PA-SurveyUSA: Clinton regains 18-point advantage over Obama

[Tender Branson]

Barack Obama lost ground among certain groups of voters in the battleground state of Pennsylvania during the last week, according to a new poll.

If the Democratic Primary were held in Pennsylvania Tuesday, Hillary Clinton would beat Obama by 18 percentage points, according to the results of a SurveyUSA poll conducted exclusively for NBC 10 and three other TV stations across the Keystone State.

The poll concluded that the results were nearly identical to one of its polls released a month ago, stalling the momentum Obama picked up at the end of March.

Women and white voters barely budged and continued to support Clinton at more than 60 percent, as they have in the previous two Survey USA polls.

However, a change in men's support has negatively affected Obama.

Last week, Obama moved within 12 percentage points of Clinton, but men who flirted with the notion of voting for Obama at the end of March appear to be moving their support back to Clinton.

The poll suggests that men make up 42 percent of likely voters.

Obama also lost ground among voters between 35 and 49, down 18 points, while Clinton continued to lead among voters 50 and older.

In the Southeast section of the state, which includes Philadelphia and makes up 42 percent of likely Democratic voters, Clinton had a good week, polling above 50 percent for the first time.

Obama trailed Clinton by nine points in the critical part of the state.

In the southwestern part of the state, which includes Pittsburgh, Clinton jumped 25 points ahead of Obama. However, the Johnstown area continued to show consistent movement toward Obama, the survey concluded.

Obama also lost ground among moderates and liberals and conservatives. Clinton only fell among conservatives.

Last week's poll, which showed Obama closing the gap, was conducted in the middle of Obama's six-day bus tour through Pennsylvania, which may have helped boost his numbers.

http://www.nbc10.com/politics/15821754/detail.html

[Alcon]

Uh, I'm going to guess 1-in-20.

[Meeker]

So what do we do when the "most reliable" pollster conflicts with all the other polling?

[exopolitician]

Okay...

[Sam Spade]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

[Tender Branson]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

Rasmussen will also come out with a new PA in a few hours. Let´s see what this poll shows ...

[Sam Spade]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

Rasmussen will also come out with a new PA in a few hours. Let´s see what this poll shows ...

Probably tightening...

[Tender Branson]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

I already had a bad feeling that things would go downhill for Obama from the moment ARG showed him tied with Clinton ...

[Smash255]

1 in 20

[Ben.]

Cant really see any reason for Clinton to bounce back in PA so suddenly, if anything she's had the worst of it over the last week or so... then again maybe we're all underestimating the impact of Bowling-gate?

[Wiz in Wis]

I think we need to see internals. Apparently this poll has her winning metro Philly by 9... which makes zero sense in light of all the other polls we've seen

[Verily]

SUSA also had the widest Clinton lead last time around. That makes suspect this is off, but they have done well so far (although California was primarily due to early voters, which aren't a factor in PA, and SUSA's been no better than average otherwise).

[Verily]

I think we need to see internals. Apparently this poll has her winning metro Philly by 9... which makes zero sense in light of all the other polls we've seen

Internals like that probably have little meaning, although it may belie an undersampling of blacks or something.

[Wiz in Wis]

I think we need to see internals. Apparently this poll has her winning metro Philly by 9... which makes zero sense in light of all the other polls we've seen

Internals like that probably have little meaning, although it may belie an undersampling of blacks or something.

yeah, my point was on the AA sample.

[Democratic Hawk]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

Rasmussen will also come out with a new PA in a few hours. Let´s see what this poll shows ...

Probably tightening...

Rasmussen (4/8):

http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/pennsylvania/pennsylvania_democratic_presidential_primary

Clinton 48% (+1) / Obama 43% (+1); Clinton is maintaining her five-point lead over Obama of a week ago

Dave

[Wiz in Wis]

1 out of 20, or SUSA's picking up on something we're missing.  I would lean towards the latter, though in general SUSA has picked up on certain things this year other pollsters haven't (I have a possible theory about that - for later).  I do want to see internals before saying anything.

I think they should have numbers out on NC fairly soon.  That will tell us something, actually - vis-a-vis this poll.

Rasmussen will also come out with a new PA in a few hours. Let´s see what this poll shows ...

Probably tightening...

Rasmussen (4/8):

http://www.rasmussenreports.com/public_content/politics/election_20082/2008_presidential_election/pennsylvania/pennsylvania_democratic_presidential_primary

Clinton 48% (+1) / Obama 43% (+1); Clinton is maintaining her five-point lead over Obama of a week ago

Dave

Survey USA looks like they dropped a stinker... O well, they still have 2 weeks to make up for it

[Wiz in Wis]

While the internals don't give us a clear indication of bias, it is instructive that they are showing Obama only winning 75% of blacks... I don't think whites could really be more than 60-40 for Clinton, so this shows her at her max in that demo. I would guess that this means its really around 10 points when the racial stuff is corrected.

[Sam Spade]

The interesting thing is that this poll and the Quinnipiac poll look roughly alike (within MOE), except for two major subgroups.  Interestingly, it's the two subgroups I would suspect would be different.  Food for thought.

[Eraserhead]

This poll has Clinton getting a solid fourth of the black vote and leading among the 18-34 age bracket. Okay...

[ChrisFromNJ]

I'll believe SurveyUSA over any of the other hit and miss polling organizations. This is a bit dissapointing...

[Gustaf]

The whole "1 in 20" business is one of the most misunderstood concepts on this forum.

[Wiz in Wis]

The whole "1 in 20" business is one of the most misunderstood concepts on this forum.

Ok, as someone with a background in stats, I would suggest that the 1 in 20 rule means that 5% of the time the actual results lie outside the margin of error... to be 15 points off is actually more likely a 1 in 100 result, but that too seems posssible given the number of polls we've seen in this cycle. My guess, Survey USA got unlucky in 2 key subsamples and otherwise would have reported Clinton by 5-10, a more reasonable response. It could also be that SUSA is weighting things differently than all other pollsters, a la PPP, and they are actually correct. Time will tell.

[Gustaf]

The whole "1 in 20" business is one of the most misunderstood concepts on this forum.

Ok, as someone with a background in stats, I would suggest that the 1 in 20 rule means that 5% of the time the actual results lie outside the margin of error... to be 15 points off is actually more likely a 1 in 100 result, but that too seems posssible given the number of polls we've seen in this cycle. My guess, Survey USA got unlucky in 2 key subsamples and otherwise would have reported Clinton by 5-10, a more reasonable response. It could also be that SUSA is weighting things differently than all other pollsters, a la PPP, and they are actually correct. Time will tell.

Exactly. 1 in 20 polls will be outside of the MoE, but in the VAST majority of those cases it will not be FAR outside of the MoE. People talk about how 1 in 20 "blows up" as if it would suddenly jump 10% away or something. That is pretty unlikely.

[Sam Spade]

The whole "1 in 20" business is one of the most misunderstood concepts on this forum.

Ok, as someone with a background in stats, I would suggest that the 1 in 20 rule means that 5% of the time the actual results lie outside the margin of error... to be 15 points off is actually more likely a 1 in 100 result, but that too seems posssible given the number of polls we've seen in this cycle. My guess, Survey USA got unlucky in 2 key subsamples and otherwise would have reported Clinton by 5-10, a more reasonable response. It could also be that SUSA is weighting things differently than all other pollsters, a la PPP, and they are actually correct. Time will tell.

SUSA doesn't weight, FYI.

[Mr. Morden]

The whole "1 in 20" business is one of the most misunderstood concepts on this forum.

Ok, as someone with a background in stats, I would suggest that the 1 in 20 rule means that 5% of the time the actual results lie outside the margin of error... to be 15 points off is actually more likely a 1 in 100 result, but that too seems posssible given the number of polls we've seen in this cycle. My guess, Survey USA got unlucky in 2 key subsamples and otherwise would have reported Clinton by 5-10, a more reasonable response. It could also be that SUSA is weighting things differently than all other pollsters, a la PPP, and they are actually correct. Time will tell.

Exactly. 1 in 20 polls will be outside of the MoE, but in the VAST majority of those cases it will not be FAR outside of the MoE. People talk about how 1 in 20 "blows up" as if it would suddenly jump 10% away or something. That is pretty unlikely.

True, though of course the above only deals with statistical error.  Since different pollsters have different sample selection methods, we shouldn't be surprised that polls from different pollsters taken over the same timeframe disagree w/ each other outside the MoE all the time.  They're sampling different populations of voters.