> Is that remotely plausible? How the hell would they know any better
> than meterologists? Some direct line to a wise old farmer who can
> tell from the look and feel of the sunset? I've ordered up the paper,
> but until it arrives - why that six-year period? Did he look at any
> other six-year periods? What was his r2? Etc.
>
> DOug
>
Roll looked at the data from 1975-1981 because while there was price data in the OJ futures market back to the early 1970s, he only had weather data from 1975. (It ended in 1981 because, I assume, he needed to pick an end point so he could write the paper.) In other words, there were no other six-year periods to look at. R2s range between 1-4%, which is minuscule, but that number is deceptive. Roll shows that the NWS forecasts explain about 90% of the variability in temperature. So the R2 is, I think, between 0.1-0.4, since only 10% of the variability was going un-forecast.
I think you're overstating the supposed implied mysticism here. It's not as if the traders don't have access to previous National Weather Service forecasts, after all. If the futures market was a predictor of the next day's weather, it would be as a supplement to the NWS, not a replacement for it. (Which makes sense, since OJ traders are only in a tangential way trying to predict the weather.) Maybe it is implausible. But it seemed to me that the earlier note about never having seen the actual OJ study cited implied that this was a fantasy concocted by market advocates. It wasn't. The data, at least, is real.
James Surowiecki -------------- next part -------------- An HTML attachment was scrubbed... URL: <../attachments/20030801/4413ec52/attachment.htm>