<HTML>The OJ futures study was done by the economist Richard Roll. (The article's in the December 1984 issue of the AER, pp. 861-880.) He looked at the Florida futures market between 1975 and 1981. What he found was that if you looked at the change in the price of OJ futures on a given day, it would effectively predict the error in the National Weather Service's forecast for the following evening and early morning. In other words, if OJ futures fell, then the evening's weather was consistently warmer than the Weather Service predicted. If OJ futures rose, then the evening's weather was consistently colder. (Cold weather makes OJ futures more valuable, and warm weather makes them less.) Now, it's possible that this was purely a random correlation. But we are talking about six years of data. And, as Roll points out, weather is likely the single biggest variable affecting the value of OJ futures. So it makes sense that futures traders would pay close attention to, and perhaps have insight into, weather patterns. Roll's study is not a homerun for people who believe in classicially efficient markets. He found that prices moved around far more than you would expect them to, and far more than could be justified by changes in the weather. But OJ futures did, in fact, forecast, however crudely, the weather. There's nothing mystical about this. After all, the futures traders had the Weather Service forecasts to go on, and they could gather whatever other information they thought might be relevant. But this, I think, was precisely the idea behind PAM: a market can often aggregate diverse pieces of information to produce forecasts that are superior to those produced by experts. James Surowiecki</HTML>