This dissertation features research that contributes to understanding the role of social interactions and social capital in the economy. Social capital (e.g. social norms and networks that facilitate coordination and cooperation in society) has been shown to correlate with desirable socioeconomic outcomes. The three chapters of this dissertation present econometric methods and empirical analyses aimed at evaluating the potential of policy interventions that enhance or leverage the stock of social capital in addressing certain market failures when they emerge.
Chapter 1 studies the relationship between neighborhood design and social capital using a household survey from California. It offers two contributions: (i) an objective measure of social capital--carpooling--that has not been used previously in this context and (ii) a precise definition of the neighborhood using geocoded data. Living on a cul-de-sac (a special case of neighborhood design) is found to be associated with a higher probability of carpooling to school, suggesting that planners can enhance social capital by favoring neighborhood designs that foster social interactions.
Chapter 2 presents a spatiotemporal model for panel count data that preserves the discrete nature of the data, incorporates different forms of dynamics and heterogeneity, accounts for spatial correlation, and is estimated via an efficient Bayesian MCMC algorithm. An empirical application of solar panel installations in zip codes reveals the presence of spatial correlation that would lead to over-confidence in the estimates if ignored. It also raises a question that is overlooked in the literature about the proper specification of the dynamics.
Chapter 3 is motivated by the well-established result that a market failure in the form of imperfect information causes consumers to under-invest in energy-efficient technologies. It contributes to the literature on peer effects in technology adoption and, hence, whether consumers' social networks can be leveraged by policy makers to fill the information gap and accelerate adoption. Estimating a spatial transition model using individual-level data on solar panel adoption in California and Bayesian MCMC methods, the analysis finds a positive but not statistically important peer effect. The results, though, reveal that failure to control for spatially correlated unobservables leads to biased estimates.
