I've always suspected population sampling of this type was bogus, and used for other purposes than actually predicting outcomes, although it never occurred to me to extend my suspicions to political polls. I've never trusted the ratings that are used to judge the popularity of media, like the Neilsen surveys. These are exceedingly important, since they are used to determine TV viewer eyeballs, or radio listener ears, which in turn allows broadcasters to market advertising time.

Broadcast programming and advertising are greatly influenced by program ratings, but a look at the methodology shows it to be based on tiny samples, and cumbersome in its data collection methods, not to mention too easily corrupted if the test homes don't follow procedure precisely. And since there is only one major company doing this, Neilsen, we have no way to test the accuracy or effectiveness of these numbers by comparing them to a competitor. And neither the broadcasters nor the advertisers are likely to want to question the rating system: if it were shown the results were totally meaningless, the whole multi-billion dollar advertising industry might collapse.