A Club for Growth poll shows Hoffman ahead, but the sample size is one that caught my attention.
The poll of 300 likely voters, conducted October 24-25, 2009, shows Conservative Doug Hoffman at 31.3%, Democrat Bill Owens at 27.0%, Republican Dede Scozzafava at 19.7%, and 22% undecided. The poll's margin of error is +/- 5.66%. No information was provided about any of the candidates prior to the ballot question.300 people is far less than what would otherwise be considered an adequate number to fair sample. Other polls from the last week or two have shown a near dead heat. Given the margin of error and the small sample size, I think it's fair to say that the poll is unreliable as an indicator of voter sentiment in the district.
This is the third poll done for the Club for Growth in the NY-23 special election, and Doug Hoffman is the only candidate to show an increase in his support levels in each successive poll. The momentum in the race is clearly with Hoffman.
Meanwhile, another unreliable poll is now being proffered to indicate that Jon Corzine has rebounded and taken a significant lead over Republican challenger Chris Christie and longshot independent Chris Daggett. It was taken by Suffolk University (cross tabs are here) and among the oddities is that it included all politicians on the ballot for governor, not just the top three candidates. It also has unbelievably high undecided figures, which would swamp any possible advantage proffered to Corzine by the poll. Only 400 people were counted in this particular poll.
On question 14 (page 17 of cross tabs), which asks those who were undecided, (which comprised nearly 25% of those polled) "if you are standing in a voting booth right now, who would you vote for" out of the 400 polled, only 104 actually answered this particular question, and of that total, 26 said Corzine, 14 for Christie, and 2 for Daggett. 56 said they were undecided.
That's a far higher number of undecideds at this stage than other races. It definitely seems that this is related to the poll's methodologies.
By comparison, Quinnipiac polled more than three times that amount - 1,200 statewide - to get a reasonable snapshot of views in their last major poll released mid-October.
I'm not alone in questioning the Suffolk poll. Other polls show that the number of undecided voters is half that what was seen in the Suffolk poll, and it seems to indicate the methodologies used by Suffolk appear to have affected the poll results. Moreover, no poll taken in the past couple of months have shown a Corzine lead of greater than three points - and those have been few and far between. It has only been in the last couple of weeks that Corzine has closed the gap to a dead heat, and that's largely due to a massive attack ad campaign against Christie.
What's troubling is that the poll indicates that a clear majority think that Corzine will win, regardless of their preference.
Small sample size relates directly to margin of error. The smaller the sample size, the greater the margin of error. That's why many reputable pollsters run samples of at least 1,000 people. Smaller than that, and results will fall within the margin of error.