Another busy polling day, and one of Barack Obama's stronger ones in some while.
The headline is probably in Ohio, where SurveyUSA shows Obama with a 9 point lead over John McCain. Hillary Clinton was not surveyed in this matchup. It's interesting to wonder, as Markos Moulistas has speculated, whether Obama polls stronger when Clinton's name is not mentioned in the same survey. There would be a fairly easy way to test this: if a pollster rotates the question order of Clinton and Obama (as they probably should be doing), we could see whether one Democrat tends to benefit from going first. We may explore this topic further in the upcoming days.
In Pennsylvania, Rasmussen has Obama +2, Clinton +11. This is another of those good news, bad news results for Obama: it's stronger than his last Rasmussen poll in the Keystone (which showed him trailing by a point), but somewhat weaker than his other results in the state. For Clinton, of course, this is unambiguously a good result, and Pennsylvania now looks so strong for her -- we have her at 86 percent to win the state -- that she might be able to get away with a limited investment of resources there in November.
We actually have two new New Hampshire polls. Rasmussen has Obama ahead by 5, and Clinton ahead by 10. But there is also a Dartmouth poll showing McCain leading both candidates: he was 2.5 points ahead of Obama and 8.8 points ahead of Clinton (they list the decimal point, so I do too). Of the two polls, Rasmussen is the one to take more seriously, simply because the Dartmouth poll had a small sample size and was in the field as much as three weeks ago. Nevertheless, this underscores one of my theories about New Hampshire: it has a very engaged electorate, and it tends to serve as a weathervane for currents of momentum.
Finally, a Research 2000 poll for Daily Kos shows McCain leading Obama by 15 points and Clinton by 19 in Mississippi. Mississippi has occasionally been discussed as a reach state for Obama. I don't buy that, as even with extremely heavy black turnout, the demographics just aren't favorable enough for him to get the state closer than about 8 points. Nevertheless, the Senate race in Mississippi looks to be competitive (Roger Wicker, who inherited the seat from Trent Lott, had a 46-42 lead in the R2K poll), so there may be incentives for the Democrats to invest some time there.
5.23.2008
Today's Polls, 5/23
by Nate Silver @ 11:17 AM...see also mississippi, new hampshire, ohio, pennsylvania, today's polls
Subscribe to:
Post Comments (Atom)

22 comments
Rasmussen's daily tracking poll has McCain and Obama tied at 45% each, which is completely at odds with his state-by-state findings.
How could the national race be tied if Obama is crushing McCain in Ohio by 9 points and has gone from 10 points behind in NH last month to 6 points ahead of McCain in NH today?
I hope Rasmussen will address this.
Rasmussen actually has McCain up in Ohio Mike.
Additionally, I don't think anyone (who knows anything about polling) is taking this Survey USA poll seriously, especially given (1) the internals and (2) the Quinnipiac survey (which is known for left leaning bias) with twice the sample size that has McCain ahead by 4.
Look at the internals: McCain got 17% of the black vote (yeah right), 62% of the Hispanic vote (I doubt this too), won independents by TEN points and got 81% of Republicans 14% of Democrats and STILL lost? Yeah, I'm not buying that. The poll was weighted 52/28 D/I, and, given the internals, that's the only reason Obama won it. There's no way the turnout on election day is going to look anything remotely like that. Obama may win Ohio, but it'll be razor thin.
Meant to say 52/28 D/R. I'd really like to hear Poblano's opinion on the survey given the information I posted. Maybe there *is* a reason to take it seriously?
The surveyusa Ohio poll overweights Democrats and underweights Republicans. So I have to question it's validity.
Is that breakdown representative of Ohio's electorate?
I have to agree with the speculation that polling multiple matchups like that (i.e., asking each respondent about both rather than rotating questions) is polluting the polls.
This is most noticeable in SUSA's polling of the effect of various VP picks on both side. For those who haven't seen it, they picked four possible picks for both Obama (Edwards, Sebelius, Rendell, Hagel) and four for McCain (Huckabee, Lieberman, Romney, Pawlenty) and, after asking about a generic Obama/McCain matchup, polls each respondent about all 16 matchups with VP.
This is producing wild variations: In Ohio, for example, the results range from O+18 for Obama/Edwards vs McCain/Pawlenty to M+2 for Obama/Rendell vs McCain/Huckabee. Now this is silly on it's face: no reasonable VP choice will produce anything like that effect on the overall ticket. 20 points of variation? No. They need to poll 16 times as many people and rotate the matchups.
In fact, even within the irrelevant universe of this poll (which might, theoretically, gauge the relative strength of various VP picks) these folks mostly have low name recognition. So, however much I'd love it if Edwards really was the strongest pick, I think his very strong showing is a reflection of high name recognition.
Also: Hagel? Rendell? While possible, how about, oh I don't know... Clinton and Richardson? or Webb and Richardson?
On other notes: 2004 convinced me that adjusting based on party id is the wrong way to go. Obviously, adjust based on things like race, gender, etc (although, guessing AA turnout will be a big topic of conversation this fall I'll bet). People change self-reported party id a lot more than I thought in '04. Hurt my side then, but may help us now (with respect to actual results vs. polls).
A little disturbed that Nevada is +6 for McCain...
Off topic:
I do have a new wish from you Poblano. This one is easy to do and I think would be very helpful. So pleeeeease!
Could you please add a new row into the (fabulous!) History tables, with how the current prediction pans out?
I think this is an important check on the reality of the prediction, especially using the relative tables.
There has been at least a few polls recently where Hillary-McCain hasn't been asked. Is there enough data to see if it really is polluting the Obama numbers? (I assume you could look at previous polls by the firm in the state that looked at both match-ups, and correct for change over time by looking at the tend from polling firms that are still mentioning both match ups.)
Apologies in that's not feasible/wouldn't be a valid comparison.
With regard to what tilthouse said, name recognition probably does play a big role at this stage, but that still doesn't totally explain why Edwards is polling so much better than the other three potential VP picks they looked at. Consider that an Obama-Edwards ticket polls about 5 points better than an Obama-Rendell ticket in Pennsylvania. That can't possibly be a name recognition advantage. I agree that it would have been nice to see them poll people like Richardson, Webb, and particularly Clinton(given all of the current speculation about Clinton as VP, it would be nice to see how well that ticket polls).
I'm also a little skeptical that VP choices really matter that much in the end, but a good chunk of voters do say that they matter when they're polled. The amount of variability is consistent with that finding.
Tilthouse wrote "On other notes: 2004 convinced me that adjusting based on party id is the wrong way to go."
As you may recall in 2004 some pollsters (e.g., Zogby) did it, and others (e.g., Gallup) did not weight or adjust by party id.
What I found, polling within one state in 2004 was that we had to weight by party ID if we were to get our results to reasonably match up to the actual voting in the November election (speaking of analysis done after the election) -- otherwise, we'd have overestimated the Bush percentage by about 5 points.
Why did this happen? Not because of our "sampling," but in fact we did find a party ID shift of about that same percentage in the surveys we did in the early fall compared with surveys we'd done on a regular basis in the last year.
We always have to keep in mind that participation in surveys is a form of political participation or even mobilization. (cf. John Brehm's work.) For whatever reason (and we've polled this state a lot), in that particular period Republican respondents were more ready to talk to us than were Democrats. No change in our sampling frame or the way we asked questions about vote preference or screening for likely voters.
When we did make that adjustment for party id (using the average of the 7-point party id distribution for 4 surveys taken in the previous year), we hit the actual state voting margin for Kerry within 1 point.
If Bob Barr win the Libertarian Nom this weekend, he'll take 5 - 10 points across the country. Sure, that isn't enought for Obama to win the nom (Barr's supporters, based on polling are about 2:1 republican to democrat), but it is enough to make McCain spend limited resources there, and it may be enough in the Carolinas or Georgia.
If Barr wins the Libertarian nomination, it will hurt McCain on a couple of fronts. There's the obvious one, that he'll siphon off votes from McCain, but there will also be a nice-sized chunk of people who traditionally vote Libertarian who can't stomach voting for Barr.
I personally know several people like that - I'm one of them - and the vast majority of them are leaning Obama.
"them" being the disenchanted Libertarians I know. I can't speak for people I don't know.
Re: Party ID comments
I'm not a pollster, and don't really presume to tell them what to do or not. The party id comment was referring to the 2004 polls where lots of people in left-wing blogosphere were dismissing or re-adjusting polls based on "obviously" too-high GOP party ID and the "fact" that party id is very stable. Doing this led a lot of people to have a false sense of confidence.
I think the same discussions will come up again with respect to turnout estimates for highly motivated AA and youth and newly-registered voters. And we won't know who is right until after the election.
Re: Party ID comments
I'm not a pollster, and don't really presume to tell them what to do or not. The party id comment was referring to the 2004 polls where lots of people in left-wing blogosphere were dismissing or re-adjusting polls based on "obviously" too-high GOP party ID and the "fact" that party id is very stable. Doing this led a lot of people to have a false sense of confidence.
I think the same discussions will come up again with respect to turnout estimates for highly motivated AA and youth and newly-registered voters. And we won't know who is right until after the election.
Pennsylvania now looks so strong for her -- we have her at 86 percent to win the state -- that she might be able to get away with a limited investment of resources there in November.
I bet she will get away with spending very little in Pennsylvania, considering the fact that she has no chance to win the nomination.
Sorry about a few technical errors in my post # 1; I was in a hurry.
But my point remains: The daily Gallup and Rasmussen tracking polls both have the Obaam-McCain race tied today yet this week has seen a wave of shocking pro-Obama polls in key states, with SUSA putting Obama ahead by 7 in VA and by about 9 in OH, for example and Obama surging to a 14-17 point lead in CA in 2 polls.
How can these nationwide and swing state polls be reconciled?
I think it comes down to clinton not being in the poll, without her around Obama's numbers look stronger
djl - As I'm also a Libertarian, I'm amused you think Libertarians disenchanted with Barr will boost Obama. For one, only 0.3-0.5% voted Libertarian in the last 6 presidential elections so a fraction of that number won't tip anything. And two, a Libertarian who would reject a party member, long-time advocate of small-gov't, and zealous convert against Iraq and the drug war, in favor of the Senate's most liberal (read: big gov't) member is motivated by something other than small gov't Libertarian impulses and would probably be leaning Obama regardless of the Libertarian nominee's identity.
情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣用品,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,情趣,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店經紀,
酒店打工經紀,
制服酒店工作,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
專業酒店經紀,
合法酒店經紀,
酒店暑假打工,
酒店寒假打工,
酒店經紀人,
菲梵酒店經紀,
酒店經紀,
禮服酒店上班,
酒店小姐兼職,
便服酒店工作,
酒店打工經紀,
制服酒店經紀,
酒店經紀,
菲
梵,
bit ilaçlama
bit ilaçlama
bit ilaçlama
bit ilaçlama
bit ilaçlama
bit ilaçlama
Post a Comment