Replies

Again, I'm not at all questioning their OVERALL accuracy, which to me is irrelevant, but their ability to get the tight races.

That's fine, and that's kind of why I provided the link to their database. So you could check it out yourself.

But I would caution you that polls aren't intended to be absolute predictions of winners and losers. They are measurements, and those measurements have limitations. If a poll predicted Thune would win 51-49, and he actually loses 51-49, the poll was not wrong. It was, in fact, extremely accurate, and well within the margin of error. We cannot expect polls to be more precise than the MOE allows. It's simply not realistic.

Yeah, I know all about standard deviations and all that jazz. That's still no reason to trust polls. My experience has been that even those that supposedly are "accurate" are wrong for MOST of the election cycle, then do some hasty adjustments at the end to be "correct." But that is ignoring their real purpose, which is, I think, to SHAPE public opinion rather than report it. See a book called "Mobocracy" about the abuses of polls.

I have taken several polls, including, in 2004, a Mason-Dixon poll. Even MD, a very accurate poll, asked loaded questions. It had me rank issues in order of how important I thought they were, and "Iraq" and "War on Terror" were separate. Well, when I saw the results of this, predictably, "the economy" or something else was the "top issue," but if you COMBINED "Iraq" and "War on Terror" as I thought more realistic, by far it was the most important issue. But this allowed the Dems to say that their platform on the economy was what "people were interested in."