Reanalysis of Iowa Political Stock Market

From IPSM’s data in 1988, Forsythe and his colleagues proved that even though there were substantial judgment bias, Iowa Political Stock Market worked quite well in the sense that its estimation of presidential candidates’ popular vote share surprisingly close, much better than the forecasts of opinion-polls. They explained that the judgment bias only referred to average traders, but market prices were controlled by marginal traders. In this thesis, I replicated their analyses with 2004 data. However, I did not have any information about traders, so I could only successfully replicate the analyses on market trades, accuracy of market estimation, volatility of opinion polls, and the relationship between opinion polls and market prices. I found that August trades in the 2004 market were not quiet as they were in 1988. The market did a poorer job in estimating the population vote share of 2004 presidential candidates than any of the six major opinion polls. Opinions polls were no longer visually excessively volatile in 2004. By applying their regression-analysis to 2004 data, one of the six major opinion polls had significant effects to market prices, which did not happen in 1988. Acknowledgements: I am particularly grateful to Professor David Lee for his invaluable advising and introducing me to such an unforgettable learning experience, without which this thesis would not have been possible. I would like to thank graduate student Mark Miller for assisting me to run STATA; IGSL librarian Paul, Mark, and Nick for helping me to find opinion polls and election results; my good friend Horace Lit, who stayed with me all the way; and also to the subjects who kindly participated in this study. Lastly, I want to thank for my family’s sincere support.