Some useful observations on the gold market in a recent Unqualified Reservations
blog entry:
One trivial example of a liquid, functioning prediction market that does not, and cannot, produce accurate predictions is the gold futures market. The gold futures market is a large, active and efficient market in future gold, but in hindsight it demonstrates little or no predictive power. The cause is not at all obscure: since both gold and dollars can be stored at minimal cost, a variety of arbitrage strategies bind the gold futures market to the much larger dollar-futures market (ie, the market for loans). The price signal is real, but it is conveying much more subtle data than the market's net opinion of whether gold will go up or down.
...
Again, the market for future gold does not function as a prediction market, because it is bound by arbitrage to the market for spot gold and the market for future dollars. At least one of these so-called markets is substantially the product of official intervention.