Reply #61: A negative WPE

[Febble]

is one in which the poll gave Kerry a bigger voteshare than the count. It got called "redshift" (don't know who coined that one, but it's useful - might have been Freeman) because the count was "redshifted" relative to expectations based on the poll.

OTOH can probably explain GEO better, but essentially it's the estimate made from the exit poll data alone (after adjustment for age/race/sex proportions of respondents vs non-respondents) and, I think some factors to do with the geographic selection of precincts.

According the the E-M report, the Geo estimate for Bush's voteshare in Alaska was 57.4%, and the final margin reported is -25.6, giving his vote share as 62.8%. But that ignores third party, and CNN

http://www.cnn.com/ELECTION/2004/pages/results/states/AK/P/00/

gives his share of the total vote as 61%.

But, ballpark, the exit poll underestimated Bush's share of the vote by quite a bit. Translated into votes, 57.4 % of the total vote (according to CNN) in Alaska is 178,978, CNN gives Bush 190,889. So, if you believe the exit polls ;) someone stole 11,911 votes, if I've done that right. Or switched 5,956


And the exit polls weren't likely to be that far off due to sampling error. But they could have been off for other reasons, and frankly there weren't rich pickings in Alaska. I wouldn't have thought it would have been worth the risk.

Better to steal votes in Connecticut.