Yesterday evening, I attended a lecture at the Stanford Graduate School of Business by associate professor Justin Wolfers on prediction markets, and the relationships between markets and politics. Wolfers started the lecture by showing data from the Iowa Electronic Markets, a real futures market hosted at the University of Iowa, on securities trading related to the outcome of the last and current presidential elections. He contrasted the data against that of the Gallup polls, and concluded that the market (even though much smaller than a traditional financial market) was more accurate in predicting the outcome of the 2000 presidential race than the poll, and less inaccurate than the poll as time before the election increased.
At this point of the lecture, I'm totally sold on the efficiency of these prediction markets. This is terribly old news to those who have been evangelizing the power of markets like these, but I'm having a rush of ideas of how they can be applied for the first time. My first idea relates to large scale software development projects. One of the chronic problems with every software project (even as small as 3 or 4 people) is accurately evaluating status and completion time. Would a market where programmers and managers could trade option contracts based on functional test acceptance time and quality more effectively aggregate the disperse information than status reports, conference calls and company meetings? Is anyone currently using a market like this to assess software projects?
Wolfers also explores the IEM data against that of real financial markets, oil prices, soft money campaign contributions, etc., making claims that politics directly (and predictably) drive markets, and vise versa. While not as compelling as the simple prediction market claims, he makes a very convincing case and presentation. Too bad Stanford will be losing him to Wharton next year.