The Vorlon Slams another CBS poll (redux)

[The Vorlon]

This is getting to be a broken record but....

The "snap shot" poll CBS did last night reporting to show a change in the electorate due to Mr. Edwards had a few problems... again...

This was not a "fresh" sample - but rather they called back a subset of the people they had also contacted for their June 23rd poll.

In the ENTIRE June 23rd sample, Bush's favorability ratings within the sample were:

Unfavorable 45
Favorable 39

For a net favorability rating of -6.

In the portion of the sample CBS chose to call back for last nights snap shot - they called a very differerent group:

In last night's sub sample, selected from the previous June 23rd sample, Bush had the following favorability ratings:

Unfavorable: 48%
Favorable 37%

A net unfavorabe rating of -11..

IE CBS knew for an absolute fact the subsample they talked to last night viewed Mr. Bush 5% more unfavorably than the comparison June 23rd sample.

This was not a "random" event - this was a hard cold fact CBS news was aware of.

Wait... it gets worse...

Ie every single question they showed us how the subsample responded on June 23rd versus last night to show a true "apples to apples" comparison

EXCEPT

(wait for it)

on the question where they compared how people planned to vote...

Here they compared last nights samples (5% more unfavorable to Bush) with the ENTIRE june 23rd sample, rather than do an "apples to apples" comparison. - The single and only question in the entire survey where they compared last night's subset to the entire June 23rd sample, rather than the exact same group of people from 2 weeks ago..

Summary...

CBS called back a SUBSET of their June 23rd sample which they KNEW (based on their June 23rd answers) was 5% more unfavorable to Bush than the entire June 23rd sample..

And then reported a 4% shift to Kerry/Edwards based on this subsample which they KNEW was 5% more hostile to Bush.

Now it is certainly possible, indeed likely, that they tried to call back everybody from their June 23rd sample, and just happened to reach a subsample more hostile to Bush (given typical survey completion patterns in a "snapshot" poll, this is indeed quite probable)

But don't you think, when they know for an absolute fact that last night's subsample was 5% more hostile to Bush than their June 23rd sample that they should have either specifically flagged the difference between the two samples, or at least shown us how last nights subsample "voted" on June 23rd so we could do a true "apples to apples". comparison...?

These people are either idiots, or deliberately trying to slant the news...

I can't think of any other explanation...

[© tweed]

What do you think of the NBC poll?

[CARLHAYDEN]

This is getting to be a broken record but....

The "snap shot" poll CBS did last night reporting to show a change in the electorate due to Mr. Edwards had a few problems... again...

This was not a "fresh" sample - but rather they called back a subset of the people they had also contacted for their June 23rd poll.

In the ENTIRE June 23rd sample, Bush's favorability ratings within the sample were:

Unfavorable 45
Favorable 39

For a net favorability rating of -6.

In the portion of the sample CBS chose to call back for last nights snap shot - they called a very differerent group:

In last night's sub sample, selected from the previous June 23rd sample, Bush had the following favorability ratings:

Unfavorable: 48%
Favorable 37%

A net unfavorabe rating of -11..

IE CBS knew for an absolute fact the subsample they talked to last night viewed Mr. Bush 5% more unfavorably than the comparison June 23rd sample.

This was not a "random" event - this was a hard cold fact CBS news was aware of.

Wait... it gets worse...

Ie every single question they showed us how the subsample responded on June 23rd versus last night to show a true "apples to apples" comparison

EXCEPT

(wait for it)

on the question where they compared how people planned to vote...

Here they compared last nights samples (5% more unfavorable to Bush) with the ENTIRE june 23rd sample, rather than do an "apples to apples" comparison. - The single and only question in the entire survey where they compared last night's subset to the entire June 23rd sample, rather than the exact same group of people from 2 weeks ago..

Summary...

CBS called back a SUBSET of their June 23rd sample which they KNEW (based on their June 23rd answers) was 5% more unfavorable to Bush than the entire June 23rd sample..

And then reported a 4% shift to Kerry/Edwards based on this subsample which they KNEW was 5% more hostile to Bush.

Now it is certainly possible, indeed likely, that they tried to call back everybody from their June 23rd sample, and just happened to reach a subsample more hostile to Bush (given typical survey completion patterns in a "snapshot" poll, this is indeed quite probable)

But don't you think, when they know for an absolute fact that last night's subsample was 5% more hostile to Bush than their June 23rd sample that they should have either specifically flagged the difference between the two samples, or at least shown us how last nights subsample "voted" on June 23rd so we could do a true "apples to apples". comparison...?

These people are either idiots, or deliberately trying to slant the news...

I can't think of any other explanation...

I repeat what I have earlier stated.

Many media polls are deliberately slanted to the left.

Even though they are inaccurate, the liberal media continues to fund them because they tell the liberal media what they want to hear.

Also please note that of the three old television networks, CBS evening news has the lowest ratings and every study I have seen shows they have the greatest left-wing slant in their 'coverage.'

In short, CBS is controlled by left-wing idiotlogues of the far left who aren't interested in profts, but rather are interested in promoting their agenda.

Perhaps you may recall a fictionalized character called 'Ellsworth Toohey.'  

[King]

I don't like CBS national news.

[TeePee4Prez]

Hey Vorlon,

                I hate to nail you here, but you have PA as <2% Bush.  Based on local factors, I'll have Kerry by 4-6%.  Gore got 4% last time and the Southeastern suburbs are going more Kerry than Gore.  Please enlighten me as to your reasoning.  Thanks.

[○∙◄☻¥tπ[╪AV┼cVê└]

What do you think of the NBC poll?

Haha, we're stuck with the Kerry +8 poll.

[The Vorlon]

Hey Vorlon,

                I hate to nail you here, but you have PA as <2% Bush.  Based on local factors, I'll have Kerry by 4-6%.  Gore got 4% last time and the Southeastern suburbs are going more Kerry than Gore.  Please enlighten me as to your reasoning.  Thanks.

Ok...  I have Bush up 0.2% in my model - call it tied.



If we throw out the most recent Fox Poll showing a decent Bush lead, and the last Quinnipiac showing a decent Kerry lead, every single poll in the last 4 months has shown the race within margin of error of being tied. (except a tiny sample University poll which I toss out without even looking at)

This looks a whole like like a tossup to me, with perhaps a modest, modest Kerry lead.

Many of the polls shown here were also done in April/May, at the height of the Iraq/Prison scandal mess when Bush was down maybe 3% Nationally.

Currently (at least prior to whatever "Edwards Bounce" we see) Bush and Kerry are basically tied Nationally, so relative to the May Timeframe, Bush is actually gained a few points.

The model I use is driven by both state polls and natiional polls.

To give you an example, lets say that on June 1st Candidate X was down 5% in State Y, and the National polls on average said that he was down 10% nationally.

If a month later the average of the National polls was Candidate X being down only 5% (ie a gain of 5% overall) I would also index the state poll that showed him down 5% also by 5%, to a "tie" in that state.

Based on the May polls showing a very modest Kerry lead, and Bush gaining maybe 2 or 3 in the last month, I have Pennsylvania standing at Bush + 0.2% (a very very light blue) Call it a tie.

If you want to adjust this to a very very light Red for Kerry I certainly would not object

I think "tossup" is a correct call at this point.

Remember, Gore only won by 4.14% in 2000, and the Bush "ground game" in Pennsylvania in 2000 was just wretched - they got absolutely killed "on the ground".

Assuming the GOP's GOTV in 2004 is even average they should, relative to the 2000 baseline, gain 2-3%.

Remember too this is a "snapshot" not a Nov 2 prediction.

My Nov 2 Prediction has Kerry holding Pennsylvania, likely by about what Gore got in 2000 or so.


Just one guys opinion..

No warranty expressed or implied

[millwx]

Ok...  I have Bush up 0.2% in my model - call it tied.

Vorlon, any clue on Nader's ballot status or potential in PA?  Makes a big difference.  In Fox/OD the Bush lead goes from 5% to 3%.  And in Quinnipiac the Kerry lead goes from 1% to 6% when you get rid of Nader.  So...

Just using the June polls (i.e., adding in SurveyUSA... which "sort of" included Nader... they had "other")...

With Nader the three poll avg is... Bush 45%  Kerry 44%

Without Nader it's... Kerry 46.7%  Bush 45.3%

Only a small difference, but it "flips" the race.  So.... any idea on Nader's status in PA?

[The Vorlon]

Ok...  I have Bush up 0.2% in my model - call it tied.

Vorlon, any clue on Nader's ballot status or potential in PA?  Makes a big difference.  In Fox/OD the Bush lead goes from 5% to 3%.  And in Quinnipiac the Kerry lead goes from 1% to 6% when you get rid of Nader.  So...

Just using the June polls (i.e., adding in SurveyUSA... which "sort of" included Nader... they had "other")...

With Nader the three poll avg is... Bush 45%  Kerry 44%

Without Nader it's... Kerry 46.7%  Bush 45.3%

Only a small difference, but it "flips" the race.  So.... any idea on Nader's status in PA?

I "think" Nader has ballot access in Pennsylvania via the "progressive party" ballot line? - I really don't know actually.

I just don't think Nader matters IMHO.

He might get 1% or so.  But if Nader was NOT on the ballot that 1% would go to some other "protest vote candidate", or at least a big chunk of it.

Nader is "useful" in the polls in that it is a good index of undecided/very soft Kerry support, But in the end... a pretty minor factor.

[stry_cat]

I "think" Nader has ballot access in Pennsylvania via the "progressive party" ballot line? - I really don't know actually.

http://www.votenader.org/ballot_access/pennsylvania/
says:
"Goal:Our goal is to get 40,000 signatures by July 26th to get Ralph on the ballot. This will involve a massive volunteer effort, but we can do it."

So I'm thinking they don't have ballot status yet.

[millwx]

"Goal:Our goal is to get 40,000 signatures by July 26th to get Ralph on the ballot. This will involve a massive volunteer effort, but we can do it."

I just saw that as well.  Also looks like the requirement for signatures is probably quite high and/or strict.  Who knows, but my best guess is they won't make it.  They're struggling so badly in so many other states, I can't see them making it in a state with heavy requirements.  But, then again, what about the Reform party?  Are they on the ballot in PA?  Does PA, like FL, require a convention (that will likely keep Nader and the Reform party off of the FL ballot)?

[CARLHAYDEN]

I don't like CBS national news.

Not too many people do.

Hence their poor showings in the Nielsons.

[TeePee4Prez]

Vorlon,

         Still have 2 objections.  One being Fox is Republican biased.  They even have NJ with a Bush lead.  I find them as non-credible as Zogby.  Second factor is most wealthier people have more phone lines.  Philadelphia(inner city) is underrepresented here.  If I recall from last time, Bush had at times a commanding lead in most polls.  This time based on your analysis, Bush is in a virtual tie.  My conclusions are based on the "ground" here, not Fox News analysis.

[I spent the winter writing songs about getting better]

most of the PA polls showing Bush with a big lead are crappy Uni ones.

[The Vorlon]

Vorlon,

One being Fox is Republican biased.  They even have NJ with a Bush lead.  

I am not aware of a Fox poll showing Bush leading in New Jersey.  Please provide a link.

Regarding polls:

Ok.. let's toss both Fox and Quinnipiac (the last Q poll was a really strange and quirky sample)

The last poll from a firm I semi trust was Survey USA which showed a 1% Kerry lead. (Rasmussen which I don't trust had a 1% Bush lead)

Re phone lines:

Single people (one phone line) have one vote per line.

Married people (one phone line usually) have TWO votes per phone line.

Bad polls thus over sample single people (who vote strongly for the Democrats) while under sampling Married people (who break for the GOP) - this is why (among other things) the last 8 LA Times polls have averaged +11 more Dems than GOP. (Have I mentioned  today that the LA Times poll is a piece of $%$%$/...? )

Actually, GOOD polls compensate for the phone lines thing you mention.

A very simplified explanation is they take census track information to know the total adult population in a give geographic area, and then overlay that with the number of active residential phone lines to get the average number of voters per active phone line.

An area that had say 2.0 adults per phoneline would get 50% more calls directed to it that an area that had 1.33 adults per active phone line.

University and other lesser quality phones do have the "phone line" issue, but it tends to skew to the Dem side rather than the GOP side as you suggest due to the single/married issue I identified. (See LA Times for more information)

PURE random digit dialing, used by many low grade firms is just plain wrong.  You end up 5-6% to the Dem side every time due to the married/single thing.

Back to Pennsylvania...

It's close... there is a long way to go...

We will see..

I have the state tied, but if you want to call it a modest Kerry lead, hey I am ok with that too...

[TeePee4Prez]

Good analysis.  Still caliing it 4-6%, but we can compromise on 1% for now.  I'm just giving you a Southeastern PA perspective on this.  Bush is getting creamed here.

[The Vorlon]

Good analysis.  Still caliing it 4-6%, but we can compromise on 1% for now.  I'm just giving you a Southeastern PA perspective on this.  Bush is getting creamed here.

Yes, but Bush got killed their last time too..

On more factor you may wish to include (this helps your case for a bigger Kerry lead) is that there has been a modest net migration (of modestly GOP leaning voters) out of Philly over the border into the New Jersey / Phillyland Suburbs.  This explains partly why Bush is getting killed marginally kess badly in New Jersey this time around.

Link to Fox New Jersey poll..?

[lonestar]

I haven't seen this NJ poll he speaks of either.

[millwx]

I haven't seen this NJ poll he speaks of either.

I searched.  The only thing I could find was a Fox/OD poll from long, LONG ago (mid-2002!!) merely asking a "re-elect/don't re-elect" question.  In that poll Bush's "re-elect" was higher than his "don't re-elect".  I can't for the life of me believe that this is the poll in question.