Thursday, October 25, 2012

What A Difference A Month Makes

I put together a little chart looking at ten prominent "battleground states" to see what the last month or so has meant for the poll numbers. (Positive numbers = Obama lead; negative numbers = Romney lead).


It certainly looks like there has been a lot of erosion of support for Obama, doesn't it?

(Note: There has been no recent data released for North Carolina as Obama is no longer viewed as being competitive in the state. I used the latest available data, October 18th, for the later North Carolina average. All data gleaned from Real Clear Politics.)

As a point of comparison, here are the current poll numbers compared with the actual results of 2008.


Many of the "predictive models" have been assuming 2008 levels of enthusiasm for Obama in the tailoring of their samples. However, its difficult to see that enthusiasm when you look at the numbers.

2 comments:

Gerard Stelzer said...

I've been noticing the wide swing in different polls on how they sample the D/R/I distribution. Some of the OH polls seem over-optimistic of D turnout, with D+9 in the 2012 polls, whereas the 2008 results were more like D+4. So the Obama/Romney results follow the weighting (since both D/R are strongly aligned to their candidate).

But Rasmussen seems to be undercutting a few of these polls. In OH, I think his last poll showed even, where the other were Obama by 4 pts, or so (e.g. Time id D+9 with Obama leading 49-44). I can't tell what Rasmussen's polls are sampling in terms of D/R/I responses.

NH polls seem to be all over the place, too.

This is much more fun than 2008!

Rich Horton said...

I'd love to muck around in the Rasmussen polls, but you cannot get all the cross tabs without buying a subscription. I'm not THAT into it.

From news reports I've seen, who do have access to the splits, Rasmussen is more in the D+4 range.

I've seen other polling that gets weird results by putting the percentage of Independents in the low teens, and that's just nuts. (And would have the affect of artificially bumping up the candidate with the lager split by as much as 2-3 percentage points.)

I've read that pollsters this time are only getting a 9% response rate, which may go a long way to explain the volatility and the lousy samples they are getting. I've got my own suspicions that the 91% who refuse are NOT evenly distributed across party ID and ideology as polling is obviously a media enterprise and Republicans/conservatives are far more likely to view the media as biased against them... so why would they talk to them? And, of course, because they won't talk to pollsters we cannot take a poll of them to find this out!!

You are certainly correct. This is a lot more interesting than 2008.