NC-Sen: Civitas Poll Shows Decline for Both Parties, but Is Wording to Blame?

[ccharles000]

McLaughlin and Associates for the Civitas Institute (R) (4/21-23, likely voters, March 2009 in parens):

Roy Cooper (D): 34 (41)
Richard Burr (R-inc): 35 (38)
Undecided: 32 (21)
(MoE: ±4%)
So down at the skunkworks at SSP Labs, the boys and I spent some time trying to figure out why on earth both Cooper and Burr would show drops like this from the last survey - especially Coop. Our initial diagnostics came up empty, but after scoping out the innards, I think we've come up with something.

You might have noticed that the very first line of this post refers to a polling outfit called "McLaughlin and Associates," a Republican firm from Alexandria, Virginia. I'm not familiar with them, and in fact, I've never seen Civitas identify McLaughlin as their pollster.

More importantly, McLaughlin changed the wording of the horserace question from what Civitas had been using in the past. In other words, we don't have a true trendline. The old question:


"If the election for United States Senate were held today and the candidates were republican Richard Burr and democrat Roy Cooper, for which would you be voting? If not sure/unsure/refused... which candidates are you leaning toward: republican Richard Burr or democrat Roy Cooper?"
And the new question:


If the election were held today, which one of the following best describes how you are likely to vote in the election for United States Senate between Richard Burr, the republican candidate, and Roy Cooper, the democratic candidate?
The choices were "definitely X," "probably," "lean," and perhaps "undecided." The key change is that the old poll actually pushed leaners - it doesn't look like the new one did. And the less-than-traditional phrasing of the new question seems a bit wonky to me. I've never really seen a straight horserace question tested in quite that way before. All in all, an unusual set of choices by Civitas, but it looks like our dedicated gang of greasemonkeys has at least solved this mystery.

http://www.swingstateproject.com/diary/4886/ncsen-civitas-poll-shows-decline-for-both-parties-but-is-wording-to-blame