Prediction markets - US election

As the US election day draws closer, I am fascinated by prediction tools that are used and how they are evolving - in both substance and presentation. While richer interfaces are available and therefore providing us a wide array of fine multimedia gadgets, I am intrigued by the fact that prediction markets are increasingly present in mainstream media.

The Yahoo! dashboard, allowing users to see state-by-state results from both polls and prediction market, and providing the chance to build a personal simulation of the way votes are distributed and, what's more, the new President elected.

Despite being a prediction markets enthusiast, I believe that this time (provided that data on the Yahoo! dashboard is reliable, and I have no reason for doubting it) they will prove an even worse disappointment than early polls. While polls predict a clear victory for Obama, markets predict an unprecedented landslide (364-174).

There is, on the other hand, a series of factors that suggest a different story:

• the Bradley effect (for which I cannot find a reasonable explanation), causing a discrepancy between polls and outcomes that harms black candidates;
• the Shy Tory Factor (a very well known phenomenon in Italy);
• the fact that a +15% result in a single state provides the same amount of representatives than a 0.1% victory, making a victory possible even with less than 50% of overall votes (as in 2000, with Bush at 47.9% and Gore at 48.4%).

So. I do not know any better than any of the curious characters that have emerged during this campaign (Joe the Plumber, the Hockey Mom or the archetypal Joe Sixpack), and my prediction is just about the substantial failure of prediction tools.

...and that's completely unrelated with what I hope.