Reply #2: What are you talking about?

[Kelly Rupert]

It worked fine in 2000. It predicted, "The election is thus predicted to be very close, with a slight edge for the Democrats," with 50.8% of the vote, but "According to the equation it is just strong enough to have the Democrats squeak by, but given the standard error of the equation of 2.1 percentage points, the election is essentially predicted to be a toss up."

A toss-up breaking very slightly Democratic. That would, indeed, reflect the popular vote, since Gore won the popular vote with 50.3%. I don't think a discrepancy of 0.5% is " very well."

http://fairmodel.econ.yale.edu/vote/index.htm

He fucked up 2004, though, claiming Bush would get 57% of the two-party vote, when he actually got 51.6%. However, the model couldn't fully take the effect of the war into consideration, being weighted almost entirely on the economy, from what I read.