More on what Prediction Markets Mean

Michael's post on Charles Manski's paper challenging prediction markets has been widely discussed. Manski's paper is difficult and a number of people wrote asking me for further explanation. Luckily, Daniel Davies has done some of the heavy lifting. Michael also offers further comments here.

My take: Manski shows that the market price is not, for example, the mean subjective belief of the market participants. But who said it was? The argument of prediction market proponents is that the market price is a good, perhaps the best, predictor of the future event. Manski does not challenge this argument. In particular, Manski does not show (or try to show) that there is an alternative way of aggregating individual information that results in better predictions. In this sense, I think the Manski paper is something of a red herring.

I would not claim, however, that information markets cannot be improved. Movements in the price of oil tell us something about trouble in the Middle East. But the oil market was not designed to elicit information about the Middle East. The information in markets is an accidental byproduct of trading. It would be a real surprise if the rules that make for good oil trading are the same rules that make for good prediction of events in the Middle East. Sundering information markets from trading markets, therefore, is a big advance and one that is likely to lead to better market design for information revelation, perhaps with help from papers like Manski's (contra Daniel, who argues that to work well information markets must be tied to trading markets).

Addendum: Victor, a former student of Manski's, at the Dead Parrots Society adds considerable wisdom to the discussion.