Discussion
The current study, utilizing a large longitudinal dataset and various panel data methods, shows how the association between volunteering and individual health differs across age groups and groups with differential levels of health. We find evidence that supports all three hypotheses. First, we find that there is a small, but statistically significant health advantage of volunteering, providing support for H1. This is in line with previous empirical studies that find volunteers report better self-rated health than non-volunteers (e.g., Kumar et al.
2012; Piliavin and Siegl
2007). More importantly, we add to the literature by showing that respondents who started volunteering have a health advantage over those who stayed inactive, and that respondents who continue to volunteer have a health advantage to those who quit volunteering.
Second, across analyses, we consistently find that the association between volunteering and health increases with age, supporting H2. The FE regressions show that only for respondents 60 years and older, within-person changes in volunteering are significantly related to changes in self-rated health. In the FD regressions, joining volunteering is significantly associated with an increase in self-rated health only among respondents 70 years and older, and continuing to volunteer is statistically significant among respondents between 40 and 69 years old. These findings provide additional evidence to the limited studies that find stronger benefits associated with volunteering among the older than the younger adults (Musick and Wilson
2003; Tabassum et al.
2016; Van Willigen
2000).
Third, consistent with H3, the results from fixed effects quantile regressions find that the association between volunteering and health decreases in magnitude along the health distribution, with the association being larger for those with poorer health and smaller for those who have better health. Previous research has shown stronger mental health benefits associated with volunteering among less healthy older volunteers (McDonnall
2011; Okun et al.
2011), and a stronger association between volunteering and life satisfaction for volunteers who are less happy to begin with (Binder
2015; Binder and Freitag
2013). Our study is among the first to provide support for a heterogeneous association between volunteering and self-rated health.
Strengths, limitations, and future research
The current mega-analysis delivers a substantial contribution to the literature by providing robust evidence on how the advantages of volunteering differ between groups of volunteers. We show results across a large number of European countries, checking for the robustness of the results. There is quite some variance in the coefficients from different countries. It is likely that findings diverge due to political, economic, social, and cultural differences between countries. Moreover, it is important to take differences in survey methodologies and data quality into account. Our main analyses used panel data methods, accounting for the country and survey fixed effects that may affect the association between volunteering and health. Notably, for the countries with the most precise estimates—Germany, Switzerland, and the UK—the coefficients are not far from each other, and all close to the average estimate. This pattern suggests that with more observations we come closer to the “true” association. Our estimates are consistent with those obtained in meta-analyses on volunteering and health (Anderson et al.
2014; Jenkinson et al.
2013), indicating no or positive associations between volunteering and different health measures like functional abilities, depressive symptoms, and mortality. We show that a mega-analysis is a good alternative for a meta-analysis, because it allows for even more precise estimates and for robust subgroup analyses.
Admittedly, this study has its limitations, which also offer pathways for future research. First, as in other research relying on survey data, we caution that our estimates provide insight only among those who are relatively healthy, at least healthy enough to participate in the survey. Longitudinal survey participants are above average in health (Golomb et al.
2012), reducing the variance in health. Particularly in surveys among older people, the dropout of the panel is health related and not observed in the data we analyze. Such selective panel attrition implies that declines in health are underestimated, and these declines are particularly likely for non-volunteers and those who stopped volunteering. Thus, our estimates are lower bound estimates of changes that are likely to be larger in the population.
Second, a recurring limitation in this body of the literature is causal inference. We stress that causality cannot be inferred from our results. Providing volunteer opportunities in field experiments (Jiang et al.
2021; Pettigrew et al.
2015,
2019) is one of the best ways to assess the advantages of volunteering, but this is not always practically and morally possible. The data we compiled provide two promising opportunities to address causal inference in future research. First, the data cover a long period in which natural experiments occurred, such as natural disasters and changes in the mandatory school age due to compulsory schooling laws, which create exogenous shocks that increase volunteering. Second, the data include measures that may be used as instrumental variables. Functional limitations of the spouse, for example, may be correlated with changes in volunteering but not with their consequences in terms of health (Gupta
2018).
Third, there may be biases or measurement issues. The survey question on volunteering in the past 12 months is sensitive to recall bias. There might also be social desirability bias if either volunteers or non-volunteers tend to report more positively on their health. If these biases are non-random across levels self-rated health, this would affect the precision of the estimates. Furthermore, some surveys have more than 1 year between the waves, so we do not observe changes in volunteering or health in between the observations.
Fourth, the current study provides robust evidence supporting the differential associations between volunteering and health based on comprehensive analyses of a large dataset; however, there are other important questions to be examined. For example, it does not directly test the potential mechanisms producing the associations or distinguish between different types of voluntary activities that may produce differential outcomes (e.g., McMunn et al.
2009; Wahrendorf et al.
2008). In addition, we did not examine the hours volunteered (Morrow-Howell et al.
2003): Because not every survey we harmonized provides information on the frequency of volunteering, we began by focusing on the dichotomous variable of volunteering to obtain the largest sample size possible. Future research may elect to use a smaller subset of the dataset to examine these questions. With its long-time frame and large number of participants, the current dataset is very well suited for extensive analyses on health decline and other trajectories. Moreover, the data include other indicators of well-being and health like physical limitations, specific diseases, mental health, and life satisfaction.
Conclusion
Our mega-analysis of six large longitudinal surveys from Europe provides robust evidence on the association between volunteering and individual health. Volunteering is not a panacea for health. Much of the association is due to self-selection of healthier individuals into volunteering and the selection of less healthy individuals out of volunteering. Nonetheless, taking such selection processes into account, we still find that health improves when Europeans start volunteering and declines when they stop volunteering. Even though the annual advantage is small, it is important because it accumulates over time. The health advantage of volunteering for volunteers is similar in magnitude to the health disadvantage of ageing 1 year. We also find considerable heterogeneity in these changes. The health advantage increases with age, with the most substantial advantages for volunteers aged 60 and over. The health advantage of volunteering among those in worse health is twice as large as the health advantage among the healthiest Europeans.
The findings of this study are relevant for public policy. If volunteering can prevent health decline among older adults and the least healthy, as our results suggest, it could reduce costs of medical treatments. For volunteers and voluntary organizations, this is a stimulus to continue their efforts. Our findings suggest that volunteering is not only a form of social engagement with positive consequences for society as a whole, but may also have positive health benefits for individuals, particularly for the more vulnerable.