As for the risibility of the claim that the market outperformed the NWS, Roll was trying to compare the correlation of the NWS' forecast with the actual temperature to the correlation of the change in futures' prices with the actual temperature. The first, as far as I can tell, was pretty good. The second was slightly better. So from 1975 to 1981, if at 3 pm you wanted to know what the Florida weather would be like that night, your best solution would be to take the NWS forecast and adjust it up or down based on what the final price of OJ futures was.
Re: an earlier point, the "predictions" Roll was measuring were of weather conditions roughly six to eighteen hours after market close. So there was no suggestion that markets could make long-term weather predictions. (Of course, some true believer may have made that claim, but I've never seen evidence for it.)
James Surowiecki -------------- next part -------------- An HTML attachment was scrubbed... URL: <../attachments/20030801/4bb1eb6f/attachment.htm>