three groups. Post hoc analyses reveal that White volunteers reported the lowest levels of depressive symptoms (least square mean = 0.75) compared to all other groups including Whites who did not volunteer (least square mean = 0.88), racial/ethnic minority volunteers (least square mean = 1.02), or racial/ethnic minorities who did not volunteer (least square mean = 0.86). Education did not moderate the relationship between volunteering and any of the dependent health variables.
Discussion
The findings from this longitudinal study support the general premise that formal volunteering protects health in the context of those older adults relocating from their homes regardless of geographical distance. The first hypothesis stated that formal volunteering will buffer older adults’ physical and mental health after relocation, and it is partially supported. Volunteering requires cognitive stimulation, social engagement, and physical exertion, which may explain why volunteers in this study reported higher levels of self-rated health and reduced ADL and IADL limitations than nonvolunteers. Our findings that volunteering after relocation improved self-rated health and reduced ADL and IADL limitations are in line with the theoretical perspectives of the social model for health promotion (Fried et al., 2004), the institutional perspective on productive aging (Morrow-Howell, Hinterlong, & Sherraden, 2001), and role theory/social integration perspective (Baker, Cahalin, Gerst, & Burr, 2005).
The second hypothesis stated that when older people relocate, non-Whites, men, and people with low levels of education will experience greater health benefits from volunteering when compared to their counterparts. This hypothesis was not supported. In terms of gender, female volunteers reported the highest levels of self-rated health. In terms of race, White volunteers reported the lowest levels of depression. These results can be understood in a broader context of sexism and racism as they inform civic engagement. From a historical perspective, women, mostly well-resourced White women, did not work for pay and were constrained to engaging in nonpaid labor market activities such as volunteering or charity work (Bojar, 1998), while most White men were able to obtain higher educational degrees and enter high-status-paying jobs. Mostly poor women, a majority of which were women of color, along with their male counterparts, were required to work in low-status and low-paid jobs, which constrained their volunteer work. These historical and structural limitations have shaped roles and expectations by gender and race and have likely informed who volunteers today.
Life-course and feminist theory posit that volunteering can be a substitute for paid-work status in retirement and that men and women will place a differential value on volunteering (Bojar, 1998). This speculation is somewhat supported in the post hoc analyses: Perhaps nonpaid roles are not as valued by men. White volunteers may have reported fewer depressive symptoms because the culture of civic engagement and the preponderance of White volunteers may reinforce a social convoy of people similar to them. Interestingly, non-White volunteers reported the highest levels of depression. It could be that racial and ethnic minorities are motivated to volunteer in religious settings, and the psychological benefits of volunteering will take longer time to detect, especially after relocation. Volunteering may be a more effective strategy for some older persons when relocation happens. For example, older women or older Whites may find that volunteering reaps health benefits. However, it is important to note that volunteering as a protective factor depends on the measurement of health that is being examined.
Research from population data supports the general premise that formal volunteering protects mental health (Greenfield & Marks, 2004) and that the relationship is not moderated by race or gender (Morrow-Howell et al., 2003); however, there is evidence to suggest that context matters. Longitudinal research from studies on Experience Corps, an AARP sponsored tutoring program, offers clear evidence that key social determinants of health may moderate the relationship between volunteering and health (e.g., race, income, McBride et al., 2011), supporting the caveat that formal volunteering can improve certain dimensions of health. With regard to respondents in this longitudinal population study, it is likely it will take more time to see mental health outcomes as they resettle and reconstitute the meanings of functional, symbolic, and emotional attachments to home, community, and identity. There may be significant emotional stress associated with relocation such as leaving possessions and residences that supported the maintenance of identity and social roles and the emotional taxation involved in meeting new persons and developing trust with civic professionals. It may be that formal volunteering after relocation may help to improve mental health later, after the 2 years measured, which would reinforce the life-course principles of timing, context, process, and meaning (Bengston & Allen, 1993; Elder, 1985). Qualitative research could be especially helpful in understanding the differential effects of volunteering by gender, race, and ethnicity.
Health professionals and community organizers may consider integrating and disseminating information on civic engagement opportunities for older adults nationally and specifically within their neighborhoods and communities. Councils on aging, senior centers, area agencies on aging, social workers, and other direct practice public health professionals can work with older adults to lay out a plan that takes into consideration health promoting activities, such as volunteering, during the relocation process. Health professionals can review potential health risks and the research in which volunteering can buffer their health during the transition. Similarly, nonprofit agencies such as AARP can disseminate information to their members on the benefits of volunteering, particularly after housing relocation.
This study contributes to the current knowledge base by employing propensity score analyses with nationally representative panel data, which enabled a quasi-experimental design to discern the causal effect of volunteering on health. These analyses helped to reduce biased estimates while controlling for important factors such as age, education, gender, race, income, wealth, and number of chronic health conditions. All of these factors are known to inform volunteering, health, and relocation (Greenfield & Marks, 2004; McNamara & Gonzales, 2011). In addition, this study confirms previous findings on improved health outcomes using cross-sectional samples (Greenfield & Marks, 2004) with longitudinal analyses. Finally, this study integrated housing transitions and productive aging literature to examine formal volunteering as a health buffer in the context of risk. Findings support our hypotheses and uncover differential effects by gender and race/ethnicity.
Limitations
While it was beyond the scope of this study, future research should aim to tease out the heterogeneity at the intersection of relocation, social engagement, and health. It is likely such relationships vary based on individual characteristics (e.g., sex, age, disability status), family structure (e.g., living alone, live with, or near children), social support (e.g., social network and capital), environmental context (e.g., motivated by negative life events, institutionalized settings), and individual agency (e.g., choice; Bradly & Van Willigen, 2010) can be triggers to relocating in later life and health (Silverstein & Zablotsky, 1996). It appears that local and nonlocal relocations as well as planned and reactive moves are different in substance and magnitude (Longino, Bradley, Stoller, & Haas, 2008; Pope & Kang, 2010; Weeks, Keefe, & Macdonald, 2012; Waldron, Gitelson, Kelley, & Regalado, 2005; Perry, Archaumbault, & Sanford, 2018) and warrant additional research as does the notion that the intensity and duration of volunteering influence health (Jenkinson et al., 2013). We were unable to perform separate analyses on local and nonlocal relocations because of the small sample size. The sequence of events (relocation and volunteering) could not be analyzed in this study due to the way formal volunteering was measured. It is also important to determine the long-term physical and mental health benefits of volunteering after relocation. There is likely great heterogeneity within and between racial and ethnic groups. For statistical purposes, we had to combine several conceptually distinct categories together for analyses (e.g., racial and ethnic minority groups, local/nonlocal relocations, and hours of volunteer engagement). Finally, although the propensity score was estimated based on a comprehensive set of covariates selected through a thorough literature review, we acknowledge that selection bias is still possible, and the results should be interpreted accordingly. We deliberated on these issues and made the decision to move forward with analyses knowing that the strengths of utilizing the statistical analyses with longitudinal population-based data and theoretical framework outweighed the limitations. Sequence analysis (MacIndoe & Abbott, 2004) may be a more useful methodological approach to discern the heterogeneity of push/pull factors to relocation and explore gender, race, and ethnicity as moderators.
Conclusion
The findings support the general premise that formal volunteering is a protective health mechanism during life transitions. The findings also support a caveat: In the context of relocation, formal volunteering is more of a buffer to health for female and White volunteers. Social policies and practices guided by principles of inclusion and health equity may help build a civic infrastructure with and for men and racial and ethnic minorities, which may facilitate higher rates of engagement and in turn, may ameliorate health inequities.
Footnotes
Author Contributions
Gonzales, Shen, Perry, and Wang contributed to the conceptualization and planning of the study. E. Gonzales led the conceptualization and methods; wrote the theory, literature review, discussion, implications, limitations and conclusion. Shen and Wang co-led the methods. Wang performed the data analysis, wrote the methods and results sections. Perry contributed to the theory, literature, discussion, and implications sections. All authors contributed to every part of the manuscript.
Acknowledgments
We thank Jasmin Choi, MSW/MPH student at Boston University, for her assistance with collecting the literature. We are also grateful to Wayne State University students, Maxwell Smith and Anita Partyka, for their support in the completion of this article.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was partly supported by the Peter Paul Professorship at Boston University; The Center for Innovation in Social Work and Health at Boston University; and the National Institute on Minority Health and Health Disparities, Loan Repayment Program (PI: Gonzales).