from: Cody Stumpo
subject: Simulating joint dynamics of InTrade’-s electoral prediction markets
I am writing you as experts on prediction markets to ask what you think of some work I have done. At InTrade, you can bet on the 51 electoral markets individually. So using these 51 time series, I generated a daily return covariance matrix to simulate the joint dynamics of how they might move forward in the days until the election. Assuming a normal copula of daily returns and simulating the 48 (IIRC) days until the election 5000 times, I get a distribution of outcomes. The histogram of Obama electoral votes (which interestingly is bimodal) is the main output. The mean is 259 – and the chance he will win the election (as of yesterday’-s close) was 44.4%. This agreed almost exactly with http://www.fivethirtyeight.com which is the best polls-based approach I know of.
Some more details are on my blog, http://houseofcramps.blogspot.com/ . You are free to use/extend the idea, with credit.