I wasn't saying that before, but I am now if Brainster's source is accurate. If all the survey companies are switching over to an in-house stable of pollees, that would go some way toward explaining the behavior that doesn't mesh with the rest of my political expectations, like how nothing seems to move the needle anymore. They can't help but preselect their people because they're the only people there to select.
When you ask follow up questions of a sample of people, you don't get another independent sample, you have a longitudinal study of a single cohort. A thousand likely voters sampled from a group of a thousand likely voters are the same people, who know they're being sampled and from the description are being actively encouraged to see themselves as representatives of their demographics.
That data can be still valuable, but it's no longer applicable to the same treatment as data from independent samples. You can't toss it into the same giant polling aggregator without skewing the results. I'm sure pollsters know this. It sounds to me like they can't think of any better solution, so they're just rolling with it. I'd say they're wrong. They'd say I'm wrong. Guess we'll find out in November.
