Civitas has new numbers out on the North Carolina Senate primary that find 77% of voters are undecided, quite a contrast from the 51% we found undecided when we polled the state a month ago.
There's a pretty simple explanation though- they're polling general election voters on the primary. In my opinion that's a pointless exercise. In 2002, the last time a Democratic Senate contest was the main draw, primary turnout was equal to 45% of general election turnout. In 1998, primary turnout was equal to 40% of general election turnout. That means more than half of the people Civitas is polling on the primary are unlikely to vote in it. That drives down the current level of support for all three candidates while also driving up the undecideds.
Let's assume that the 55% of people Civitas is polling who aren't going to vote in the primary are all undecided. If you take Elaine Marshall's 14%, Kenneth Lewis' 5%, and Cal Cunningham's 4% and divided them by 45%, you get Marshall at 31%, Lewis at 11%, and Cunningham at 9%. Those numbers are almost identical to what we found a week ago.
I think Civitas' polling is generally methodologically sound (question wording is a different issue,) but if they want to release numbers on the primary I think they should poll primary voters. Elaine Marshall has a lead in the high teens right now, and Cal Cunningham has more support than the Civitas poll is showing as well.