Social networks affect many economic outcomes ranging from which jobs people find to which products they buy. A prominent aspect of network structure is homophily: the tendency of interactions between individuals to be biased towards others who have similar characteristics. Substantial homophily can lead to the fragmentation of societies and communities and thus can significantly affect who interacts with whom, what people learn and what they believe, and how people behave. This project, continuing studies from an earlier grant, includes investigations as to how homophily and other social network properties influence behaviors such as informal borrowing and lending among individuals, word-of-mouth communication, and participation in microfinance programs in the developing world.
The first part of this research concerns how homophily in social networks affects word-of-mouth learning and diffusion. The focus of the research is on how the specifics of the learning process affects the extent to which homophily can retard learning and diffusion. An important question would be whether or not adjustments made by individuals in a society in the way that they process information can help correct for biases that the structure of a social network might introduce. Interestingly, preliminary investigations show that societies in which individuals follow simple rules of thumb for how they process and relay information to and from friends can learn more and learn faster on an aggregate level than societies where individuals are more sophisticated and ``rational?? in their information processing on an individual level.
The second part of this research concerns the relative roles of choice and chance in social network formation and in producing homophily patterns. Do high school students choose to form a disproportionate share of their friendships with others of their own race, or does institutional structure (e.g., assignment of students to classes, participation in various clubs, etc.) lead students to meet others of their own race in disproportionately high numbers? Data analysis based on a model developed under the previous grant suggests that both biases are present and significant; and that the relative roles of choice and chance differ significantly across races. For example, Asians, Blacks, and Hispanics face three to six times higher biases in the rates at which they meet their own types than Whites. This part of the project would involve tracing the sources of these differences across races, and developing a general statistical model of network formation for identifying the roles of choice and chance in combination with other factors in a wider variety of settings.
The third part of the research involves analyzing network data from 75 villages in rural India together with participation by residents of those villages in a recent microfinance loan program. This rich data set includes social networks of eleven different varieties including friendship networks, borrowing networks, advice networks, business networks, and family ties; as well as information about which residents have taken out microfinance loans and what their repayment rates are over time. The network properties influencing microfinance participation will be investigated, including things like caste and subcaste homophily. The data should also provide novel insight into structural contrasts across different sorts of networks. The research will include continued development of and testing of a new theory that suggests how borrowing and favor networks should differ from other social networks.
The broader impacts of this research include the interdisciplinary importance of new tools for analyzing and understanding social networks. The project will also shed new light on policies designed to encourage desegregation and economic development.
