Quote Originally Posted by Haloti92 View Post
Yeah, I don't think we were ever on completely different pages, but again, that bold part doesn't tell us anything meaningful. He says he "accounts for" the possibility that the polls are off by looking at how much they have been off in the past. But the distant past is only marginally meaningful at best, imo, and meaningless at worst, imo. The recent past is another matter but then we have the problem of insignificant sample sizes, probably 2000, 2004, 2008 only, and 2000 had the late-DUI effect which Silver just ignores. And 2008 had the unprecedented "perfect" storm for Democrats. Like I said, I don't think the way Silver says he "accounts for" the possibility of the polls being off means much in terms of their actual chances of being off.

I would much prefer some kind of rationale that specifically addresses (in terms of reasoning, i.e. why and how) the discrepancy between the tight national polls, the almost even party-ID when polled specifically, and the less-tight battleground states, and uneven party-IDs being found in these states.

For example the most recent PPP polls for Ohio and Virginia that came out today. Shows sample of D+8 in Ohio; when 2004 was R+5 and 2008 was D+8. And it gives Obama a 5 point lead. Then there is the Virginia poll which shows sample of D+5; when 2004 was R+4 and 2008 was D+6. And gives Obama a 4 point lead. If you split the difference between the two turnouts, Romney is ahead in both polls. If it is 2004 turnout (which due to demographics is fairly unlikely, imo) then Romney wins by several points.

To me, those polls and their results seem extremely unlikely, yet there they are, and there they get put into Silver and Wang's models along with many like them. Could they be dead on? Sure. Do they seem strange to me in terms of a common sense comparison of the situations at election time of 2004, 2008 and now? Yes, to me, they do, even if they are eventually proven to have been accurate.
Right and Nate doesn't weigh all Polls the same, so how he decides which individual poll is debatable. He has an article called "house effects" where he explains and ranks which polls are more biased and then says that he reduces/increases their effect in his calculations so yes, PPP won't have the same influence as say Susquehanna since it is blatantly left leaning. If you guys think mitt is going to eek this one out you may want to put some money up on intrade, the returns on there look crazy. Last I checked Obama was a 64% favorite. Other gambling sites, especially the euro ones have him in the high 70s...