Thursday, September 17, 2009

Rasmussen Polls NC

Rasmussen polled North Carolina Tuesday night and found results that in some ways are very similar to our picture of the state and in others are very different.

The similarities:

-They find Obama's approval at 47/52. We recently measured it at 45/51.

-They find voters in the state opposed to Obama on health care, 53/44. That's relatively similar to the 54/40 spread we recently found.

-They found Richard Burr with a ten point lead over Elaine Marshall and a 16 point over Kenneth Lewis. Last week we found margins of 11 and 16 points respectively in those match ups.

The differences:

-They find North Carolinians a whole lot more happy with their elected officials than we do, with 58% viewing Richard Burr favorably and 40% with a positive opinion of Bev Perdue. We recently found their approval numbers at 38% and 26%. I think Civitas and Insider Advantage are the only other companies to poll on this in the last six months and they have found numbers considerably more similar to ours than these Rasmussen ones.

-The one thing I find truly impossible to believe in the poll is that 66% of voters have an opinion one way or the other about Kenneth Lewis. The one time we polled on his favorability we included his party identification and 64% of voters still had no opinion of him. If we hadn't included that I'm sure it would have been up in the 80s.

It's always good to see a wider variety of polling in North Carolina.


Will Matthews said...

Just the Ockham's razor explanation here, but from the looks of it, you guys state that there is a "Not Sure" option in each approve/disapprove question, whereas Rasmussen does not (per the link).

Given that this is IVR, I'm not sure how people know what to press for "Not Sure". In any case, they do not seem to emphasize it being an option, which likely pushes respondents toward stating an opinion.

On a side note, do most IVR polls vary gender of questioner? Could this play a role in desire to respond definitively? You do all of the recording for PPP, correct? Methodologically, do you think random variation in questioner demographic would add truth or noise? It seems to me that one questioner for all respondents likely causes some kind of bias but that it increases the precision of the measurement. Varying questioner demographics might increase accuracy but at the cost of repeatability.

Will Matthews said...

Forgot to specify the explanation of why the Obama numbers would still be similar...

The effect of pushing people toward an opinion whether or not they have one is more visible among less known candidates. This theory then explains why Rasmussen matches your numbers on Obama and the divergence grows progressively larger as you move to less known candidates.

Web Statistics