An extensive body of research documents the strong influence of social relationships, social support, social integration and social networks on well-being. Nonetheless, conceptual clarity remains elusive and these terms are often used interchangeably, precluding confident conclusions and hindering cross-study comparisons. Guided by social network analysis, the social convoy model and the life course framework, I measure social network structure and composition through the use of typologies. I then examine the influences of social network structure and composition on an array of health indicators, including self-rated health, psychological distress and self-esteem.
This study uses data from the Americans;; Changing Lives Survey, a nationally representative longitudinal panel survey of adults aged 25+ interviewed in 1986, 1989, 1994 and 2001/2002. I use hierarchical cluster analysis to create social network typologies from data on respondent reports of close confidants and develop two typologies, one for social network structure and the other for social network composition. In cross-sectional analyses, I use logistic regression and Poisson regression to examine the associations between these two social network typologies and poor/fair self-rated health, high self-esteem, and counts of depressive symptoms. I also perform two sets of longitudinal analyses to determine the predictive utility of network structure and composition for health. First, I use OLS regression to examine whether the social network typologies predict residual change scores for self-rated health, psychological distress, and self-esteem both 3 and 8 years after the baseline survey. Second, I use autoregressive cross-lagged models within a structural equation framework to disentangle the effects of social causation and social selection on the relationship between social network structure and the three indicators of health mentioned above.
The typologies representing social network structure and composition are strongly related to important social and demographic factors. In addition, there are strong and significant cross-sectional associations between these typologies and indicators of mental health, although their association with self-rated health is weak at best. The typologies are highly predictive of changes in mental health across waves, although again, they are not strongly related to changes in self-rated health. Lastly, this dissertation finds strong support for both social causation and selection processes at work in the relationships between social network structure and self-rated health and psychological distress. Support social selection, but not social causation, was found in regards to self-esteem.
