The present study sought to investigate the links between social capital and self-rated health among middle-aged and older adults while exploring the roles of LTPA and SES on its relationship. In line with previous research, the findings demonstrated a positive relationship between social capital and self-rated health. After adjusting for study confounders including demographic characteristics as well as history of chronic disease, greater bonding and bridging social capital were significantly associated with better self-rated health. This suggests that both stronger connections within homogeneous groups and weaker relations between different groups enhance the feeling of health among middle-aged and older adults, which is aligned with argumentation of Putnam [41] stating that bonding social capital may enhance the experience of health through mobilization of reciprocity and solidarity, and on the other side, bridging social capital may have health effects through external benefits secured.
However, after controlling for SES characteristics in the model, the associations of bonding and bridging social capital with self-rated health were attenuated and lost their significances, indicating that SES may play a crucial role on the relationship between bonding social capital and self-rated health. Whitley [42] explained this phenomenon by considering qualitative approaches, which suggest that social capital can play a minor role in protecting some aspects of health, but this pales into relative insignificance when wider SES factors are considered. Following this argumentation, prior studies have suggested that higher levels of social capital may simply be an epiphenomenon of more influential socioeconomic processes [43]. The follow-up test of interaction effects revealed significant moderating role of household income on the associations between bridging social capital and self-rated health. We found that, among individuals with low household income, bonding social capital was positively associated with greater self-rated health; yet, these associations were not apparent among those in middle and high household incomes. Our findings support one of the hypotheses in the literature, called the “buffer effects”, which suggests greater health benefits of social capital among people from disadvantaged populations when compared to their counterparts from non-disadvantaged populations [44]. Specifically, our results are aligned with a recent systematic review which concluded that bonding social capital may play a buffering role on the effects of low SES on health [22], suggesting that promoting social supports or networks within the people sharing common social identifies such as family members or close-friends would positively affect self-rated health among the people with low SES.
Additionally, our study demonstrated that the association of bridging social capital with health varied by the levels of household income, where beneficial effects were more apparent among the people with low and high household incomes, but not for those with middle household income. Interestingly, among the people with middle household income, social capital, both bonding and bridging, were not significantly associated with self-rated health, but LTPA, which was not a significant predictor of health among low and high household income groups, was positively associated with self-rated health. It is unclear as to what underlying mechanism explains such variations by household income levels. But given that the people with middle class tend to have different norms and values than other social classes [45, 46], personal human capital such as investment to health by engagement of PA [47], rather than social capital, may be a stronger influential factor promoting better self-rated health among middle-aged and older adults who are in the middle class. Collectively, our findings may imply that social networks and relationships established between the people from dissimilar social identifies may play a positive role in promoting the self-rated health of individuals who are at the lower and higher end of the SES spectrum. However, for those who are in the middle class, individualized behavioral factors such as LTPA rather than building social connectivity may be more influential to subjective health.
In line with previous study [48], engagement in LTPA that meets the current recommendations of the WHO, was found to be significantly and positively associated with better self-rated health. The findings also demonstrated positive associations of LTPA levels with both bonding and bridging social capital, supporting the previous results [49]. However, there was insufficient evidence observed to claim the mediating role of LTPA on the relationship between social capital and self-rated health, in which the inclusion of LTPA in the model did not significantly alter the association of bridging social capital with self-rated health even after the analysis was stratified by annual house income level (results are not reported but provided in Additional files). Our findings are generally aligned with former results from Boen et al. [50] who reported insignificant mediating effects of PA on the relationship between the indicators of social capital and self-rated health among middle-aged and older adults (≥ 55 years old). However, the present results are in contrast with the results from Mohnen et al. [51] who found PA as the behavioral factor at individual-level mediating the positive associations of neighborhood social capital and self-rated health among adults ≥ 18 years old. As previously noted, social capital is a broad concept and its operational definition is largely inconsistent across the studies [14, 33], which makes a direct comparison of the results with previous studies difficult. In other words, the discrepancies of the results may be attributed to the differences in measurement of social capital including the level of focus (e.g., individual- or, neighborhood-level social capital), dimension (e.g., cognitive/structural social capital, or bonding/bridging social capital), and measurement methods (e.g., surrogate indicators or structured questionnaire). In this study, social capital was operationally defined from the network perspective and measured using a structured questionnaire assessing bonding and bridging social capital at individual-level [25], which have been considered as part of structural social capital [14]. However, some researchers suggest that cognitive social capital, which refers to what people feel regarding social relations such as norms of trust and reciprocity, is more strongly associated with self-rated health [11], implying that the role of PA on the relationship between social capital and health may likely be influenced by how the social capital is operationally defined and measured.
The present study contributes to the body of literature by identifying socio-ecological factors that enhance self-rated health among middle-aged and older adults, helping to better understand the process of successful aging. However, caution is necessary when interpreting our findings due to several methodological limitations. First, as previously discussed, the interpretation of the results concerning social capital should be limited to bonding and bridging social capital as measured by PSCS. Second, the present study hypothesized PA as a potential mediating factor explaining the links between social capital and self-rated health based on the previous literature showing the directional association of social capital with PA (i.e., social capital → PA). However, it is also highly plausible that PA and social capital are reciprocally associated in that greater PA may lead to greater social capital as reported in the previous studies [52, 53]. This implies that PA may not be the complete mediator but a confounding factor influencing the association of social capital with health. Due to the cross-sectional nature of the present study, we could not further elucidate the causal associations not only between social capital and PA but also between other study variables, which we shall recommend being addressed in the future study. Lastly, the survey sample of this study was recruited from the Qualtrics research panel that may not fully represent the target population. Although the Qualtrics panels are shown to provide a reliable and valid survey data, the generalizability of the present findings should be tested in different, more representative sample of middle-aged and older adult populations.