26 Days: Q&A round 3, ask me anything about politics and the election!

Before I open the floor, here are the latest numbers.

Electoral collage, then and now

Talking Points Memo
08/10: Obama 323, Romney 191
10/10: Obama 270, Romney 206

Real Clear Politics
08/10: Obama 247, Romney 191
10/10: Obama 217, Romney 181

08/10: Obama 281, Romney 191
10/10: Obama 263, Romney 206

08/10: Obama 247, Romney 206
10/10: Obama 237, Romney 191

Nate Silver
08/10: Obama 301.1, Romney 236.9
10/10: Obama 296.7, Romney 241.3

Quick analysis: Obama has lost a lot of projected electoral votes as swing states that Obama lead by safe margins all throughout September begin to tighten after the first presidential debate. But not all projections are equal. Real Clear Politics has almost as many tossup EVs (140) as it has assigned to Mitt Romney (181), and has states like Pennsylvania which the Romney campaign as reportedly pulled out of, listed as a tossup.

That’s ridiculous.

My confidence in these projections go like this, most to least: Nate Silver, TPM, Pollster, CNN, RCP.

If you’re not sure, ignore everything except Nate Silver’s numbers. Most of these other projections are driven by state polls. Silver’s model has about a million different things in it and all of them matter to some degree. That’s why it’s so stable.

National polling

Quick analysis: A mixed day. Obama gained two points in both Gallup polls, lost eight tenths of a point from RAND, and Romney gained a point with Rasmussen. If not for Rasmussen’s bias, Obama would be leading 3/4 national tracking polls today that have released data. I’ll update this story when Ipsos comes out.

Perhaps the takeway is this: Obama has one poll out of the margin of error, the Gallup RV model. Romney isn’t out of the margin of error in any of these polls. And these polls are not directly comparable, despite the way I’ve got them graphed. These are rolling averages comprising different length samples. I’ve updated the graph to show that. The shortest samples are on the left, and longest on the right. So polls on the left will respond to recent events much quicker, but also be much more unstable and prone to noise. Polls on the right will respond much slower, but be far more stable. And most important, registered voter (RV) polls are not directly comparable to likely voter (LV) models. The only RV model in this graph, I think, is by Gallup, and marked as such.

Now, the mic is yours Newsvine and anyone else who happens by.

What questions do you have about the political world today?

[Note: The actual Q&A took place over here.]

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.