Today's installment was from one of the sillier events on the campaign trail: Donald Trump's endorsement, which apparently is going to go to Mitt Romney today. Now, in real life no one is going to care one way or another that Donald Trump endorsed a candidate. About the only effect would be a very short blast of publicity, but leading presidential candidates get plenty of that anyway. This isn't something that will be forgotten by November; this is something that will almost certainly have been forgotten by Saturday, when Nevadans caucus. In other words, it's not going to affect vote choice at all. And yet if you ask voters, it turns out that some will tell you that they would be more likely, and a somewhat larger number will tell you that they'll be less likely, to vote for someone with a Trump endorsement. Hey, reporters: don't believe those polls! You can take it as a measure of what respondents think about Trump, if you care about such things, but there's no reason to believe that this kind of self-reporting about vote choice is meaningful at all, and it shouldn't be included in stories about a Trump endorsement as if it was meaningful.
Similarly, there was a ton of coverage about exit polls in Florida that asked about whether ads or debates had influenced vote choice (sorry, no links; most of what I heard was on TV and radio). Hey, reporters: don't believe those polls! People have no real way of knowing how they were influenced in these sorts of things even if they try real hard, and there's no reason to believe that exit poll respondents did any such self-examination. Don't believe me? Ask a room full of people if they vote based on political party. You'll get only a handful of people who believe that they do -- and yet we know very well that party is far and away the biggest factor in partisan elections.
The bottom line here is that polling is a really good tool for reporters to use in many cases, but remember: what polling tells you for sure is only what people will say if they're asked a question by a pollster. We can be confident (if it's a competent pollster) that the answer can be extrapolated out to the full relevant population, but only to the extent that we can be confident that everyone would give similar answers to those questions when asked by pollsters. It's the reporters job to stop and think whether those answers have anything to do with real attitudes or real behavior. They might -- polling about vote choice the day before the election is usually very accurate! But in cases when there's no good reason to think the poll is telling us something meaningful, it's a disservice to readers to report those poll results.