Social learning in labor markets and in real estate brokerage .

摘要

This dissertation presents three studies of social learning in economic decisions. In the first chapter, I present an analysis of social learning among home sellers as they select real estate agents to list their homes. We are able to identify this social effect by exploiting a natural experiment that arises from the manner in which members of the Mormon Church are assigned to congregations. We argue that the assignment to congregations is essentially random conditional on observed geography. Using real estate transaction data from the largest multiple listing service in Utah, we find that the average home seller is almost twice as likely to choose the same real estate agent as his neighbor if they are assigned to the same congregation. We also present evidence that some of the social learning in this setting is due to word-of-mouth communication and not simply observation, a distinction with important welfare considerations. We show that home sellers respond more to the real estate transaction outcomes of neighbors who reside in the same congregation than to the outcomes of similar neighbors in other congregations.;In the second chapter, I investigate the hypothesis that informal job information networks exist among residential neighbors. Using data that includes information on the timing of residence and job transitions, I investigate whether correlation in workplace and residential location can be interpreted as evidence of social learning in neighborhoods about job opportunities. I find some evidence of a neighborhood peer effect in job choice that persists even in specifications in which reverse-causality can be ruled-out. However, I also find evidence of sorting across micro-neighborhoods on employment-related characteristics.;In the third chapter, I present a simple model in which real estate transactions are more likely to feature high quality listing agents when selling agents are in a position to learn about the quality of listing agents prior to advising buyers. I test this prediction using multiple listing service data and county records data. I employ instrumental variables to identify the relationship between measures of listing agent quality and measures of the network connection between the two agents in the network of past transaction relationships.