I don't know...I think the latest round of Mason-Dixons reflect a closer race than that.
Compare the latest M-D results with the 2000 election:
State 2000 M-D Diff
FL TIED GOP +3 GOP +3
CO GOP +8 GOP +6 Dem +2
MO GOP +3 GOP +5 GOP +2
NV GOP +4 GOP +10 GOP +6
NH GOP +1 GOP +3 GOP +2
NC GOP +13 GOP +8 Dem +5
OH GOP +4 GOP +1 Dem +3
WV GOP +6 GOP +5 Dem +1
IA TIED GOP +6 GOP +6
OR TIED Dem +1 Dem +1
PA Dem +4 Dem +1 GOP +3
WI TIED TIED TIED
Six show a gain for Republicans, five a gain for Democrats.
The mean is a gain for Bush of less than 0.5%.
So I don't see at all how these show a Bush lead of 3-4% at all. It's more likely that they show a popular vote tied, seeing as how Bush lost the popular vote by about 0.5% in 2000.
You are not taking into account population size.
GOP +3 in Florida means more than the Dem +3 in Ohio
You're right that this is very inexact. Averaging GOP +3 in Florida and Dem +3 in Ohio may not result in a tie, but it certainly doesn't result in GOP +3 or 4 as AuH20 was claiming. It's much closer to even than it is to such a wide margin.