Final NBC News/WSJ poll out: Numbers are bad but not nearly as bad as Gallup

[Politics_Guy25]

Just posting this for anyone who was concerned about Gallup's final poll.

NBC/WSJ disagrees that it will be 15 points.

They have it
Reps-49%
Dems-43%

among LV's and 46-44 GOP among RVs.

That matches up with Pew's final numbers EXACTLY btw.

Link here:
http://www.msnbc.msn.com/id/39940472/ns/politics-decision_08/

Between 2/6 points is a lot more manageable than 15 and that is why I post it here. Just to try to debunk Gallup a little.

[FBaggins]

1. They actually build their number in a different fashion.

They ask who you would vote for among republican, democrat, libertarian, green... then they ask the question "if it came down to just a republican or a democrat" of the people who selected green/libertarian/not sure... then they recalculate their data to combine the two questions. I'm not sure how directly that compares to the more common version of the question.

I agree that the "real" number is probably in between numbers like this one and Gallup/CNN/etc. It's been pretty steady in the six range for 2-3 months now (IMO). Maybe it's opened a tad since then, but I find double digits to be unlikely (or at least hard to accept).

The difference, though, is pretty clear (ignoring that noted above)... this poll takes a much broader approach to who is a likely voter. They polled 1,000 people and came up with about 750 LVs. The Washington Post poll was similar (greater than 75% of RVs made the LV cut). The polls that indicate a double-digit gap pare the sample down to more like 55% of the RV set.

[Politics_Guy25]

3. Don't forget though F that NBC/WSJ is the industry gold standard poll

and nailed 2008. They had President Obama up by 8. He essentially won by 8. Gallup had it like 12-13.

[FBaggins]

6. When did they become the gold standard?

The hardly do any polling.

They had President Obama up by 8. He essentially won by 8.

According to one list, they were ranked 13th out of 20. Some polls not only came close on the gap, but called the specific percentages correctly as well (and polled scores of other races). Many of the pollsters also broke it out by state and came close in most of those (as well as EVs), while NBC/WSJ only had a dozen or so polls all year (many of the top pollsters did that in the last few days).

http://www.fordham.edu/images/academics/graduate_schools/gsas/elections_and_campaign_/poll%20accuracy%20in%20the%202008%20presidential%20election.pdf

[Cognitive_Resonance]

2. One of these polls has to be radically wrong, and it looks like Gallup is the outlier. Two questions

1. Why is it so wrong?
2. What is else do they have wrong?

[FBaggins]

4. It's all in the LV screen.

The three recent polls with smaller 4-6 point gaps have LV screens with great big holes in them (letting through lots of people who won't actually vote). The three polls with 10-15 point gaps have LV screens that come much closer to the number of people who will vote, but not necessarily the correct subset of who will vote.

And on that difference lies the entire election. When Gallup/CNN/Fox/etc try to get down to the right sub-sample size... how close are they to identifying which RVs are actually LVs. If this poll is correct, then it's still a bad night (we have only rarely trailed at all on the generic poll)... if the tighter screens are correct, it's a nightmare.

[Politics_Guy25]

5. F-the senate polls would tend to support the...

hypothesis that the race is closer than 10-15 points. If it really was 10-15, Manchin would be doomed in WV, not up, Murray would be going down to defeat in Washington and Harry Reid would not be within 3-4 points etc.

Let's see what the final round of senate polls show, especially from PPP. If we're right in the thick of things, then Pew/NBC/ABC are probably more correct than FNC/CNN/Gallup. No way would we come anywhere near close to holding the senate if we were really down 15-16 points.

[tritsofme]

7. This Senate class is pretty favorable to Democrats

2004 was a pretty good Republican year so they had ground to defend, and the battle for control is being waged in bluish states.

If the class if 2006 or 2008 were up this year, it would likely be a bloodbath for Senate Democrats as well.

[FBaggins]

9. Not really.

What the senate polling shows is that we're unlikely to lose the senate. But that's not the real story if you're trying to use those polls to judge the accuracy of a national figure.

No way would we come anywhere near close to holding the senate if we were really down 15-16 points.

You have to remember that only roughly 1/3rd of the senate seats are even up this year and we're defending comparatively few of them. We're expected to lose a little less than half of the seats we're defending and win none of the seats held by republicans.

A comparable House performance would mean almost triple-digit loses. If the entire senate were up this year and the numbers looked the same in the other 2/3rd of the senate, we wouldn't be wondering if we were going to lose the Senate... we would be debating whether their win would be filibuster-proof or even veto-proof. If anything, the Senate polling shows us under performing those polls. This is because they focused on a dozen of these races early with tons of cash... making competitive races out of what should have been easy hold. That's not indicative of the national mood any more than our recent (apparently successful) "triage" defense of a handful of those seats speaks to anything national either.

If it really was 10-15, Manchin would be doomed in WV, not up

A guy who just won statewide in WV by 30%+ wins by four points and that's evidence that the national mood hasn't even shifted by 15%? I don't know.

[mahina]

8. But when only half the people show up to vote,

we can make a huge difference.

I still believe we're going to whoop ass on Tuesday, based how tight the primary was polling here prior to the election, and how bad we whooped em in the end.

Can do guys!