Someone has pointed out in the past that SUSA's numbers just before the primary tend to be really accurate, while their polls a few weeks out tend to be less reliable. Evidently, SUSA pushes leaners hard, causing its earlier polling to reflect a lot of soft support that may very well change. However, on the eve of a primary, the reported leanings are unlikely to change before everyone goes to the polls.

In any case, I wouldn't ignore SUSA's findings. And, I'm happy to avoid high expectations for Obama.

indeed I find it easier to say yes, Hillary is up by 18.

then just wait for election day to see what happens, if she wins by 18 k well we all saw it coming, and Obama has to do a bit better in NC.

but if she say she is at 18, and she only wins by 7, well thats kinda hard to spin.

so I saw let the media narrative be +18, and Obama and his supporters just keep on trucking and working hard.

What about the idea that robocalls might eliminate a "Bradley" effect in the other polls?

Given Rendell's comment about how a lot of people might not be ready to vote for a black man, SUSA might actually be more reflective of that reality.

Doesn't mean it's not accurate, though.

If Hillary can squeak out a 92% to 8% win, I'd say she just may have clawed herself back in the race.

Regardless, its absurd to think Hillary will win PA by 18 points.

10 -12 at the most in the worst case scenario (worst case for Obama), but 18 is ..well, you have to be on drugs to seriously think she will win by that much.

Even in CA and Ohio, and even NJ, it was like 10-12 points. So I don't believe SUSA because its just not realistic and in line with voting in other states even nearby.

Maybe the automated nature of SUSA polls is weeding out those who might be afraid to say they're for Hillary due the race issue.

Heck, this poll comes out just two weeks before the primary, and if Clinton comes close to this result, she's in it until June 3.  Could things be looking up for her just a little bit?  She was vindicated on the Ohio insurance story, she finally gained a superdelegate (albeit in Arkansas), and if she can stay close in Indiana and win Indiana, who knows?  A lot of ifs, but
Obama can't close the deal (a lot like Memphis last night), so at least Hillary has a slight chance.

Now if she completely get rid of Mark Penn....

Just some comments

Scranton/Wilkes Barre area has the second (or third) eldest population in the US outside of Florida.

According the the Quinnipiac poll, she is leading in the north East 66/33.

have consistently pushed the polls most favorable to him.

SUSA has been the gold standard for polls.

I can't imagine after all the coverage so far this year that there is really gonna be that much movement (unless there is some new development)

People know who they plan to vote for.

You can throw a dart or shake the Magic 8-Ball until you get the poll numbers you're hoping for.

The fact is, unless there's a 30+ point blowout in actual votes Pennsylvania, the elected delegate numbers for the two candidates won't change very much. So I'm all for letting the voters vote.

Obama better beat Clinton in Pennsylvania with the ton of money he is spending their per week.

Check out the headline in the page.

OBAMA'S WEEKLY PENNSYLVANIA TV BUY BREAKS ALL RECORDS

Democratic media consultant Neil Oxman, who isn't for either candidate: "Nobody has ever spent 2.2 million in this state: not Rendell, not Specter, not Casey, not Santorum, not Bush, not Kerry...That's unbelievable."

Nobody Has Ever Spent 2.2 Million in the State"

If he is spending this much per week and he is still trailing by 18 according to SUSA then there is a problem.

If she trumps him by that margin on election day , she is going to run out of the barn with that figure.

That is just incredible.

People seem to think that an outlier poll indicates that the pollster did something wrong.  But 1 out of 20 polls will be outside the margin of error no matter how good the pollster is.  Moreover the margin of error on this SUSA poll is 4.1%, which applies to both Obama and Clinton's numbers.  Thus, Clinton + 10% is actually within the SUSA margin of error.

This is why from a statistical perspective it makes more sense to rely on poll of polls than any one poll.  The quality of a pollster cannot prevent outlier results on occasion.  Given that this SUSA poll is about 8 points more pro Clinton than any of the 6 other polls from the last few days, it seems highly likely to be an outlier.

One particular reason I see this SUSA poll as a likely outlier are the internals showing something like 24% of blacks voting for Clinton and also showing Clinton winning in Philadelphia.  The margin of error on these kind of subgroups will be way higher than the 4.1% and there is just no way those numbers make any sense imo.

SUSA has done at least 20 states without being terribly out of their m.o.e.  This one they are wrong on.

The Obama people are doing a brilliant job of spending and planning for a potential PA upset, while deflating expectations - have they finally learned how to handle the expectations game?

If we're going to put the old SUSA poll up might as well use the silly-ass PPP poll too :-P

of the time, and those people are starting to figure it out.

Why am I denied this privilege?

I have asked this at least 4 times.

and time to campaign and money, if Obama cannot win this state, he is unelectable. Any other candidate with these kind of advantages should no doubt win the state. This is the real test of Barack Obama.

hillary has to blow out pennsylvania, therefore she will do so.  no question about it, hillary is leading by 18 percentage points...

what's going on with the SUSA poll are 3 things:

1. As Noam Scheiber points out, the latest poll indicates a significant softening of Obama's African-American support:
http://blogs.tnr.com/tnr/blogs/the_stump /archive/2008/04/08/obama-s-african-amer ican-problem.aspx

But if you compare the crosstabs from the two polls, you see that declining black support accounts for almost all of Obama's dropoff. In the last poll, Hillary led Obama 61-33 among whites; this time she leads 61-32. Among African Americans, on the other hand, Obama went from leading 83-17 to leading 74-24.

The black share of the electorate has dropped from 15% to 14% in this poll. Taken together, this explains half the six point swing (not "almost all"; Scheiber corrected himself in an update). I'm not at all sure it is a real shift, though. The poll's total number of black likely voters is 86. Unless they oversampled the AAs (no mention of that in the methodology), the MoE for the AA subsample must be enormous.

2. While Clinton's overall share of the Female vote remained steady at 61-33 (last week it was 62-34), the percentage of likely women voters increased in this poll to 58% from 54%.

3. Accordingly, the Male vote dropped from 46% to 42%, and Obama did even worse in that shrinking electorate, plummeting from a lead of 50%-43% last week to a deficit of 45%-48% today.

All of those shifts easily explain Clinton's 6 point gain from last week's 53-41 poll to this week's 56-38. But are those shifts real? After a generally positive bus tour through PA, with the new support of the very popular Senator Casey, what good reason would there be for Obama's black support to plummet 9 points? Why would his white male support drop a net 10 points?

I haven't had the time to compare these internals to those of all the other recent polls. But on its own, this new SUSA PA poll certainly looks like an outlier.

nothing dueling here. according to this:http://www.surveyusa.com/index.php/surve yusa-report-cards/
SUSA has 2% of error out of 26 recent election polls,
Rassmusssen has 7% error out 34 polls and
Quinnipac did only 6 polls and has 4.5% error.

And don't forget, PA will create huge waves, which will change entire picture into Hillary's favor.

Olbermann is only covering the Quin poll. The SUSA poll is not being mentioned at all. Is media bias this blatant or did I miss something here?

What has been the PPP margin of error..anyone?

Many people are wisely pointing out that with 2 weeks out, the polls have limited predictive power. So can we all apply that wisdom to the arguments that "McCain is leading [Clinton, Obama] by _points in _____. This proves that [Clinton, Obama] is guaranteed to lose in the general."

The general election is seven months away, folks. You can use those McCain v. ___ polls to line your birdcages.

This poll seems right to me. I just don't see Obama doing that well in PA. It's in Senator Clinton's backyard, the Senator has every big time politician but Casey in her pocket (including the big city mayors) and she has the most excellent professional  GOTV folks in the country in a team that includes Clemons who engineered NH and Texas. What else does she need to take a 15% lead out of there? C-mon. I am a pod for Obama who would love to see this thing end, but my Senator has every right to flex her PA muscle.  Obama is right and realistic to be working hard for a respectable 10% spread.

While polls this far out arent necessarily tremendous indicators of what the actual outcome will be, they are, like it or not, the best measure we have available.

I havent really followed polls much in detail before in this primary season as we never had a six week layoff with no other contest on the same day.  So I think gauging PA polls is going to become the equivalent of the NCAA college basketball bracket betting pools for the next two weeks.  It will become a sport to 'pick the margin'.  We will all become 'experts' and postulate our theories as time goes along.  

While the polls have some value as indicators, its seems odd to me so many of them are so different.  Is this common to have such disparity?  Are there are any measures to look at composites of polls and find how those composites do against actual results?

The best things these polls are worth right now is to shape narratives in the media.  Both sides will pick and choose what works for them and discard what doesnt.  And in the end it will probably be a bit closer than originally expected but still a cushy win for Hillary.

I do believe that if Obama can not close significantly it does give rise to some minor concerns about him and validates Hillarys position to stay in until the end.

The math may be fun for us geeks and it may be irrefutable on a purely math basis, but the math discounts emotions and irrationality which is essentially what the super delegate votes will be based on (not meant to imply that for supers to vote for Hillary is irrational - just implying humans dont always use reason or logic in making decisions).  

So I will be the first to put my purely scientifically derived bet down on the end result of PA - Hillary +12.  Someone care to put a bracket together?  Or is that gauche?

Good news is that if you look back at those 3 polls . . . SurveyUSA got it right and Rasmussen and Quinnipac both got it wrong...both had them polling closer together than SurveyUSA who hit it right on target.

It may not be 18% in PA but I'm feeling a 12%-15% win for Hillary in PA...and won't that suck for Obama who will probably have out spent her 4-1 in the state on advertising to lose by 12-15%...

I meant...look back at Ohio

also SurveyUSA hit CA right on the nose as well.

chuck todd said tonight on hardball that both camp's internals put the race in single digits.

I have to believe Obama thinks he has a shot given the amount he is spending

Seriously, the polls are not what they used to be. There was a time when you would call say 1000 people, ask them nicely a few questions, make a few statistics and voila. But not any more.

Just think about it, if a stranger calls you asking for some of your time and or money, how long does the conversation last? for me, I manages to detect those calls and hang up in less than 0.5 s. I suspect most people do the same.

That's what telemarketing did to us. Do you know what they do then. They take the result from the last election, the results from the last polling, tweek it a little and come up with a number straight out of their ass.

The business of polling is huge ( marketing new product etc... ) and they will never admit that it doesn't work any more.

Coming out this morning shows Clinton 45 to Obama's 42. No idea on their reliability, but it's another poll in line with the majority. As for Survey USA, someone on Politico had a good explanation.

"A reader points out that the difference between two Pennsylvania polls out today -- SurveyUSA has Clinton up 18, Quinnipiac has her up just 6 -- is fairly easily explained:

   All depends on whether you believe it's more likely for Obama to win Southeast Pennsylvania (including Philadelphia) 54%-40% or lose it 52%-43%"

That being said, I still think Clinton wins by a 10-12 point margin. I think there are many people in Pennsylvania who would lie about their racism, which would inflate Obama's numbers. The demographics of the state are tailor made for her.