Abstract:
The TAIPEX, which is one of the existing prediction market platforms in the world, was first set up in early 2004 in Taiwan as an experimental tool to study the voting behavior of the people in Taiwan. After the first successful experimental launch on this platform, it was soon realized that one could in fact study financial market behaviors by using this platform. In subsequent years, that is, from early 2004 till the end of 2010, eight experiments have been carried out on this prediction market platform. These include: three presidential elections (2 from Taiwan and one from the US), one parliamentary election, three city mayor elections and one on bird flu. From the results of these experiments, we have observed many stylized facts that are known to exist in everyday financial markets. For example, Figure 1 shows the probability density of normalized price returns of TAIPEX in the 2004 Taiwan parliamentary election with different time lags equal to 55(red), 148(black), 403(yellow), 1097(green) and 8103(blue) minutes. The figure illustrates two interesting features. The first feature is the heavy tails on the two ends of the curves. One can see this easily when compared to a Gaussian distribution as shown in the figure. Heavy tails of price returns in financial markets are stylized facts that are known to market practitioners for a long time and our experiments also exhibit such a feature. Another feature that we can observe in this figure is the universality of different time lag curves. One can see that the different time lag curves can indeed be represented by a single distribution curve. We also observe that other well known stylized facts also appear in our prediction market.
Coffee and tea will be served 20 minutes prior to the seminar.