I've pointed out several times here that I am struck by how many more people say they're decided when SurveyUSA polls in North Carolina than when we do.
For instance our last Democratic Senate primary showed Kay Hagan at 19% and Jim Neal at 7%. Theirs showed Hagan at 37% and Neal at 29%.
Republican Governor is another good example. We showed Pat McCrory at 18%, Fred Smith at 16%, Bill Graham at 13%, and Bob Orr at 8%. They had McCrory at 27%, Smith at 18%, Graham at 15%, and Orr at 6%.
Well Pollster.com answered the question for me
SurveyUSA, the pollster with the smallest undecided in South Carolina (1%), typically inserts a pause in their automated script, so that respondents have to wait several seconds before hearing they can "press 9 for undecided."
The effect that pause is having in North Carolina is to get 'leaners,' people who are mostly undecided but might be leaning toward one candidate, to say they support that candidate rather than saying they're undecided. So SurveyUSA's numbers are probably a greater reflection of what would happen in the ballot booth if the election were today, but they're likely also more volatile because 'leaners' are much more likely to change their mind about who to vote for between now and the election than people who have pretty firmly made up their mind.
PPP and Civitas probably give a more solid look at what the state of the races really is today, but SurveyUSA gives more of an indication of how the vote would come down if the election was already here.