This has been a tough couple of days. I went about a week between posts here because the national polling looked terrible. In April of this year I began a diary with the following: "It is quite possible to drive yourself crazy following National Polling. In particular, the two pollsters that conduct daily tracking are both among the least reliable." You would think I would know better. But I didn't.
On top of this we get the PPP national poll this morning reported on the front page (with no attempt to explain how that poll could be in any way consistent with PPP's state polling). Finally, we get the clear attempt by right wing pollsters with absolutely no track record, or a bad track record, to flood the zone with dubious polling.
It took a call from a good friend to talk me down and get me back to looking at the actual data. And when I did the results surprised me. Then I read this article from Charlie Cook, dated yesterday.
Here is the bottom line: Charlie Cook (who has been skeptical about Obama's chances) rates Ohio as leaning Democratic. And as StellaRay pointedly noted this morning, as long as Obama is ahead in Ohio he is still going to win.
But let's get to the numbers.
The simplest way to see if national polls are right is to compare them against state polls. It is pretty easy to do this. You take the margin in the poll, and then you compare that to the margin in 2008. Then you average the change across all of the state polls, and you get a projection of the state of the race. Let's take the recent PPP Polling:
Are you fucking kidding me? How the hell can PPP say Romney is up 4 nationally and produce these state numbers? There are quite simply NOT consistent. How the fuck can the national number get front paged and this not noted?
So this table shows how you can check if national polls make any sense. And right not they don't. The table uses a 5 day moving average for state and national polling.
Do I think Obama is up four? No - there is at least one outlier in this sample - an Arizona poll showing Obama up 2 (from a pollster I don't trust). But I do believe Obama is ahead, and by more than 2. Because there is no way Obama can be ahead in Ohio, close in Florida, close in Iowa, leading in Nevada and close in Virginia and not be. To paraphrase Bill Clinton, how do I know Obama is ahead:
arithmetic.
You might ask if I have ever seen a gap this large? Yes, I have. Here is the same analysis for 2000:
It is useful to remember that the state average I am computing above is heavily comprised of polls from right leaning pollsters. In fact, 38% of the post debate polls have come from either Rasmussen, Gravis, or We Ask America. Think about that. Gravis has no track record at all, We Ask America is a right leaning group, and Rasmussen was absolutely awful in 2010, and in 2008 understated the Obama margin by 2.92 points. And yet these crappy polls are driving narrative. But even setting that aside, on average THEY STILL SHOW AN OBAMA lead.
Cook has some good advice in this regard:
The problem with state polls is that most are in the extraterrestrial category; robo-polls are often all over the map. Aficionados would be well advised to focus on state-level polling offered by NBC/Wall Street Journal/Marist University; CBS/New York Times/Quinnipiac University; and by ABC, CNN, Fox News, and other brand names that specifically use live interviewers calling voters with landlines as well as the 30 to 40 percent of voters (mostly young people and minorities) who have only cell phones.
I have written about cell phones before and how they lead polling to understate the Democratic margin in virtually every close race in 2010. His advise on Rasmussen and these other fly by night pollsters is well taken.
Here are the state polls post debate, and the projected probability of winning. This based on the work of Chris and 2010. Essentially, what that work found was the empirical odds of winning a race with a certain lead in the polling average. To summarize:
A candidate with a .5% lead wins 56% of the time
A candidate with a 1% lead wins 64% of the time
A candidate with a 2% lead wins 78% of the time
A candidate with a 3% lead wins 83% of the time
A candidate with a 4% lead wins 89% of the time
Go here for more detail:
http://www.openleft.com/...
With these probabilities you can compute Obama's odds of winning the election (just create a spreadsheet with these odds, use a random number generator, and then calculate the result for each state. Run this 20,000 times, and you find that Obama wins 74.7% of the time. EVEN WITH ALL THE CRAPPY POLLS, HE HAS A DECISIVE ADVANTAGE.
Numbers on the number of state polls by pollster are here: