Open Access
Xiong, Wenying
Graduate Program:
Information Sciences and Technology
Master of Science
Document Type:
Master Thesis
Date of Defense:
May 12, 2011
Committee Members:
  • Tracy Mullen, Thesis Advisor
  • trading volume
  • uncertainty
  • information aggregation
  • social network
  • prediction markets
  • clustering
A prediction market is a speculative market designed to predict event outcomes. It is an information aggregation method that aggregates participants’ information into a single prediction. Prediction markets have a wide range of real-world applications including predicting political events, sports competitions, current events, entertainment box office results, business sales figures, dates for scientific and technology breakthroughs, and more. The content of this thesis can be divided into two parts. First, I describe the motivation for and history of prediction markets, along with prediction market applications, and mechanisms. Second, I review the limited research that combines prediction markets with social networks and describe my initial research in this area. The intersection of prediction markets and social networks is a promising area, since in most cases participants in prediction markets are actually interacting in some kind of a social network context, but such a context is often neglected in the modeling of prediction markets. In my research, I build on previous research by Chen et al [1], introducing uncertainty to participant’s beliefs, and analyze the effect on the prediction market, analyzed in particular its influence on market prediction accuracy, trading volume, and social network clustering. Two social network theories that motivate and help explain my experimental results are Festinger’s theory[2] and Burt’s theory[3]. Festinger’s theory states that people with similar attitudes and perceptions tend to be attracted to one another, which changes the social network structure. Our experimental design follows Festinger’s theory that people with similar Party registration tend to cluster together. Burt’s theory asserts that overall network structure as well as individual network characteristics influence information flows. My results support this theory, showing that when participants are more clustered, the market has better overall prediction. My preliminary study finds that adding uncertainty causes several effects: First, more uncertainty tends to reduce clustering. Second, and not surprisingly, more uncertainty tends to result in lower accuracy of information aggregation. Third, more uncertainty tends to cause lower trading volumes.