Ignore Those Polls!

There's a Pew poll out this morning about endorsements, in which Republican voters are asked how an endorsement by a variety of people would affect their vote. Ignore those polls! We (humans, that is) are really, really bad at determining why we voted for a candidate; we're almost certainly even worse at hypotheticals about how some new piece of information would affect our vote.

That's not just true of endorsements; it's true of any new piece of information. Ask people whether they're more likely to vote for Barack Obama if the economy improves, and their answers are going to be less meaningful than examining past examples of the relationship between the economy and actual voting behavior -- because we're just not very good at predicting these sorts of things. And we have biases. For example, we know that given a two-party election about 80% of us are absolutely locked in by our partisan inclinations, but very large numbers of those locked-in voters will tell you they vote the candidate, and not the party.

There is information in the Pew poll; it can be taken as an indication of how popular the various people they test are with Republicans. But of course it's easier to get at that with a straight favorability question. I'm not convinced that asking it this way adds any value at all.