Gallup Poll Analysis by Headclunker
Essentially, the poll numbers appear to be biased by the relative number of Democrats and Republicans in the poll.
Obama's numbers are random with respect to the ratio of D/R polled, but McCains correlate too well. I use weekly aggregates because they do not give the party breakdown by day. The next weekly aggregate will come out next Tuesday or Wednesday. For Sept 29 - Oct. 5 they had a D/R of 1.17 (approximately 17% more Democrats than Republicans). I think this week it has been more like 30% more Democrats than Republicans.
Pollsters may not be doing it intentionally, but the validity of their polls is questionable when such a correlation exists.
As I understand it, Gallup randomly generates the digits for a particular phone exchange that is weighted based on expected population. A person with a database of the phone numbers of Obama Supporters could bias the system if they had access to telecommunication systems and knew the source phone numbers of the pollsters. It would be a matter of redirection. I do not know what validity checks of phone numbers that the pollsters are doing.
Gee, I wonder why Obama flipped-flopped on FISA?
Update:
Gallup posted the weekly aggregates for last weeks election poll. It is apparent that they polled around 31% more Democrats than Republicans last week AND the bias in the sampling is determining the poll results.
McCains poll results has been correlating with the bias for over two months. This latest datapoint made it clear that Obama's data, with the exception of two outliers, also correlates. The coefficients of the correlation are close and opposite in sign. The correlations indicate that McCain and Obama would be tied if the bias were limited to 11% more Democrats than Republicans.
Why are there so many more Democrats than Republicans is the Sampling Pool? Are all those Republicans on a Do Not Call List? Or are there strange things happening in the world of telecommunications?