Mark Frauenfelder blogged today about the Iowa Electronic Markets (IEM), a futures market run by the University of Iowa for studying predictions of political elections and other events. I love the idea of these markets and have been following them closely. While the Policy Analysis Market caught some well-deserved criticism last summer, I didn't think it was a crazy idea, but that it was more poorly explained than poorly conceived.

I was following the IEM during the Democratic Presidential primaries, and was disappointed with the results of what I saw. These markets are sometimes claimed to "predict the future," and rightly so in that they can expose upcoming events that are for whatever reason non-obvious by conventional wisdom. It was not obvious to me that Howard Dean would suffer such an upset in the Iowa Caucuses, placing behind Kerry and Edwards when all the polls were showing a different result. After the shoes had all fallen, I went back to the IEM to see if these futures markets -- based, after all, in Iowa -- had been able to predict what many of us had missed.

It seems they did not. Graphs of the market's activity show Dean and Clark (who placed sixth, behind Kerry, Edwards, Dean, Gephardt, and Kucinich) leading right up until the day of the caucuses, January 19th. Edwards, who placed second, wasn't even offered as a separate contract until after the caucuses were closed (he was included in the "rest of field" contract). I've created a detail graph of the end-of-day contract prices (drawn from here) for Dean, Kerry and Clark in the days leading up to the caucuses, and they don't show any clue that Dean was headed for a fall.

No one has claimed predictive markets are perfect, nor that they can forecast everything -- just that they tend to outperform other predictive measures. I'll be interested to see if this idea takes greater hold over time, and gains in effectiveness as it involves more participants.