[december12]

The trouble with models like this is that the parameters are selected to fit past data, so a fit is assured. Therefore, an excellent fit doesn't necessarily mean that the model has excellent predictive value. Still, Bush ought to be favored to win.

[rebel_yell2]

Give this man a cigar!!! If the future deviates significantly from past data, forecasts can be way off. Just ask the principals of Long Term Capital Management, who included a couple of Nobel Laureate Economists.