Friday, October 29, 2010

GPI update; SurveyUSA goes cellphone style, GPI stabilizes

SurveyUSA (10/28, likely voters, 10/14 in parentheses):
Mark Dayton (D): 39 (42)
Tom Emmer (R): 38 (37)
Tom Horner (I): 13 (14)
Undecided: 6 (4)
(MoE: ±4%)

This is the first poll in a while thats shown Mark Dayton below 40% and while his numbers have slipped since the last SUSA poll, that could just be float within the margin of error rather than a decline.

SUSA did something on this poll that they haven't done in Minnesota yet:

Using both RDD landline and cellphone sample from Survey Sampling, SurveyUSA interviewed 900 Minnesota adults 10/24/10 through 10/27/10. Cellphones were hand-dialed.
Unlike the findings in some academic research, and unlike SurveyUSA data in California, CPO respondents in Minnesota are not politically different than respondents interviewed on their home phones.

This SurveyUSA poll also features the much discussed "18-34 years olds love them some GOP," showing Emmer winning the demographic 37-34.

The combination of cellphone only households being split and 18-34 years olds supporting the GOP candidate more so than is happening elsewhere in the nation or in other polls of this state makes me wonder still more about the automated polling, live interviewer split in the data that I discussed yesterday.

Here's that chart again with the addition of this SUSA poll.

Image Hosted by

Not only do automated polls show a larger Emmer share on average, there is absolutely no cross over between the two types of polls. Emmer's largest share in a live poll is 34, his smallest in an automated poll is 36. The two types of polls are telling us a completely different story about the race.

Incidentally Nate Silver just talked about this very subject yesterday:

Most of the automated polling firms have a Republican-leaning house effect. For instance, it’s about 2 points for Rasmussen Reports (our estimate for Rasmussen includes polls conducted by its subsidiary, Pulse Opinion Research) and 4 points for SurveyUSA. Another automated polling firm, Public Policy Polling, has almost zero house effect. But some of the smaller robopoll firms, like Magellan and Merriman River Group, also have a Republican-leaning effect.

On average, the robopoll firms have a 2-point Republican-leaning house effect, whereas the live interviewer polls have a 0.7-point Democratic-leaning house effect. The difference between the two, then, is 2.7 points.

The entire post is worth a read. Obviously we won't know who was right until after the election, but in the meantime, averaging the automated polls with the live ones should give us the most accurate picture of where the race stands and that's the purpose of the GPI.

The addition of this poll didn't change the wAve of the candidates, which is a sign that there are enough polls in the GPI to be confident in it's results.

Image Hosted by

Image Hosted by


"Rating" is from Nate Silver's pollster ratings. "Age" is the number of days since the median date that the poll was in the field. "Weight" is a combination of pollster rating and age of the poll, with age being the dominate factor.

"Ave" is an average of all polls conducted since the primary. "L3Ave" is an average of the last three polls. "wAve" is a weighted average using the "weight" variable.

The graph is of the wAve in the GPI, not specific polls. The dates used for data points are the date the GPI was published.

No comments:

Post a Comment