Bacon King and I were going over the
crosstabs and methodology of the poll last night. Usually, SRBI does produce fairly accurate polls and I am not totally sure that the overall margin here is that far off, but there are no fewer than a dozen issues we both spotted in regards to double-weighting, sample sizes and the like.
The most prominent one was the results among white and non-white voters. It currently shows Deal leading 67-21 among whites (believable and probably higher for Deal),
but it also shows Carter leading among non-whites by a paltry 60-21. Ludicrous - no statewide Democrat has received less than 80% of the non-white vote in Georgia since Dems really began losing control in 2002.
If we assume that 60% of white undecideds would break for Deal if the election were held today and Carter performs as well as any other statewide Democrat (receiving 80-85% of the non-white vote in a midterm year), then the real race looks more like:
Whites: 74-24 Deal
Non-whites: 84-14 Carter
Deal 51,
Carter 47