Happy Birthday to Dan Rather, 81. From when broadcast network anchors were a big, big deal; and then, from when they weren't.
Some good stuff:
1. John Sides (with Lynn Vavreck) looks at undecided voters again; again, their conclusion is basically that late-deciding voters shouldn't tip the results very much, although this time they suggest undecideds might slightly break for the president.
2. More on insider information from Henry Farrell.
3. Nate Silver, or mathematics, or something like that has been under attack; it produced some useful posts (including the one above); here also is Brendan Nyhan with a nice column making the key point that, not to be insulting or anything, Silver isn't anything special. That is: there's no reason to go after Silver in particular, and not the other people doing basically the same thing. I like Silver a lot, but that's exactly right. Ezra Klein jumps in, too, with an optimistic post. A quick point: Silver's value added, to me, isn't his poll aggregation (which is in my view good, but there are other good ones out there) or his forecast model (it's fine, but again there are other comparable ones). His value added is that he consistently pumps out lengthy, interesting posts on interesting topics. Sometimes polling, sometimes not; sometimes I agree, sometimes I don't -- but he's consistently worth reading. I actually do think there are some people who mistakenly have treated him as a wizard; I don't think his detractors made that up. But so what? He's good anyway.
4. And I didn't do an "elsewhere" post yesterday, but I did one over at Greg's place about ways in which the polls could be wrong. Simon Jackman has a very nice post that actually goes nicely with what I was talking about; he writes about how to model those possible errors, and what including them should do to our confidence in final polling averages.