Asking someone how they feel about the other candidate in the midst of an intense primary is not a good way of predicting what they'll do in November. And "who would you vote for if the election were held today" general election polls taken in May are completely, utterly, 100% useless as predictors of the general election.

In the spring and summer of 2003, national "who would you vote for today?" polls for the 2004 Democratic primaries all showed Lieberman with a commanding lead over all other candidates. That did not shake my very firm belief that Lieberman had 0% chance of being one of the lead candidates in the actual voting. But it did mislead a lot of people into thinking he had the best chance of becoming the nominee. That's where believing early polls can get you. There *are* ways you can make educated guesses about an election in advance, and those ways *do* involve some of the things polls can bring to light (such as trends in candidates' positive/negatives among different groups of people), but "who would you vote for today?" numbers are completely, totally, fully, utterly useless. Really, I can't say that enough. There is zero value in them whatsoever. ZERO.