The problem with this midterm polls is that their sample sizes are usually very, very small. If I recall correctly, most of the polls that tracked the Specter/Sestak race had samples of less than 500 people each, which is hardly large enough to be statistically representative.

Dan, do you have any plans to interview these guys, to get a sense of how midterm polling differs from general election polling, in terms of sample sizes, tracking, and resources expended? I’d be curious to see how they differ, if at all.

Thanks for the reporting!