On Wednesday morning, Democrats were buoyed by the new polling data released by AP (and conducted by their longtime polling partners at GfK). The numbers, to the eyes of Dems, were almost too good to be true: 60% job approval, 53% re-elect, and the best right track/wrong track numbers in years.
It didn't take long for some number-crunchers on the right to insist that, indeed, the numbers were too good to be true. Case in point, the resident horserace analyst over at The National Review, Jim Geraghty:
46 percent [in the poll's sample] identify as Democrat or leaning Democrat, 29 percent identify as Republican or leaning Republican, 4 percent identify as purely independent leaning towards neither party, and 20 percent answered, “I don’t know.”
For contrast, the AP’s immediate preceding poll was 45 percent Democrat, 33 percent Republican; the likely-voter pool in October 2010 was 43 percent Democrat, 48 percent Republican. The poll’s total sample in October 2010 split 43 percent Democrat, 40 percent Republican.
With a poll sample that has a 17-percentage-point margin in favor of the Democrats, is anyone surprised that these results look like a David Axelrod dream?
Does Geraghty have a point? Yes...and no.
Without question, a seventeen-point spread on the question of partisan identification, in either direction, would demonstrate considerably less parity than we typically expect in American political polling. Democrats have historically enjoyed a narrow edge on the issue of party ID (even in the absolutely abysmal 2010 midterms, where the Democratic base stayed home in droves, the party ID of that electorate was evenly split). However, Democratic advantages to the tune of seventeen percent are, to put it charitably, unusual.
The problem is, when analyzing a polling sample for potential tea leaves, party ID might be one of the least reliable ways to do so. Anyone who has ever read the comments section of any political website (including this one), in the midst of a major news event, knows that a pretty sizeable share of the electorate can shed their partisan identification, or reapply it, with astonishing ease. In a time when the President is enjoying a lift, it's not surprising that more voters are willing to identify with his party. Conversely, in a time when there is a wave of discontent in the ranks of the GOP, it is not surprising to see fewer people self-identify with the party.
What's more: AP goes out of its way to press Independents to get off the fence and pick a side. So, Independents are a MUCH smaller proportion in this poll (4%!) than they are in any other poll of recent vintage. Normally, this means elevated numbers for Republicans and Democrats (which split 43-40 in the AP/GfK poll in October 2008).
Why were the Republican numbers not elevated in this poll (indeed, this is fewer Republicans than have been noted in an AP poll in three years)? The answer may very well be the Teabaggers. A startlingly high 20% of respondents said that they "don't know" their partisan affiliation. It's likely that these folks are those self-same teabaggers who espouse all things right-wing, but cannot bring themselves to call themselves Republicans.
By other metrics, that are quite a bit more stable, this is a pretty fair sample. If it tilts left, it is only by a small measure.
Take, for example, ideology. A look at the complete poll (PDF) shows that the sample is 23% liberal and 38% conservative. That is a net edge of 15 points to the conservatives. That actually strikes a pretty decent midpoint between the 2010 exit polls (where the split was 20/42, or Conservative +22) and the 2008 exit polls (where the split was 22/34, or Conservative +12).
Looking at other demographic details: the AP/GfK poll is only slightly less white (71%) than either the 2008 or 2010 exit polls, but there are pretty sizeable differences on income (only 15% identified their income as over $100K, compared to roughly a quarter of voters in 2008 and 2010) and age (this sample skewed quite young).
Given that younger and less wealthy voters historically would be among the Democratic base (although perhaps a bit less so recently), those would tweak the numbers a bit. But the key words here may well be "a bit."
Democrats would probably be doing themselves a disservice to presume, from this single political snapshot, that all is well with the electorate. But Republicans would be equally misguided to look at one demographic detail in the poll, declare it an invalid document, and presume that their own electoral viability is still assured.