INDIVIDUAL BEHAVIOR IN SOCIAL NETWORKS

Abstract

This dissertation studies the role of social networks in individual health outcomes focusing on three topics: costs, homophily, and the spread of cardiometabolic disorders within individuals’ social network. This study employs a top-down approach to disentangle these topics by examining and analyzing both novel population level and community level datasets. The novel population level dataset consists of five-year longitudinal data on diabetic individuals and their household medical expenditures and where they lived. It is used to gain deeper insight into individuals’ social networks in order to study how the effects of individual, family, and neighborhood peers influence individuals’ health expenditures and the likelihood of becoming diabetic. The novel community level dataset contains cross-sectional data on individuals’ cardiometabolic disorders and their social network, detailing the type of relationships, their cohabitation arrangement, and self-reported perceived influences from the network. It is used to investigate individuals’ health similarities and how friends and family influence individuals’ cardiometabolic disorders as well as how their relationship characteristics may help explain such influences. The chapters in this study employ several empirical techniques to prove and study the impact of friends and family on individual health outcomes in both population and individual aspects. In Chapter 2, the study takes a broader perspective utilizing the five year longitudinal data of diabetic individuals and households in different locations. The results show that family and neighborhood effects significantly influence an individual medical expenditure and his/her likelihood of becoming diabetics. In Chapter 3, the study shows the importance of individuals’ health in network formation by applying the Exponential Random Graph Model (ERGM) on the novel community survey data to study the homophily of health outcomes. The results provide insights on how obesity and diabetes risks can influence network formation. In Chapter 4, the study focuses on proving the existence and investigating the impact of peer effect on individuals’ cardiometabolic disorders, specifically obesity and diabetes. We use the types of relationships, time spent with peers, and cohabitation arrangement to estimate the peer effect among individuals’ social networks to achieve more precise estimation than previous studies. Moreover, perceived influences on lifestyle habits such as eating, drinking, and exercising are used as alternative measures of the characteristics of relationships. The results of this dissertation generate at least two implications on the relationship between social network and cardiometabolic disorders. Firstly, it proves the existence and impact of friends and family’s effects on individuals’ cardiometabolic disorders and the associated costs on both population and individual levels. The population aspect points to the significant impacts of family and neighborhood on individuals’ health expenditures and the likelihood of becoming diabetic; the individual aspect shows how the peer effects vary across the characteristics of relationships, such as the types of relationships, cohabitation arrangement, and time spent with peers. Secondly, it explains the phenomena of clustering of similarly healthy individuals in their social networks, which leads to meaningful insight of how public health administrators can more effectively intervene in the presence of clusters of unhealthy individuals.