Friday, October 8, 2010

GPI update; Ras now says no to leaners

Rasmussen (10/6, likely voters, 9/22 in parentheses):
Mark Dayton (D): 40 (34)
Tom Emmer (R): 38 (36)
Tom Horner (I): 15 (18)
Undecided: 5 (2)
(MoE: ±4%)

A few things change with this update of the GPI, the first of which is something truly radical, standardizing the way I format new poll results. I put a lot of thought into what the most effective way to do it would be and after careful deliberation I decided to crib Swing State Project's poll formatting style, so if it looks familiar you know who to blame.

In this new poll Rasmussen backtracks from one aspect of their last poll.

In the Minnesota governor’s race, Rasmussen Reports has made a decision not to use our traditional leaners model. Normally, that model shows support falling off for a third-party candidate. However, in Minnesota, third-party candidates often defy that trend, and a look at the initial preference data suggests that may be happening this year.

Joe posted about this apparent methodology issue when that poll was released and it looks like Rasmussen has come to the same conclusion. This resulted in another change to the GPI, I'm using the without leaners numbers in the GPI now instead of the with leaners numbers they originally published as the top lines.

Additionally I was using the pollster rating for the StarTribune in their weighting, not for PSRA who they now contract with for polling, that has also been changed. I've also added the averages for undecided voters into the GPI.

The big change though is the switch to election countdown mode, which basically means increasing the decay rate of the age weighting. The GPI is going from a linear decay to an exponential decay which will put even more of an emphasis on the newness of the poll. With that out of the way here is the updated GPI.

Image Hosted by

No comments:

Post a Comment