Crowds are wiser than you think

I'm reading James Surowiecki's The Wisdom of Crowds - so far, it's very good! Steve Sailer posted the following comment on the book:

...he doesn't really explain why pseudo-markets like the Iowa Election Market tend to be more accurate than traditional predictive tools like the Gallup Poll, although the answer is obvious: because they piggyback off traditional sources of information like Gallup. For example, participants in the IEM look at not just the Gallup Poll but a dozen others. Without these pollsters spending large amounts of money to generate information, however, the market players would be pretty clueless. Similarly, if the crowd takes the average of the 11pm weather forecasts of the weatherguys on Channels 2, 4, and 7, they may well beat the best individual forecast, but that doesn't mean the crowd could beat the professional weathermen -- unless the pros first tell them what they think the weather is going to be.

Clearly, there is some truth to this - Hayek said markets are "marvelous" he didn't say "miraculous." But a lot more goes on in information markets than averaging. The green line in the figure below (click to expand) shows the Iowa market prediction for the 1996 Presidential election. The blue line is what actually occured. The spots are various polls. Now notice that from February through August every single poll overpredicted Clinton's victory and every single poll, with but one exception, was above the market prediction. This means that the market prediction could not possibly be an average of the polls.

Surowiecki gives a number of other examples where the wisdom of the crowds cannot be explained by averaging of expert opinion even though averaging is an important reason why crowds can be wise.