Interpretation
As in previous studies, we investigated the association between volunteering and PA [
9,
10,
30]. However, to the best of our knowledge, this is the first study investigating the association between changes in volunteering status and changes in PA, with special focus on the different effect sizes for starting and stopping volunteering, in combination with a stratification for sex.
The three studies on the association between volunteering and PA included in a systematic review could not conclusively identify an association. One of the studies only included women [
16]. Another investigated PA as mediator between the association between productive engagement, including volunteering, and no longer driving a car [
31]. The third study focused on physical functioning rather than PA [
32]. The large differences in the included study population, design, focus, and details may explain the inconclusive result of the systematic review. In future studies, more focus should be given to the details and the context.
In an RCT that investigated the association between volunteering and PA, predominantly African American older adults were included. The intervention group assisted children at school for at least 15 h per week [
9]. Besides cultural dissimilarities in this context, the major difference between that RCT and our study lies in the comparison. While the RCT performed group comparison, we investigated within-individual changes. Nonetheless, the results point in the same direction, highlighting a positive association between volunteering and PA.
Another difference between the RCT and our study is the broader context of this work. While in the RCT, one specific task of volunteering was investigated, we included any officially volunteering task. However, we did not include details on the time spent in volunteering; but most volunteers spend less than 15 h, which was the minimum in the presented RCT [
12]. These two factors: tasks associated with the volunteering, and the time spent volunteering, could be the reason for the insignificance of our observed findings. Future studies should therefore consider the task of, and time spent, volunteering when evaluating associations.
Furthermore, the underlying organisation and type of volunteering might be the reason for differences between men and women, who are likely to have different preferences in terms of volunteering [
16]. While men are more likely to volunteer in sports and recreational organisations, women prefer to volunteer in religious organisations; volunteering for these different organisations include different levels of PA [
16]. On the other hand, a lack of power may explain the lack of statistically significant differences between both sexes observed in our additional analysis that investigated the interaction between starting volunteering and sex. Another explanation for differences in our sex stratified analyses—but no statistically significant differences between sex—might be that in our sample, women were more active than men. This difference in levels of physical activity does not correspond with the existing literature [
4]. Thus, we assume that women, independent of their volunteering status, already had a high physical activity level, and therefore increasing the amount of physical activity might be particularly challenging for these women.
The difference in effect size and significance, at least in men, between starting and stopping volunteering and its association with PA, might be explained by the fact that volunteers stop this role due to competing time demands. That is, they may replace their volunteering with another duty or task e.g. taking care of grandchildren, which may require the same amount of PA.
In summary, starting to volunteer was associated with an increase in PA in men, while stopping was not significantly associated with a decrease in PA. Thus, in addition to the benefits of volunteering for health status, as reported in previous studies, it also supports enhanced PA in men [
9,
13].
Strengths and limitations
To the best of our knowledge, this is the first study to investigate the association between starting and stopping volunteering and changes in PA. Our finding that starting to volunteer is associated with a larger increase in PA, than the decrease in PA when stopping volunteering, in men has never been reported before. Moreover, our data from a large national representative longitudinal study is a major strength of the study. It allowed us to investigate associations of rarely investigated phenomena, such as starting to volunteer in older adults.
The use of FE regressions, which use within-individual changes, and which eliminates the influence of time-constant (both, observed and unobserved) factors on the estimates, was an additional strength in this study.
A limitation of the study is the use of self-reported information, which is always prone to recall bias, and an overestimation of desired behaviour, such as PA. This effect, however, could be small, as it is likely that recall ability and overestimation of desired behaviour is constant within individuals over time, and thus would not have influenced the findings drawn from within-individual change observations.
Special attention should be paid to our variables of interest: volunteering and PA. Participants were coded as volunteering if they performed any task and any timespan of volunteering within an organisation. This means that there is limited information on the type of volunteering and time spent volunteering, leading to potentially significant heterogeneity in volunteers. However, the aim of this study was to investigate if volunteering in general leads to increases in physical activity. Nonetheless, we would recommend to include information on the task, time spent, and type of volunteering organisation in future studies. Another limitation regarding the change in volunteering status is that we do not know the exact time when the participant started or stopped volunteering during the study period and if this happened at the same time as the change in physical activity occurred. Measuring our dependent variable (PA) in a valid and reliable way is well-recognised as difficult [
33]. However, a systematic review concluded that the IPAQ is suitable for evaluating within individual changes [
34].
Finally, previous research has shown that a small selection bias exists in the DEAS study [
17]. Despite the design of this representative national cohort study, the low response rate might have slightly skewed the inclusion of participants in the present evaluation, with a slightly higher likelihood of participation among men, participants living in rural regions, and among people aged between 40 and 54 years [
17]. However, the DEAS study is disproportionally stratified by age, region and gender [
17]. This is done to oversample groups expected to contribute to panel attrition, so as to ensure the sample to remain representative in the upcoming waves.