|dc.description.abstract||This dissertation consists of three chapters that study social influence and the diffusion of information in decision making contexts with limited observable outcomes.
Chapter 1 studies social interactions and female genital mutilation (FGM), a traditional procedure of removing the whole or part of the female genitalia for non-medical reasons. Using survey data from Egypt, this paper attempts to identify effects of peer adoption and medicalization on a household's decision to opt for FGM. We find that households are less likely to adopt if their peers adopt less and (in certain areas) if medicalization is more widely used by their peers.
Chapter 2, using a lab experiment, studies how influence of any given agent in a social network is driven by assessments of their reliability by network members based on observations of their past behavior. Agents repeatedly make choices, the optimality of which depends on an unobserved state of the world; they are able to communicate those choices with their social peers; and earn a reward after the last period. We enrich the non-Bayesian DeGroot model by postulating that the extent to which network members are influenced by a peer member depends on the extent of nonconformity, variability and extremeness of their past choices. We find that inferred reliability has an effect as significant as network centrality on social influence; when weighting the views of their peers, individuals are sensitive to their observed behavior, especially for those peers with low centrality.
Chapter 3 analyzes the effects of a large-scale randomized intervention which provided incentivized block grants with the aim of improving twelve health and education outcomes. Communities were incentivized by having grants sizes dependent on performance. Our goal is to refine an earlier intention-to-treat evaluation, by examining the intervention's heterogeneous effect on the different subpopulations of households defined by their participation in health information outreach. We find that incentivized grants have a strong effect on immunization rates of children from households participating in outreach activities: as high as a 14.3% increase for children aged six months or less, compared to a maximum average treatment effect of 3.7%.||en_US