In 1996 I started comparing the swing in state polling to national polling. And it was odd: the state polling never shows as big a lead as the national polling did, and in ’96 polling pretty badly missed on the high side.
The basic theory behind the analysis is two fold:
- There are many more state polls than national polls, which makes them less susceptible to being distorted by outliers.
- There are more pollsters – so you are not dependent on about 5 pollsters for the state of the race.
Here is a summary since 2000.
It was significantly better in 2000 and 2012. In 2000 the state polling always showed a close race. In 2004 national polling was better. In 2008 they were both pretty close, but there was a late decline in state polling that the national didn’t see. In 2012 the state polling was SIGNIFICANTLY better.
This shows the moving average as of the morning of November 5th. By the end of the day it would predict an Obama lead of 3.1 – much better than the national polling average did. The diary I wrote this in – of which I am proud – is
here:
So this is the data as of this morning:
For the first time Trump actually leads in the 5 day moving average. There are reasons to be suspicious of this number – it is heavily weighted to Emerson numbers, and those have leaned right. The 10 day contains better news – suggesting a 4 point lead. Almost exactly where Obama was at this point.
One more thing to say about 2012. On the eve of the first debate Obama’s state numbers looked very close to his 2008 numbers. He was, I think, headed for a significantly larger margin than he wound up with.
And then the first debate happened…