Rasmussen is an automated-calling operation similar to SurveyUSA. In the 2000 cycle, Rasmussen kicked ass in the 2000 election cycle but blew the general in a big way -- predicting an improbable 8-point Bush victory over Gore. Rasmussen's reputation took a serious hit.
While it's always healthy to retain skepticism about any poll's results, it's silly to completely dismiss them. Two reasons: 1) Rasmussen has had four years to work on its methodology. It may be a better poll these days, just as it could be an even worse poll. But point is, it has likely changed.
And 2) this is the only presidential tracking poll this far out. And given the sheer quantity of polling (every day for eight months) we're going to have plenty of trend lines to gauge. So while the numbers may be off, the trends will be critical to understanding who has and doesn't have momentum.
One more thing -- the sheer quantity of national polling means that Rasmussen will be able to release state-specific polls with great regularity. And we've got our first polls out of Michigan and one of this election's two biggest swing prizes: Ohio. MoE 4.5%.
Michigan
Kerry 48
Bush 44
Ohio
Kerry 45
Bush 41
Pennsylvania
Kerry 45
Bush 44
That's the third poll showing PA a statistical dead heat, which is a worrying trend. It'll be extremely difficult for us to win the presidency without PA in our column. That state went down to the wire in 2000, and it looks like it will be the source of much heartburn this time around.
Thankfully, Bush is experiencing similar problems in Ohio and Florida. (Rasmussen gives Kerry a 48-45 edge in Florida.)