Strategic Poll Responses

A few new polls have come out in the last day or two, showing President Bush's approval ratings at an all-time low and Congressional approval at a mere 13%. Both of these numbers are about as low as I can imagine them being; in order for a president to fall beneath the 40% level, he has to be abandoned by his own partisans. The new dip in Bush's (and Congress') approval, assuming it's not related to sampling, question wording and order, etc., is perhaps some kind of response to the ISG report. But it seems to me that Bush's approval ratings have been consistently 3-5% lower since the election.

We're inundated with polling data in today's political world, and I think more and more people are actively aware of how pervasive it's become. Now, survey research is statistically sound, in that once you have a large enough random sample of the electorate, you can make certain conditional statements about it within a 95% confidence interval. The degree to which a sample matches the real distribution of opinions in a population, however, is more complicated. The 2000 and 2004 presidential exit polls somehow systematically undersampled Republican identifiers, and in so doing presented an inaccurate view of the population's vote choice.

Looking at the Bush and Congress numbers, I can't help but think something else is at work. I don't have any evidence to back this up, but I've long suspected that high-information partisans provide strategic answers to pollsters in the period leading up to an election. For example, if I'm right, a lot of conservatives who disapprove of Bush have an incentive to tell pollsters they approve of him. They might be hopeful that they can neuter the impact of poll results, which are used to frame the tone of national reporting and the annoying "conventional wisdom" that runs through AP, NYT, WaPost, etc. stories and the TV talking heads. Plus, Democrats use them as a cudgel against Republicans during red-faced political debates, which I imagine gets irritating after a while.

Is there anything to this? I don't know -- it's a bit of a paranoid conspiracy theory. But why would Bush's approval numbers drop by 3-5% overnight, merely because Republicans lost the election? Maybe people simply don't like a loser. Or maybe lots of Republicans who disapprove of the president's performance no longer have an incentive to over-report their approval.

UPDATE: Newsweek has Bush at 32%.