The October issue of the academic journal PS: Political Science & Politics is out. The journal is a publication of the APSA (not a Bush friendly organization).
The 7 models predicting the vote on November 2 find GW winning in all of them.
The predicted vote for Bush is as follows (name of model author followed by percentage):
1) Abramowitz: 53.7%
2) Campbell: 53.8
3) Wlezien & Erikson: 51.7-52.9
4) Holbrook: 54.5
5) Lewis-Beck: 49.9
6) Lockerbie: 57.6
7: Norpoth: 54.7
How have they fared over the years? They have been consistently accurate EXCEPT for the 2000 election. Note: These models are predictive models which mean they are not the same as polling and therefore only seek to predict the percentage on the day of the vote.
My prediction? I have been saying it for 2 months now–my optimistic prediction is Bush gets 40 states.