The Poll Sampling Issue
11.01.12 1:39 PM ET
Romney and "Independents"
The one argument from conservatives about the polls that I've been taking kind of seriously is this question of whether Democrats are being oversampled. Conservatives are trying to laugh off yesterday's NYT/Quinnipiac polls as being too skewed toward Democrats in their sample. Here's one representative critique.
It does seem implausible that the turnout in Ohio is going to be D+9. So can we rely on this poll at all?
Well, here's the counter-spin, and I think it makes some sense. First of all, as I've explained, pollsters dont seek out for their respondent pools a certain number of Democrats and Republicans and independents. They seek out a demographically representative sample--white, black, young, old, middle-income, lower-income, upper-income; from this or that region of the state. Then they ask people, do you consider yourself to be a D, an R, or an I.
That's just the best they can do. And so, as I've explained, "party ID" is thus a barometer of how people are feeling about the two parties. And indeed if you look at the history of these things, party ID tracks pretty well with how the parties are perceived. National GOP identification, for example, was at its highest in recent history at the start of the Iraq war. Democratic identification was through the roof at the very beginning of Obama's term. Et cetera.
Now we get to the explanation of how it could be that these polls are "oversampling" Democrats, and how Romney could be leading by double-digits among independents but still behind. Might it be that fewer people are calling themselves Republicans? Steve Kornacki of Salon thinks so:
Actually, there’s a reasonable answer: The term “independent” is a lot more fluid than most people realize. When it comes to their voting behavior, most independents are actually partisans, regularly favoring one party or the other. Very few are authentic swing voters. Plus, when it comes to polling, party label is a matter of self-identification. Pollsters don’t seek out a fixed number of Democrats, Republicans and independents, and people won’t always provide the same answer when asked; that is, a registered Democrat might call himself a Democrat one day and an independent the next.
In the case of the Romney-Obama race, this suggests that a disproportionate number of functionally Republican voters are identifying themselves as independents instead of Republicans – and, perhaps, that a disproportionate share of functionally Democratic voters are calling themselves Democrats, and not independents. This would explain why Romney is doing so well among independents without gaining a comparable edge overall. It would also explain why pollsters seem to be “oversampling” Democrats.
What this means, if this is correct, is that public perceptions of the GOP are generally low. Now I recognize that I'm biased, but that makes sense to me. The GOP Congress has barely double-digit approval ratings. No one outside the LDS church is genuinely excited about Romney. So Republican stock is low, and more Republicans are calling themselves independents.
This would also explain, incidentally, this alleged high enthusiasm among Republicans in polls. If the only people willing to ID themselves as Republican to pollsters are the hard-shellers, then of course enthusiasm is going to register as higher.
I'm not completely sure about all this, but conservatives, I think it's something for you to consider seriously (as if they will--just watch the bile on this thread!). The inescapable fact is that pollsters aren't just inventing these party ID numbers. For two months, they've been coming up D+7, D+9, and so on. There must be a reason, and no, that reason is not pollster bias. We'll find out soon enough.