And who got it right and who didn't?
Most polls I saw showed Bush winning narrowly. Which polls were off outside the margin of error?
But that doesn't really matter - electoral polls are a different kind of poll entirely. They're meant to predict behavior, and therefore, they use all sorts of past statistics to improve their prediction rate. Thus, they adjust their numbers to reflect past voting trends.
But this isn't about behavior - it's about opinion. Just randomly call a couple thousand people and get their opinion on the President's job approval - it doesn't matter what party they belong to, or whether they vote. If the sample is large enough and adequately randomized, then the answer you get is probably close to the truth.
Rasmussen and Battleground
Have a good night