Standard rational expectations models assume away direct communication among speculative traders and resort to psychological traits such as traders' overconfidence in information processing ability to explain the observed enormous trading activity in financial markets. Yet, a growing literature documents that social communication affects individual trading behavior and market trading patterns, and over- and underconfident traders may coexist. The dissertation thus has two general objectives: (1) to analyze the impact of social communication on asset pricing, agents' trading behavior and welfare, and (2) to examine how traders' changing confidences resulted from communication affect asset pricing and trading behavior.
Chapter 2 develops an asset pricing model in which agents communicate information in social networks prior to trading. An agent who is more confident in her private information puts greater weight on her private signal than on signal received through communication when aggregating her information. The results are as follows: Proximity between agents in networks affects correlation of agent demands; Individual agent exploits information and influences price distinctly in different networks; Under certain circumstances social communication is welfare improving for all agents; Irrespective of different network structures, market trading patterns such as market liquidity, trading volume, price volatility and informational efficiency of prices are all higher in communication economy relative to those in economy where agents exploit private signals exclusively; Market trading patterns are strictly decreasing in agents' confidence in private signals. Interestingly, social communication can alternatively explain some intriguing empirical facts such as "gender trading differences" which were attributed to overconfidence.
Chapter 3 develops a multi-period market model to examine the evolution of risk averse agents' confidence degrees in learning their abilities to obtain precise information and the properties of resulting price volatility, trading volume, and expected profits. Agents initially do not know their abilities which are related to the qualities of private signals. They assess abilities from communicating and comparing quality of their own signals with that of others. Agents are assumed to credit (blame) themselves strongly for favorable (unfavorable) outcomes. I demonstrate that under reasonable circumstances excessive trading activity can be associated with underconfidence.