http://www.nytimes.com/2014/10/09/upshot/why-georgia-may-be-bluer-than-it-appears.html?rref=upshot&_r=0"Of course, the black share of the electorate might not stay as high in 2014 as the 28.2 percent share in 2010. Black turnout, for instance, might well go up, but white turnout might go up even more. But we can safely dismiss the possibility that the black share of the electorate will crash into the mid-20s — and that’s where several of the most recent polls put it.
None of this should be interpreted to mean that Ms. Nunn is the favorite, nor should it be interpreted to mean that she’s assured to do better than the polls suggest. Changes in voter sentiment, or the possibility that the polls aren’t accurately measuring voter sentiment in the first place, could easily cancel out any benefits that Ms. Nunn might accrue from a more diverse electorate.
Even if Ms. Nunn does win on Nov. 4, she probably won’t receive the 50 percent necessary to avoid a January runoff. I don’t hear very many Democratic pollsters or campaign operatives screaming about how Ms. Nunn is really in the lead, either.
But there isn’t much question that the most recent polls underrepresent the nonwhite share of the Georgia electorate. Even if there’s something else wrong with the polls that cancels it out, it’s another reminder of the mounting methodological challenges facing pollsters in an era when random sampling is no guarantee of a representative sample. Weighting isn’t always easy; neither are likely-voter screens."