What can reverse causation tell us about demographic differences in the social network and social support determinants of self-rated health in later life?

Few studies of the association between social networks (SN), social support (SS), and self-rated health (SRH) address the role of demography in determining that association. Yet demography defines social-structural context, differentiates family from friend networks, and influences network structures. This study examines the SN-SRH association through cross-cutting analyses of four demographically defined groups (Males, Females, Partnered, Unpartnered) and three networks (Family, Friend, Group). By distinguishing between 'healthy' and 'unhealthy' samples, the underlying causal mechanisms are explored. The positive causal effect of SN on SRH is almost entirely confined to the healthy. In this sample, Friend SN is operational among Females and the Partnered, and Group SN is operational among Males. In the unhealthy sample, reverse causation accounts for all but a weak positive effect of Group SN on the SRH of the Partnered, while worse SRH among Females has the causal effect of greater emotional SS through confiding in friends. Among the Unpartnered, only the effect of SRH on confiding in family members is significant. The findings call into question the validity of studies which assume only positive causation, and underline the importance of demographic differentiation of both population and networks for understanding the SN-SRH association.