I've read the last couple of pages of this thread, and I also read Nate Silver's most recent article, and I think you're closer to agreeing with him than you realize.
Yeah, we're both hedging our bets!
The essence of both of your arguments are: the battleground state polling could be wrong, and if it is, Romney could win. The difference is he puts the probability of that being the case at about 17% while it seems like you think it's more likely than not.
He's right down the middle with the bias argument. He feels its just as likely that Obama is being under represented in the polling as it is that Romney is. I'm just super skeptical of a turnout that is pretty much like 2008. I need very good evidence to believe that this is going to be the case, such as an actual vote. There is evidence in some polls that it won't.
I re-read his latest post and he doesn't talk about turnout at all. Which I think is a big mistake. Silver is, I think, "too smart" if that makes any sense. He's being too analytical and counting on the data too much and not factoring in other pieces of information that don't have to do with raw numbers.
I'm not sure if you remember this but Kirk Gibson in 1988 was voted MVP of the National League. His stat line wasn't all that great compared to Daryl Strawberry.
Gibby
.290BA/25HR/76 RBI
Strawberry
.269BA/39HR/101 RBI
Statistically Strawberry had a better year. Stats, however, weren't the only thing that made Gibson who he was as a player. There are intangibles that both players had that made Gibson overall a better player for his team that year.
I know Silver is huge in sabermetrics and that is where he first made a name for himself. I wonder what he'd have to say about that MVP race.
Anyway the point is that I think Silver's model is too mechanical and it would pick "Daryl Strawberry" as MVP over Gibson because it ignores some intangibles and relies on the numbers too much.
I don't blame him at all however. It's going to take a massive failure for him to want to change his model and I hope he gets the chance starting November 7th.
What will be interesting if you end up being right is how Silver will respond. Obviously he would want to figure out what went wrong and adjust his model to account for it, but at the same time, the pollsters would be doing the same thing, adjusting their methods. So it would be difficult for him to interpret the numbers in 2014 or 2016
I'd think he'd do it in a way that would rely less on the top line of polling reports and would try to figure out what the likely turnout would be and adjust his polls accordingly.
Facebook
Twitter