Thursday, February 28, 2008

Um...no

Yesterday Ryan Teague Beckwith wrote a reasonable summary of the polling landscape in North Carolina.

Somehow GOP gubernatorial candidate Bob Orr extrapolated that to mean:
None of the polls being run actually poll likely primary voters nor release the sample or weighting given.
That's neither what Ryan said, nor is it true. PPP has always polled likely primary voters and been very transparent about how we derive our sample and what we weight for. In case Mr. Orr has not been paying attention, here's that info:

-Our sample model is based on people who voted in the 2004 or 2006 primaries. We may change that to 2004 primary or 2006 primary or 2006 general if our primary ends up mattering in the Presidential race because turnout in key states has been closer to midterm general election patterns than standard primary patterns.
-For Republicans we weight by age and gender. For Democrats we weight by race too but that, um, isn't really necessary for the GOP. Usually too many women and older people respond, so we randomly delete respondents fitting those descriptions so that our poll is as accurate a reflection of the electorate as possible.

Mr. Orr can claim all he wants that the polls are all wrong, but I'm confident enough that he's not going to win the GOP nomination based on our numbers that if he does I'll eat a live cockroach.

2 comments:

Anonymous said...

How do you think SurveyUSA (which was not mentioned in the Under the Dome post presumably because it only has been polling the current NC races for the past few months, but seems like it will poll the races through at least the end of the primary season) rates against the other polling firms Ryan mentioned? He wrote "Basically, if you're looking for a human-operator survey of likely primary election voters from a nonpartisan pollster regularly working in North Carolina, you're not going to find it." It basically meets two of the criteria detailed. It is nonpartisan and seems to have started polling North Carolina voters on a regular basis. I think Civitas and PPP are in the top tier and Elon's poll and SurveyUSA are in the bottom tier because of Elon's questionable methodology and therefore results and SurveyUSA's outlier polls. Civitas and PPP seem to, on average, produce more reasonable results.

Tom Jensen said...

I think SurveyUSA is one of the best polling companies in the country. Some of its North Carolina numbers, especially in the Democratic Senate primary, have seemed a little off to me. I think that's because of the pause they have in offering respondents the choice of saying they're undecided. Their polls may be a better barometer closer to the election than they are this far our. But on the whole I think they are pretty good.

I'm with you for sure that Civitas and PPP are more reasonable than Elon though- that may be the only thing our two organizations agree on!

 
Web Statistics