So, as Nate Silver said today, the only way Romney wins is if the polls are systematically biased. Thats the only non-crazy argument Romney fans have, and it basically boils down to disagreeing with the weights by race. (I don't care about how many people say they are Democrats, that doesn't matter)
So, I did a bunch of math to see if I could quantify what it means if the weights are wrong. Not many polls have the data needed to determine the weights they are using, but I found a couple national polls and an Ohio poll that gave me everything I needed to play around with it.
National (CBS/NY Times 48-47 Obama, and Pew 47-47 tie)
Both of them are using 46.5% men/53.5% women, which aligns perfectly with the census data (people who told the census that they voted). If we presumed 50/50, it would move things a fraction of a point, but there's no reason to believe that'll happen. For race, CBS is assuming 79% white, 21% non-white. That seems very aggressive to me. Pew has a far more reasonable assumption of 82% white, 18% non-white.
If we presume that turnout will be the same as 2008 (based on census data of people who said they voted), CBS goes to 46-49 Romney, and Pew goes to 46-48 Romney. If we assume 2000 turnout (which is the GOP's desperate hope), Obama's losing by 4 or 5 points in those polls.
I don't think that'll happen, the white vote has been in a steady decline for decades. We also know that this year there's been problems with the national polls being in sharp disagreement with the state polls and the state polls are more accurate, so turning to the Ohio poll from PPP, which had 51-47 Obama:
The gender split seems off, PPP is using 46% men/54% women, but in Ohio men have been in the low to mid 47% range for the last 2 elections. That wont make much of a difference though, because even if we shifted it to 50/50, Obama only loses a point, so we're talking a small fraction of a point here. Anyway, by race if we assume 2008 turnout based on what the census found in 2008 for Ohio, we're at 50-47 Obama. If we use either 2000 or 2004, the poll only shifts to 50-48 Obama.
So, in this critically important swing state, even if the Republicans get their most optimistic presumption that non-white turnout will plummet to 2000 levels (so in Ohio, instead of 14% from 2008 or maybe higher, they are only 11% of the vote), Obama is still winning Ohio. And if he wins Ohio, Romney is probably toast.
<WarMoose> Think about how stupid the average person is. Now realize that half of them are dumber than that.
<Chunda> Why half?