Wisconsin, Colorado, NH, Florida

Those are the four states that make up the current difference between the two electoral college projections on my blog.

Electoral-vote.com has Wisconsin, Colorado, and NH going to Kerry and Florida too close to call.

Electionprojection.com has them all for Bush.

That's a swing of 50 electoral votes.

Comments

[Jason Chervokas]

Fred-

The way I see it, and I follow the current polling via the LA Times app, Kerry needs a sweep of Minnesota, Wisconsin, Michigan...he needs both Oregon and Washington....and then, assuming he loses Ohio and New Mexico (I think the former is certain, the latter less so), he needs both Florida and Pennsylvannia

[Shivering Timbers]

There are a few differences between the sites: First, there's no question that both sites are partisan: electionprojection.com (EP) favors Bush, and electoral-vote.com (EV) favors Kerry. Despite that, I think both authors are trying their best to be objective.

EP uses some sort of adjustment formula, which, while he goes to a lot of effort to explain, seems a little obscure to me. He doesn't just take the most recent head-to-head poll, but mixes together a number of different questions (including job approval, right/wrong track) to come up with a prediction. How he gets to the precise number--and more importantly, why--is a bit opaque.

EV, on the other hand, uses a much simpler technique. He just takes the most recent state poll from each state, and uses it at face value.

You can argue that EP is using a more sophisticated projection, but it isn't necessarily more accurate. "Data enrichment" is always a risk when you're trying to precisely forecast an event which only happens every four years.

I should also point out that EP hasn't been updated in almost a week, while EV has been updated daily. Most of the recent changes on EV have been pro-Kerry, so if EP had been updated since 9/5, it might give the same electoral vote tally as EV.