Polls taken before some of the Primaries showed that Obama was either slightly behind or dead even; he then went on to win these Primaries by a factor larger than the percentage of error.

In short, it was a sort of "Reverse-Bradley-Effect."