Monday, November 03, 2008

Latest polls; Reassurance for Dems in NBC/WSJ Survey

Whatever McCain "surge" there might have been has disappeared in the latest polls and Obama's back to being 96.3% likely to win at

Today's Polls, 3 AM Edition (11/3)

Barack Obama's position has become somewhat stronger since our update this afternoon. We now have him with a 5.8 point lead in the national popular vote, and winning the election 96.3 percent of the time. Earlier today, those figures were 5.4 and 93.7, respectively.

I continue to find a hair's worth of tightening on balance in the state-by-state polls -- even as Obama's position in the national trackers seems to be roughly as strong as it has ever been. This, ironically, is the exact reverse of the position we saw earlier in the week, when the national polls seemed to be tightening even as the state polls weren't.

However, Obama's win percentage has ticked upward again for a couple of reasons. Firstly, he's gotten some relatively good numbers out of Pennsylvania since our last update, with PPP and Zogby giving him leads of 8 and 14 points, respectively, and Rasmussen showing his lead expanding to 6 points after having been at 4 before. (The Zogby poll is probably an outlier, but may serve to balance out outliers like Strategic Vision on the other side).

Secondly, McCain's clock has simply run out. While there is arguable evidence of a small tightening, there is no evidence of a dramatic tightening of the sort he would need to make Tuesday night interesting.

Related to this is the fact that there are now very, very few true undecideds left in this race. After accounting for a third-party vote, which looks as though it will come in at an aggregate of 2 percent or so (after doing some work on this tonight, I concluded that I had been slightly underestimating the third-party vote before), I am showing only about 2.7 percent of the electorate left to allocate between the two major-party candidates. Even if John McCain were to win 70 perecnt of the remaining undecideds (which I don't think is likely), that would only be worth a net of about a point for him. Frankly, McCain's winning scenarios mainly involve the polls having been wrong in the first place -- because of a Bradley Effect or something else. It is unlikely that the polls will "tighten" substantially further -- especially when Obama already has over 50 percent of the vote.

It's very late, obviously, so we won't get into too much more detail, but a couple of quick notes.

--> Don't worry too much about that SurveyUSA result in Minnesota, which shows Obama just 3 points ahead. SurveyUSA's polling in Minnesota has been very, very weird all year; they've never shown Obama with larger than a 6 point lead in their likely voter model, and had McCain ahead in the state as recently as October 1st. SurveyUSA does not have a Republican lean in general, but in Minnesota, it has consistently had a huge one.

--> A couple of the national polls have now started to predict how undecided voters will behave and allocate them between the two major-party candidates. I use the versions of these surveys before any such allocations are made, as from my point of view it isn't the pollster's job to get into the prediction business (our model has its own ideas about how to handle undecideds).

Sunday, November 2, 2008

Reassurance for Dems in NBC/WSJ Survey

Of all the polls out late tonight -- and I do hope to have some sort of midnight update to the polling thread -- the one that ought to give Democrats the most reassurance is the new poll out from NBC and the Wall Street Journal, which gives Barack Obama a 51-43 lead. What's to like about this particular survey?

Firstly, all of the interviewing was conducted today (Sunday) and yesterday, so it's about the freshest set of data that we have.

Secondly -- and this is an underrated factor -- the NBC/WSJ poll always behaves intuitively. It goes up when the other polls go up, and goes down when the other polls go down:

Date             Obama   McCain
11/1 - 11/2 51 43
10/17 - 10/20 52 42
10/4 - 10/5 49 43
9/12 - 9/22 48 46
9/6 - 9/7 47 46
8/15 - 8/18 45 42
7/19 - 7/21 47 41
6/6 - 6/9 47 41
4/25 - 4/28 46 43
3/24 - 3/25 44 42
3/7 - 3/10 47 44
1/20 - 1/22 42 42

Those numbers very closely match our "supertracker" trendline at any given time period, albeit with perhaps a 1-2 point Obama house effect. (The antithesis of this is something like the IBD/TIPP poll, which had its best numbers of the year for Obama (+11) in mid-May, in the midst of the Jeremiah Wright crisis.)

Thirdly, it's hard to accuse the poll of partisanship, as it is co-directed by a Democratic pollster (Peter Hart) and a Republican one (Neil Newhouse), and is co-branded with a conservative-leaning newspaper and the most left-leaning of the three broadcast networks.

Finally, it includes cellphones (which may be part of the reason for the "house effect"). An updated version of the cellphone chart is below. (The polls that include cellphones are highlighted in orange, those that don't are in gray). Between the NBC poll and the final Gallup numbers, the discrepancy has now grown even greater: Obama leads by an average of 10.0 points in the cellphone polls, versus 5.1 in the landline-only's.



Post a Comment

<< Home