and go with that.

Actually, the "safer action" at this point is just wait until Tuesday's results come in. It's close enough to Tuesday now, and the PA polls so different (and variable) that I think it's probably just a waste of time to see where in the +0% to +20% victory range Clinton falls. With the significant caveat that Obama and Clinton are, and have been for some time, within 10% of each other (signifying there's no national blowout factor at play), the national polls are just as irrelevant as they always were.

that info about pre and post debate is on Page 2 of the Newsweek story:
http://www.newsweek.com/id/132721/page/2

Stacy DiAngelo of Princeton Survey Research Associates, which did the April 16-17 polling, says that of the registered voters who were surveyed 517 were interviewed after the debate and 692 before. She added that the views of those surveyed remained largely constant.

Obama's down from an 11 point lead to only 3 points in a week, according to Gallup. He's tanking.

in the newsweek article:

Stacy DiAngelo of Princeton Survey Research Associates, which did the April 16-17 polling, says that of the registered voters who were surveyed 517 were interviewed after the debate and 692 before. She added that the views of those surveyed remained largely constant.

Johnathon - Any idea why Obama performs better in multi-day polls with large samples?  It seems like he always is further up in them than the tracking numbers (I also think Rass uses bad voter frames for their numbers)

all I know is that my poll can kick your poll's butt hands down

That last paragraph was some sage advice.

I saw the Newsweek poll and it is such an anomaly I don't take it seriously, especially reading posts and various polls from PA.

I don't trust these polls.  I think there will be a real change towards Hillary on Tuesday.  Her supporters are coming out in full force this weekend.  Thousands of people have volunteered and there is very strong support for her in Northern/Western PA.

Poll numbers PA and National are all over the place.

There will be nothing concrete until Tues, no matter what anybody says.

At least the "Hillary should step down!!" choruses seeem to have quieted down as the Bamabots realize their man is in deeep trouble.

He'sgone from Hope to Change to Bitter in two months, and he looked out of his league last night.

It'z just amazing what a record-setting media buy can do, ain't it?

Money talks, 'Bamabots walk!

LOL

It's appreciated!

Jonathon,

I read on Ambinder's blog that the Newsweek poll includes "leaning Democratic voters," which probably means that it includes Independents and the Gallup poll is registered Dems only.

If that's true than that would more than likely explain the differentiation of the two polls.

Knowvox,

Don't be so shortsighted.  He's not tanking.  He's dipped before. He'll be back up again... Like he always is.  I apologize if more people like him than Hillary.  I really do, but... more people would rather him be president than Hillary.  Sorry.

They are just out of sync with every natinal poll out there.

Frankly, Newsweek, like NBC/MSNBC, is in the tank for Obama.  Just like media twists its propaganda, they're probably doing the same thing with their polling.

They employ Markos and Karl Rove. That's all I need to know. They are not interested in unbiased reporting.

Counts from RCP

Comments?

http://www.haloscan.com/comments/ega...0 6622/?src=hsn

Quote:
omfg..it's as maddening as reading The Hunting of the President.

Guerilla Women...read this and weep. The thugs stole OUR PRIMARY with the complicity of the DNC!

They wanted to force FL and MI to hold undemocratic caucuses instead of (rethug) state-sponsored primaries!!!!

link and excerpts...

http://www.huffingtonpost.com/wayne-...u _b_94158.html

Dems Take the Hit for the GOP

The Republican role is not some irrelevant anecdote. The DNC is charged, under its rules, to determine whether the Democrats in a noncompliant state made a "good faith" effort to abide by the party's electoral calendar, and to impose the full weight of its available penalties, namely a 100 percent takedown of a state's delegation, only if Democratic leaders in that state misbehaved. So the fact that it was Republicans who fomented the move-up of primaries in both these states to dates out-of-line with the DNC calendar is at the heart of the matter.

...Back in June, a DNC spokeswoman, for example, told the Associated Press that neither Dean nor the Rules Committee "has the power to waive the rules for any state," explaining that "these rules can be changed only by the full DNC." Yet a few months later, on the same day that the Rules Committee stripped Michigan of its delegates, it waived the rules for New Hampshire, Iowa, and South Carolina, each of which had also moved up their primaries.

Though Dawson and others on Rules now say, as they did in recent interviews, that states whose contests were always scheduled before February 5 were free to shift dates without sanction, that's not what the delegate selection rules adopted in 2006 say. Those rules provided an automatic 50 percent loss of delegates for any state party that moved its contest to any day "prior to or after the dates" spelled out by the DNC.

That's why Rules powerhouse Donna Brazile said she would "grudgingly support the waiver," warning New Hampshire shortly before the December committee vote that "the days of 'privilege' may end soon."

Not only did "first-primary-or-die" New Hampshire switch from January 22 to January 8, it moved ahead of Nevada, whose January 19 caucus had been deliberately scheduled by the DNC to precede New Hampshire's. But New Hampshire's Democrats got a DNC waiver because their back was up against the wall, due to a decision by the South Carolina Republican Party to move its primary up to January 19. That unilateral decision -- which the Carolina Democrats declined to join in -- forced New Hampshire's hand. The waiver was, in other words, a reasonable response to a Republican provocation. What's unclear is why one Republican provocation is more equal than another.

A DNC official claimed that the Michigan party had sponsored so-called "firehouse caucuses" in the past and could have set their own date and done them again, ignoring the state-run January 15 primary. The Florida party, the DNC source added, was "offered $880,000" by the DNC to host their own caucus on a date in compliance with the DNC schedule and chose to participate, instead, in the state-financed primary, a "bad faith" decision.

But Florida party officials said the $880,000 would've only covered the cost of 150 caucus sites, with the capacity to draw a maximum of 150,000 voters out of the state's 4 million Democrats. "It wasn't a real offer," a spokesman said. Michigan's party would have had to self-finance caucuses, which, even with added Internet and mail voting, drew only 165,000 voters in 2004, a fraction of the 600,000 who voted in 2008. Stripping both states of their full delegations because the state parties in each refused to run these limited-participation caucuses--which would have occurred a couple of weeks after an official, state-financed primary -- is a bit like punishing Democrats because they like democracy.
Zee | 04.10.08 - 2:27 pm | #

As someone who spends quite a lot of time working statistics, I'd like to point out that a move from a polling result outside to one inside a margin of error is not, itself, necessarily a statistically significant move. And even the apparently significant 8-point swing over the week could represent anything from a wiggle to a big Obama drop.

The MOE on these Gallup tracking polls is +/- 3. That means >6 point separation is required for statistical significance at the 95% confidence level.

Yet, it also means that at that confidence level, the results of each finding in any poll are subject to a fluctuation of +/- 3% entirely as a result of sampling error.

So, Obama 49, Clinton 42 +/3 could be expressed as a "confidence interval" as Obama 46-52, Clinton 39-45. That entirely encompasses the results from the next day's rolling sample, O: 47 / C: 44.

Even though one day's results are "statistically significant" and the other day's are not, the difference between the two days is another matter. These poll results alone actually prove nothing about movement from day 1 to day 2 - that "movement" could be fluctuation around the true mean based on sampling error.

On the other hand, the movement from 51/40 to 47/44 across several days is more convincing. 51 +/- 3 is 48-54; 47 falls outside that range. Yet even there, not by so very much.

There are ways to test the level of confidence with which we can say 51/40 to 47/44 represents a real movement that require more data than Gallup publicly (at least to my knowledge) releases, but we can look at these MOEs and say yes, there likely is some real movement. Yet how much? Well, with full raw data from Gallup we could create a maximum likelihood model describing the most probable degree of movement. We can't say for sure that there has been an 8 point swing (O -4, C +4). That's merely one possibility in a probability cloud, that probably ranges from something like a "true" but rather trivial 2 point swing (O -1, C +1), to a much more impressive 14 point swing (O -7, C +7).

And in response to what? Well, just as the actual degree of separation between the two candidates has sampling error, and the meaning of a "movement" has sampling error, the timing of any movement has even more potential for statistical error. Looking strictly before the debate, the movement is all reasonably within the MOE. Looking strictly after the debate, the movement is all reasonably within the MOE. It's only combining the two that we have what appears to be significant movement. A cumulative effect of recent developments? Maybe.

Finally, we have to bear in mind that comparing two samples based on their margin of error would normally involve an assumption that the two samples are independent. Unless you're comparing across > 3 days, they're not independent (two consecutive rolling samples share 2/3 of their data). So it takes the Apr 14 - 17 separation before the statistics are simplified by meeting the assumption of two independent samples, rather than sharing some of the same error variance.

In other words, while we can look at these results over the last week and say yeah, this hasn't been Obama's best week of the campaign, what I think people really want to know is: a wiggle or a real drop?

Based on what Gallup releases... we don't actually have a basis to draw a conclusion. Even if we all had the statistical background to deal with all these issues, we lack the raw data to be able to run all the stats. We can just make rough guesses based on means and MOEs. This could be either a wiggle, or a plummet, or something in between.

Overall, honestly we spend way too much time dwelling on changes in horse race polls, especially just one poll, without really knowing what we're looking at.

And she clearly does better than Obama versus McCain.

Why does Obama, a "transformative" leader do so much worse than Hillary versus McCain?

Massachusetts:

Clinton 56
McCain 41

Obama 48
McCain 46

New Mexico:

Clinton 46
McCain 49

Obama 44
McCain 50

Missouri:

Clinton 47
McCain 46

Obama 42
McCain 50

New York:

Clinton 55
McCain 39

Obama 52
McCain 43

Ohio:

Clinton 53
McCain 42

Obama 45
McCain 47

Oregon:

Clinton 47
McCain 46

Obama 51
McCain 42

Virginia:

Clinton 39
McCain 55

Obama 44
McCain 52

Minnesota:

Clinton 47
McCain 46

Obama 49
McCain 43

Wisconsin:

Clinton 46
McCain 46

Obama 49
McCain 44

Kansas:

Clinton 36
McCain 57

Obama 37
McCain 54

Alabama:

Clinton 34
McCain 60

Obama 32
McCain 64

California:

Clinton 53
McCain 40

Obama 50
McCain 43

Iowa:

Clinton 42
McCain 48

Obama 49
McCain 42

Kentucky:

Clinton 46
McCain 48

Obama 29
McCain 63

http://www.surveyusa.com/electionpolls.a spx

http://www.surveyusa.com/index.php/2008/ 04/17/april-head-to-head-contests-six-st ates-flip/

jonathan, the fact that you're citing newsweek and olbermann is telling.  i don't know if you've already endorsed obama, but as someone who doesn't know whether you have, i can assure you that your bias is showing through loud and clear.  you clearly have no clue what the obama campaign has been doing stealthily for a year--and now not-so-stealthily in the past week in PA.