Love this poll. Only 10% black when the state was 23% in 2012.
This poll does appear to be a bit crap if they are that far off among AA voters, especially in a Southern State.... although it is a B- Pollster with a +1.1% R house effect overall, but still an epic fail on the demographic breakdown of the electorate.
All being said, it does look like the NBC/Marist poll was a bit high for Clinton, and considering that Nate Silver is indicating it does look like a +5-6 Clinton lead nationally, PPPs NC poll is probably closer to the actual current state of the race, and Clinton is likely up about 2-3%.
Do any of these outfits ever actually explain why their demographic samples are so off? I try not to nitpick polls too often, but some of them have ludicrous samples that are blatantly not representative of the actual electorate.
So I have to ask.. why? How can they mess something like this up? It doesn't take a genius to put the numbers together beforehand. I just don't get it. Even a crappy firm should be able to do this without an issue. All I can think of is that they are terribly managed and staffed by idiots who just don't care. If that's not it, then I got nothing.
It's a
random survey. Randomness and the fact that some groups is less likely to participate in polls implies that crosstabs might be skewed. So it's pollster's job to weight accordingly to voter history/pattern/their intention to vote etc.
Gravis indeed weighted, but I prefer when pollster publish their "after-weghted-data" for more transporacy. Their SC crosstabs looks better.
Nate Silver about unskewing/crosstabs.
With that being said, Gravis is B- pollster with R-house effect and should be treated accordingly.