Greetings, all. My name is Sam Wang. I founded the Princeton Election Consortium. The site's Meta-Analysis of State Polls started here in 2004 as a diary during the Bush v. Kerry race. Now, in collaboration with Andrew Ferguson, it's grown from a single page of HTML to a full-fledged site...
...until today, when our server went down. Because we're in Sandy-struck New Jersey, we'll be operating on a temporary basis for a little while - no more than a day or two, I hope.
I wonder if I could please commandeer DailyKos for commenting purposes. I'd be grateful. It's a bit retro - and sentimental for me.
Come see what's happening over at the Princeton Election Consortium.
-Sam Wang, Princeton University
For those who are new to PEC: What we do is convert all recent state polls into a high resolution "snapshot" of state polls, as viewed through the Electoral College. Unlike another DailyKos alumnus, poblano (Nate Silver), there are no adjustments and no corrections to polls. Just some math to optimally extract win probabilities and all 2.3 quadrillion possibilities. This makes the estimate a snapshot that is quite sharp in time.
Today, the Meta-Analysis indicates an outcome of Obama 303 EV, Romney 235 EV.
I calculate a second quantity, the "Meta-margin," which shows how much state polls would have to swing toward Mitt Romney to tie the race. Today, the Meta-Margin is Obama by 2.18%.
This Meta-Margin differs by about 3.0% from national polls. I regard this as the big current mystery. While we bring PEC back online, I am doing analysis to understand the difference. Swing-state polls were dead-on in 2004 and 2008, suggesting either national-polling methods or non-swing state opinion as the culprits.
I'll be blogging over there as usual. Please come over.
By the way, at PEC I welcome non-progressive commenters. I hope that some of them will register on DailyKos so that they can comment, even if temporarily. I am proud of the quality of the discussion threads. Please keep it civil!
Update: I've been asked to explain the Meta-Margin concept. Briefly, the Popular Vote Meta-Margin is the same concept as the two-candidate margin we see in national tracking polls - except as measured through US election mechanisms. It is defined as how much swing would have to happen in state polls to create a perfect toss-up in the Electoral College.
The Meta-Margin is driven mostly by state polls in swing states. There are so many such polls that the accuracy of the Meta-Margin is far superior to national poll-based aggregates. I estimate that at any given moment, the Meta-Margin is accurate to within 0.2 to 0.5% - about a tenth the margin of error of a single poll.
For true geeks, detailed methods are cached here.