Bell Curve The Law Talking Guy Raised by Republicans U.S. West
Well, he's kind of had it in for me ever since I accidentally ran over his dog. Actually, replace "accidentally" with "repeatedly," and replace "dog" with "son."

Sunday, November 02, 2008

State of the race -- 2 days to go

This is a daily feature of this blog. For further detail, please see this post.

StateRCPPollster538538 Win %

Plenty of movement toward John McCain here as we finish up this last weekend before the election. Still, Barack Obama holds solid leads in Colorado, Virginia, Nevada, and Ohio. We'll see what happens for the other states.

Nate Silver reveals John McCain's most likely path to victory and it involves Obama winning only states either Gore or Kerry won. In other words, Obama would drop all the states I am tracking above. It's not a likely scenario -- Silver reports it only happens 169 times out of 10,000 -- but it's something to keep in mind. This is why I've been saying that Obama only needs one of the states we're tracking to win this election. It's much more likely that he'll win four or more than just one.


The Law Talking Guy said...

At this point 538 has more or less admitted that it is aggressively weighting the latest polls, more so than the other sites. The mason-dixon polls of today have had a bigger effect on 538 than the others. Tomorrow will be the biggest poll day, however, as the last pre-election polls are released.

The Law Talking Guy said...

If he only wins Nevada of those, he will at best tie in the electoral college.

The Law Talking Guy said...

RCP has a polling average with an all-time high for Obama at 50.7%. The gap is not an all-time high -- around 6% -- but the over-50% is a big deal. This means that undecideds are starting to shrink and both candidates are benefiting to some extent from these decisions.

It is unclear how undecided voters will break. Arguments can be made for both sides. Some say that if you haven't decided to vote for Obama by now, it's probably because he's black, or you're too scared, and you're not gonna. Others say that if you're still undecided about the incumbent party at this late date, you will probably break for the challenger. And for all his efforts, McCain is JAR: Just Another Republican.

I think the way to handle this intellectually, if you are caluclating Obama win scenarios, is to assume conservatively that 2/3 of the undecideds will vote for McCain. Similarly, if you are calculating a McCain win scenario, assume no more than 1/3 of the undecideds will go for him.

What you discover is that it doesn't matter. Unless the polls are simply wrong (which they can be) in any state where Obama is regularly polling close to or above 50% (VA, PA, CO), the chances of a McCain victory are negligible.

Raised By Republicans said...

I've seen some analysis on 538 I think that suggests that some of the pollers who sympathize with Republicans (like Fox, Strategic Vision, Mason Dixon etc) are suddenly changing their sampling distribution to produce a closer result.

But I think a lot of what is going on is the undecideds are finally deciding and they're breaking more for McCain than Obama. So the number to look at is not the gap between the Obama and McCain numbers but the Obama number itself. So as LTG says, in states where that number is already very close to or over 50% then the narrowing gap won't matter.

So states I'm fairly confident about (assuming the poll averages at RCP are reasonably accurate): PA, VA, FL, CO, and NM.

I'm worried but hopeful about: OH, NC, and MO.

Also, keep in mind that voting has already been going on for weeks in some of these states and the early exit polls on the early voters show pretty strong leads for registered Democrats. Which by the way, suggests that the polls that are sampling larger numbers of registered Republicans than the others are inaccurate - in other words, the polls showing things like McCain ahead by 2 in Ohio are based on nonsense.

The Law Talking Guy said...

The wide disparity in poll numbers represents a strong disagreement about who will vote, not really the underlying facts (except for the cell phone only voters, where we do see a 1-2% poll differential). If Democratic turnout is as many pollsters (but not all) expect, Dems will win very big. If it is more like historical levels (See Fox/Mason-Dixon) then it might be close.