Social Interactions and Network Formation -- EmpiricalModeling and Applications

Evidence of social interactions is found everywhere for various kinds of economic outcomes; for example, students' academic performance, use of tobacco and alcohol, and the spread of obesity. Studies reveal that social networks play an important role as channels to disseminate information or facilitate activities. By using the spatial autoregressive (SAR) model to study social interactions, one can utilize the detailed network information to construct a spatial weight matrix and generate the necessary nonlinearity to solve the reflection problem faced by the linear-in-means model. In my research, I consider the issue of endogenous network formation which will potentially cause a problem of selection bias in interaction outcomes. In order to handle this selection bias problem, I propose to model social interactions and network formation under a general framework.In the first chapter, an structural modeling approach for both friendship network and economic outcomes is proposed. To embed the concept of strategic network formation into the model, I motivate the model specification by an economic game in which individuals interact with others and choose both friendship links and economic activities to maximize utilities. By allowing individuals, when making friendship decisions, to respond to economic incentives that stem from friend interactions on certain activities, my structural setup generates two advantages. First, one can evaluate the importance of incentives from certain interactions when individuals choose their friends. Second, the possible friendship selection bias in interaction outcomes will be corrected when the network formation is explicitly modeled. I propose to use an exponential distribution framework to model observed networks.In the second chapter, I consider a different modeling approach to handle the potential friendship selection bias in interaction outcomes. I note that if there are unobserved factors that contribute to both friendship formation and economic outcomes, the spatial weight matrix of the SAR model will correlate with the disturbance term. Thus, a selection bias in outcomes happens. To solve the potential bias, I specify unobserved latent variables in both the network model and the SAR model so that I can capture the correlation between the network formation process and the outcome process.In the last chapter, I apply the proposed modeling approaches to study American high school students' friendship networks and school behavior in the Add Health data. To use the approach in the first chapter, I consider two activity outcomes, students' GPAs and smoking frequencies. From estimation results, I find a significant incentive effect from GPA, but not from smoking, on friendship formation. These results suggest that helping one another on academic learning is important in the formation of friendship networks, while the joy of smoking together is not. However, outcomes of both GPA and smoking frequency are subject to significant positive network interactions. Next, I apply the model of latent variables from the approach discussed in the second chapter to study the effect of social interactions on students' GPAs. The results show that, when enough latent variables are specified in both the network model and the outcome model, the problem of friendship selection bias in outcomes can be resolved.

Name thesis:degree_name

Doctor of Philosophy

Level thesis:degree_level

doctoral

Discipline thesis:degree_discipline

Economics

Grantor dc:publisher

The Ohio State University

Year dc:date

2013

Author dc:creator

• Hsieh, Chih-Sheng

Contributors dc:contributor

• Lee, Lung-Fei

• Economics
• social Interactions
• network formation
• Bayesian estimation
• selection bias

Statement dc:rights

• unrestricted
• This thesis or dissertation is protected by copyright: all rights reserved. It may not be copied or redistributed beyond the terms of applicable copyright laws.

Language dc:language

English

Repository record dc:identifier

http://rave.ohiolink.edu/etdc/view?acc_num=osu1366024830

OAI identifier oai:identifier

oai:etd.ohiolink.edu:osu1366024830