Essays on Social Influences in Economic Decision Making
- Author(s): Argyle, Daniel Wecker
- Advisor(s): Startz, Richard
- et al.
Understanding social influences is vitally important to understanding human behavior; unfortunately, observing and measuring these effects is extremely difficult. This dissertation focuses on applying and developing techniques to measure social influences on individual decision making. The first chapter develops two techniques for estimating social influences using friendship networks, even when individuals differ on unobservable attributes. The second chapter uses social networks sampled from villages in India to demonstrate strong peer influences on an individual's decision to take a microfinance loan. The final chapter examines changes in juvenile criminal behavior in response to their schools changing to a four-day school week. An abstract for each chapter is provided below.
Chapter 1 Abstract: Existing methods for identifying peer effects in social networks require the assumption that there are no unobserved factors that determine both network links and the outcome of interest. Since this assumption is likely to be violated in most networks, I provide two distinct methods for estimating peer effects in the presence of individual heterogeneity. First, I show that if repeated observations on individuals are available, accounting for time-invariant individual attributes can restore identification. Second, I suggest a semi-parametric Bayesian instrumental variables technique that can be used to estimate peer effects models when panel data is not available or if fixed effects assumptions are undesirable. I demonstrate the properties of both strategies using simulated network data and provide evidence that both are able to estimate peer effects, even when networks exhibit sorting on observable and unobservable attributes. Additionally, I apply both techniques to estimate peer effects in the United States Congress using cosponsorship networks and voting outcomes.
Chapter 2 Abstract: I use friendship networks from villages in rural India to provide a careful examination of peer effects in decisions to take up microfinance. Using social networks resolves the traditional identification problems associated with peer effect estimation and allows each individual to have their own peer group. Since the networks in question are sampled and some friendship information is missing, I use an analytic correction to ensure unbiased estimates as well as a technique known as graphical reconstruction to provide bounds on estimators accounting for possible configurations of the unobserved links in the data. In addition to including a variety of individual covariates, I control for observable characteristics of an individual's peers, examine different ways of specifying network links, and account for the probability of being informed about the program. Across these specifications I find positive and significant peer effects; an additional friend participating in microfinance increases the probability that an individual participates in microfinance by 6%-10%.
Chapter 3 Abstract: There has been a recent trend to implement four-day school weeks, especially in rural areas. We show that changing to a four-day school week influences youth criminal activity in jurisdictions where these policies take effect, especially impacting property crime. Focusing on Colorado, where four-day school weeks are especially prevalent, we show that switching all students in a county to a four-day week leads to 0.75 additional juvenile arrests for larceny per 1,000 residents. Effects on other types of crime are inconclusive; estimated signs on drug and violent crimes are negative, although neither attains statistical significance.