Posted this again

(Seems to have disappeared the first time)

According to major American polling organizations Obama is so far ahead of John McCain that some pundits are predicting a landslide – and I am too: but for McCain/Palin.

There’s an emotional bias to this: I simply don’t believe Americans are stupid enough to vote Obama into office – but then I couldn’t believe anyone would vote for Clinton over the first President Bush either.

Luckily, there’s a more rational basis too: specifically, I think there’s some evidence that there’s something fundamentally wrong with polling these days – and it’s possible to guess that the underlying cause is a difference in willingness to answer polls between the two sides.

The evidence is based on human behavior, specifically on a consequence of Leon Festinger’s theory of cognitive dissonance. That theory predicts that people will seek out information confirming already held views while actively avoiding contrary information.

Consider, with this in mind, the complete contradiction between the polling numbers reported with respect to attitudes toward Sarah Palin and the actual turnout and enthusiasm you see everywhere she turns up. The polls are uniformly negative but the reality is that Republican organizers can’t find stadiums big enough to hold the crowds she draws.

Similarly polls done for movie studios uniformly show the strongest market for anti-war movies since the end of the sixties – but the actual movies flub horribly at the box office.

The dissonance reducing behavior predicted by Festinger’s theory has been demonstrated many times – and not least by the success of sites like the dailykos and hotair.com. But, if the polls are right, movies like Zucker’s An American Carol should be box office failures – and things like Redacted blockbusters – while John McCain and Sarah Palin should draw only a handful of depressed supporters to major market events.

Since the theory suggests a disconnect between polling results and what we’re seeing people spend time and money on, we have two choices: either the theory is wrong, or the polls are wrong.

A lot of people have put a lot of effort into debunking Festinger’s theory – but, in general, its predictions seem to hold up.

When we look at the polls, however, the field is both much more varied and much more questionable – it’s easy, for example, for a very small group of committed people to mislead panel surveys by working on a few panel members, or for a slightly larger group to mislead pollsters relying on random samples by giving misleading answers on the questions used to compute sample weights.

And what that tells me, is that the polls are wrong.

Paul Murphy on October 7, 2008 at 3:45 PM