It's been said that the 2004 election would be a challenge for pollsters, and today's polls are a fine demonstration. Put simply, side by side they make little to no sense.
In the national arena, the Associated Press finds Bush up by nine points, 52-41. An Economist poll, conducted on the same days, finds Kerry leading by one point. Both surveys were of registered voters, so previous disputes between pollsters regarding who constitutes a likely voter can't account for the discrepancy.
Some new state polls also lead nowhere, with Wisconsin going from a two-point Kerry lead to a 14-point Bush romp in 24 hours, and Oregon, which went from a 12-point Kerry rout to a one-point Bush lead in the same span. And two Florida polls released today are at odds. Gallup gives Bush a 47-45 lead among registered voters, while Quinnipiac University finds Bush ahead 49-41. At least the hurricanes have left a clearer trend in their chaotic wake: The more they batter Florida, the more popular Gov. Jeb Bush appears to get. According to the Quinnipiac poll, "voters approve 87-9 percent of the way Gov. Jeb Bush has responded to the hurricanes, pushing Gov. Bush's overall approval to 62-30 percent, up from a 45-44 percent split August 12."
Finally, there's a bit of controversy brewing about the gender split in the election. The Washington Post announces that "female support for Kerry slips" today, while the AP finds "Bush builds advantage among men." Both articles convincingly use recent polls to argue that Bush is winning in each category, and essentially debunk the other's suggested key demographic. If Bush is:
A: more popular among men
B: more popular among women
Then (unless there's a gender we're forgetting) he is:
C: generally more popular.
Donkey Rising has a good explanation of why looking at women as a single voting block is dubious. It turns out that marriage is a far stronger determinant of how women vote than their sex -- and one recent survey says Kerry leads by 22 points among single women, while he lags among married women by four.