The absentee ballot requests are good data. So is the trend of party registration in states that register by party. So are the “exit polls” (given that we have early voting, an increasing proportion of the polls we have are exit polls). But, so are polls that actually are polls of voting preference.
In 2004, Ohio had a 5 point Republican advantage and, in 2008, an 8 point Democratic advantage. This was a huge swing. Then, from 2008 to the mid-term elections of 2010, another huge swing, 9 points.
Also, looking at Gallup’s tracking of party affiliation among adults (or possibly registered voters), from 2009 to 2011, a similar 9 point swing.
So, there is the potential for a huge swing in the partisan mix from 2008 to 2012. This year, who knows, I’m thinking a 1 point Democratic advantage is conservative (from a pro-Romney view).
Another way to look at this is the demographics of voters as revealed by the post-election Current Population Survey. The big shift in 2008 was due (A) to blacks and (B) to younger adults. Possibly black turnout will remain at its elevated level of 2008, but it is not credible that young adult turnout will. This argues that the demographics of this election will be something between those of 2004 and 2008, which would support the argument that the partisan mix will also be something between.
Now, let’s look at the recent polls in Ohio. Four recent telephone polls (not including Rasmussen) and two recent internet polls involve Democratic partisan advantages ranging from 4 to 9 points. That does not look right. What I suspect is going on is that the pollsters are either imposing the partisan mix of the 2008 election, and/or the demographic mix of the 2008 election onto the responses, by weighting the responses.
Now I’ll look at Raz. He imposes a moving average partisan mix. This is fine for the country (although it temporarily stifles a shift in voter preferences that coincides with a shift in party affiliation, which, I believe, explains the current cleavage between Rasmussen’s and Gallup’s nationwide numbers).
What Raz does at the state level is adjust the state spread by the changes in the national spread. But the national spread has been changing slow relative to the changes in the spread in Ohio during the past four years. So, I think Raz’ methodology results in him being off by a couple points in Ohio.
When I adjust the polls to reflect a 1 point Democratic advantage in Ohio, I get Romney up by 2 points. This still makes the state a toss-up. But, there is reason to be optimistic.
One final point. The foray into Pennsylvania is not merely a feint. At this point, it might not be possible for the other side to gear up a GOTV effort. If Romney is within 2 points in this state, and if he is flush with cash and Obama is strapped, Romney can steal this state (and a couple others) during the week prior to the election.
You are really onto something here, and I saw an article BUT CAN'T FIND IT NOW, SO HELP IF YOU CAN!! :), that was something like "Obama wins OH if 400,000 white people don't vote." The gist was that all the polls are not JUST oversampling Ds, but heavily oversampling blacks and hispanics. They are likewise undersampling whites, but especially white Rs.
I've Googled, looked on FR, Big Government but can't find it. If anyone can help, I'd appreciate it.