March 09, 2004

How Predictive are Idea Markets?

Not very, if the Iowa Electronic Markets following the Iowa "primary" for the US election are to be believed.

Barry L. Ritholtz's Big Picture blog observed recently that the Iowa market didn't converge on the right answer until the last 24 hours. Hence these markets have no real predictive power ... goes the conclusion - they are only act as an opinion poll of another flavour.

To some extent this is true - but the presence of money makes for a much more immediate feedback loop than is otherwise present in representative democratic voting. Few people can quantify the direct benefits to their pockets from voting in one or other choice, while everyone can quantify the result of being paid for being correct.

The continual opportunity to measure ones own prediction as against the end result, for money, makes every voter into a profit maker. That's what makes the "market poll" idea special - suddenly, the voters have an incentive to recognise even their own party is on the way out.

And that's what I'd suspect happened - looking at that graph, I wonder if the Dean supporters decided to let their monatery interests rule over their political preferences, and sold out as they got close to the day.

(practical comment: idea markets should be interpreted as contract trading markets in the FC sense, and then they make much more sense.)

Posted by iang at March 9, 2004 05:06 PM | TrackBack