Monday, September 27, 2004

27 September's average Poll on the front page of www.pollingreport.com
(plus Rasmussen)

Bush 48.2 % Kerry 43.6 % Nader still about 2 %
if WYSW(S)HG (0 undecided voters vote for Bush) then Kerry wins . If Bush gets about 13% of the undecided voters, the race is tied.

I should have kept track of the polls on the first page. Today multicandidate likely voter polls on the first page are Time, GWU, AP, Marist, GQRR(Democracy corps) , NBC/WSJ, Tipp. Also CBS Registered voter and two way and Fox likely voter and two way.

Also I should include at least the Rasmussen tracking poll which would make it
Bush 48.16 Kerry 43.92 so we would say that the race is tied if we thought Bush would get about 14% of the undecideds.

three updates only one important: Pollingreport has new polls from Gallup and ABC/WP and retired the Tipp poll so the current list are multicandidate Gallup, ABC/WP Time, GWU, AP, Marist, GQRR(Democracy corps) , NBC/WSJ and Zogby. Two way CBS and Fox. All likely voter polls except CBS and NBC /WSJ. The new polls are good for Bush and the retired poll was good for Kerry so the new polling report average poll gives
Bush 49.1 Kerry 44. For others including Nader around 2% Bush would just barely win even if all undecided voters voted for Kerry.

including Rasmussen makes 12 polls with Bush 49 Kerry 44.2 so we would consider the race tied if we excpected that 8% of undecided voters will vote for Bush.

The change is more than entirely explained by the change in which polling agencies are included, since all of Gallup, Time and ABC/WP show the gap narrowing although, shockingly Gallups RV poll is even more favorable to Bush than the LV poll.

With 12 polls, each with a "MOE" of 3 or 4 %, the standard error due to sampling alone of Bush-Kerry is about 1% so a 95% interval is Bush ahead by 2.8% to Bush ahead by 6.8%. Of course it is obvious that most of the error in polls is not sampling error.

Earlier today, pollingreport described the NBC/WSJ 2 way poll as a poll of registered voters and the multicandidate poll as a poll of likely voters. Looked like a typo but I didn't save a screen shot. Now they make it clear that NBC/WSJ is a poll of registered voters.

I left out Zogby. My list should have been Time, GWU, AP, Marist, GQRR(Democracy corps) , NBC/WSJ, Zogby, Tipp, CBS and Fox.


No comments:

Post a Comment