I usually like SurveyUSA, but this one is rather odd. First, here are the top-line results.
Bob McDonnell: 55%
Creigh Deeds: 40%
Bill Bolling: 54%
Jody Wagner: 42%
Ken Cuccinelli: 53%
Steve Shannon: 42%
Obviously, from a Democratic perspective, these are not the results we want to be seeing right now. However, before we all get too discouraged, we need to look at some of the "internals" of this poll and compare to the internals of SurveyUSA's last Virginia poll (in early June). For instance:
*SurveyUSA now assumes an electorate that is 52% male and 48% female, compared to 52% female and 48% male back in June. The flip is rather odd. I'd also point out that it's almost always the case that there are more female than male voters. So why isn't that the case here? Strange. Also, keep in mind that women consistently tend to vote more Democratic than men (i.e., the "gender gap"), which means that this flip in "internals" skews the poll "red."
*SurveyUSA now has just 17% of voters aged 18-34, compared to 24% of the sample last time around. That's a big shift; the question is "why?" Also, keep in mind that younger voters tend to be more Democratic, so this sample shift would also tend to make the poll skew less Democratic and more Republican.
*On party affiliation, SurveyUSA now assumes 38% Republicans, 32% Democrats and 29% Independents. This compares to the June "internals," which assumed 34% Republicans, 36% Democrats, and 27% Independents. What's up with that? Again, a significant shift in "internals" towards the Republicans that needs to be explained...
*On ideology, SurveyUSA now assumes 42% conservative, 41% moderate and just 13% liberal. This compares to the June assumptions of 33% conservative, 46% moderate, and 15% liberal. Again, a big shift towards the conservatives (+9) and a shift away from moderates (-5) and liberals (-2). Why?
The three possibilities I can think of for the shifts in "internal" assumptions by SurveyUSA since June are: 1) they simply messed up; 2) they intentionally skewed the sample to favor Republicans; or 3) they are picking up a big gap in enthusiasm that skews towards older, more conservative, more Republican voters. I rule out #2, so that leaves #1 (which is possible, but SurveyUSA is usually pretty good) and #3 (which I definitely think is a possibility, but to this extent?).
The bottom line: until these internals are explained, I'm not sure what to make of these poll results. Having said that, I'm not happy with them: a 15-point lead for Bob McDonnell, a 12-point lead for Bill Bolling, and an 11-point lead for Ken Cuccinelli are NOT what I was hoping to see when I heard a new poll was out. Sigh.
UPDATE: I just heard from SurveyUSA news director Ken Alper, who said, "just saw your post and wanted to make sure you see the poll we released today is of likely voters – previously, during our pre-primary polling, we asked the November matchup questions of registered voters."
UPDATE #2: Also a bit strange, SurveyUSA's assumed electorate says it voted 52%-43% for John McCain last year. As we all know, Barack Obama won Virginia 53%-46%.
UPDATE #3: The more I think about this, I believe it's possible that core Democratic voters - liberals, African Americans, Latinos, women - aren't nearly as motivated this year as they were last year. There's also the issue of "federal voters" not voting in non-federal elections. In short, what we could be seeing is that Virginia '08 was a totally different electorate than Virginia '09 will be.
Blue Virginia
