That's very interesting. My calculation of the accumualtion of state-by-state polls (mostly using Mason-Dixon, SurveyUSA, Strategic Vision, and ARG) is as follows:
Bush 48.90%
Kerry 45.73%
Turnout 110,000,000
Margin 3,500,000 for Bush
(I projected turnout in 2004 by increasing the total of voters by the growth in population in each state, then multiplying the state polls by that number.)
That implies that about 0.6% of each candidates support level in the national polls is not being caught yet in the state polls. But otherwise, it shows an identical 3.25% race.
I'm cautiously optimistic.
Not my calcs but Jay Cost's. Sounds like you both are on the same statistical wavelength. I agree with both of you.