A quick DEFENSE of the FoxNews/OD poll...

[The Vorlon]

Don't worry... I will be posting another new thread pointing out the flaws of these polls, (National sample has a few glitches, still waiting for data on the states to get in) but I have to start some where....

You can "flame" this quick take yourselves... or wait for me to do it .... (just be kind... ok..? No Trolls please!)

---actually, on second though... please "flame away" then I can just "cut and paste" to build the new thread

I am going to start on the "counter thread" entitled "A quick ATTACK on the Fox Polls" some time soon too... ...

The National Poll...

Step 1 - Apples to Apples.. in the last 6 polls the only "type" of poll they all share is a poll with Nader.  I prefer the head to head, but 3 way will do...

Fox has Bush +7
Harris has Bush +6 (or +10 if you want to use "likely" voters)
Pew has Bush +4
Tipp has Bush +3
Gallup has Bush +1
Washington Post has Kerry +4

Pew also notes a strong shift to Bush on Day three of their poll when Reagan dies - if they has started 2 days later they would have shown Bush +6 or so also...

TIPP a bit lower, easily explained, they don't push the leaners, hence lower numbers overall.

Gallup, below the other 4, but still within MOE.

Lets use the old rule of thumb when a batch of polls conflict - toss out the highest and the lowest, and average the rest.

Fox has Bush +7
Harris has Bush +6
Pew has Bush +4
Tipp has Bush +3
Gallup has Bush +1
Washington Post has Kerry +4

Average is Bush +3.5% of the other 4.

Fox is the most recent, the economy is improving, and maybe the beheading of the Korean and American has certainly shocked a lot of Americans.

In every single poll, Bush's best issue is Terrorism - Bringing terrorism to the top of mind (as the news certainly has the last few days) should thus help Bush as it puts the focus on his best issue.

The Fox poll is 3.5% higher than the average, and may be factoring the above argument as it is more recent.

Is it a bit high? - Likely - is it Crazy? - No.

Michigan:

Step #1 - adjust what you think is "about right" up 4% from last month!

Last month the average of the national polls had Kerry a bit ahead.

7 of the last 8 published polls (Add Rasmussen at (today anyway) Bush +2, and the NPR poll with Bush +1 into the mix) have had Bush ahead from anywhere from 1% to 10%.

The average of these 7 polls is Bush +3.7%, lets say that is crazy, and say it is really +2 - That is 4% better than last month when Kerry was up an average of 2%

Last month Kerry was up a couple points, now Bush is up a couple points.

Back to Michigan...

In the Head to head ballot, Fox has Kerry up 1%

The last four polls prior to Fox in Michigan:

Epic/MRI - Kerry +2
Survey USA - Kerry +4
Rasmussen - Kerry +6
Detroit News - Bush +4

Average of these 4 polls is Kerry +1.5

Lets "adjust" this average a bit now.
Rasmussen has moved Michigan to "tossup" (ie 4% or less) and let's also lets toss out the Detroit News (Bush +4 - just silly), and replace the +6 robo-rasmussen with a +4 updated Robo Rasmussen.

Epic/MRI - Kerry +2
Survey USA - Kerry +4
Rasmussen - Kerry +4 (Updated electoral projection)

Average - Kerry + 3.33%

Now regardless of if we use the average of Kerry + 3.33% or Kerry + 1.5%, the Fox number of Kerry +1 is not even close to being "way out"

It only represent a shift of 2 or 3%

If we factor in a 3-4%% national shift to Bush, and add 3% to the average of the May Michigan polls, Bush being down 1% is not at all crazy, and looks very, very rational actually.

Ohio

Bush up 4% in Fox Poll

POS says 3%
Rasmussen says 2%
Mason Dixon says 7%

LA Times I pitch out...
ARG has Kerry +7 a month ago, a poll immediately repudiated by polls from POS/Rasmussen/Mason Dixon.

Remember ARG does market research for consumer products as their main business.  Determining if people prefer "spingtime fresh" to "touchably soft" is rather different than sorting out a choppy electorate.  

Remember too, ARG has NEVER had Bush ahead nationally - this poll runs cold for Bush, so I will toss this one too..

Is +4 crazy...?

If you look at:

POS says 3%
Rasmussen says 2%
Mason Dixon says 7%

+4 is NOT crazy, especially if you factor in the fact than 7 of the 8 most recent national polls have Bush up, versus May when the majority of polls had Kerry up.

Florida

First, Florida is a bitch to poll... I have commented on this several times.

1) Huge non-citizen population
2) Overseas military who have Florida as "home" (no state income tax)
3) Snowbirds (Where do they "live" for voting purposes)
4) Felons do not count
5) Language issues

For this reason, let's toss the ARG, a poll which runs cold for Bush and is a less than "top tier" firm...

What do we have left from QUALITY (or at least decent) firms..?

An old mason Dixon had Bush +7
POS has Bush +5
Snell Lake Perry has Bush +4
Survey USA has Bush +7

Average Bush + 5.75
Kick out the old M/D Bush +7

Taking just the last 3 has Bush + 5.33%

Factor in a bit of a bump from Bush being up a bit nationally versus down a bit last month, (say 3 or 4%?) and +9% (head to head) is maybe a point ot two high, but again NOT crazy.

Pensylvania...

With the sole exception of Today's Quinipiac, (and a 400 sample size Uni-poll which I toss on principle) the last polls have all had this state within MOE of even.

Survey USA - Kerry +1
Rasmussen - Bush +1
Pew - Even

Fox has Bush up 3% head to head - Also within MOE of even.

Take "even" as a baseline in May, factor in the 3-4% National bump in the polls Bush has enjoyed, and Bush +3 is entirely sane...

Flame away!

[CARLHAYDEN]

A very thoughful analysis.

Some polls are just unbelievable.

I noted that the Harris poll showing Bush up 10 per cent was unbelieveable, and the ABC poll showing him down several points was also not credible.

I believe that the Fox poll (on its face) is at the upper end of credibility and the Gallup Poll at the lower end of credibility.

Also, I must question any poll which shows the undecided vote below seven per cent.  Too many polls are pushing respondents too hard.

I will be interested in seeing the internals of the Opinion Dynamics poll.  Think the internals of the Pew poll was the best public poll I have seen recently.

[The Vorlon]

A very thoughful analysis.

Some polls are just unbelievable.

I noted that the Harris poll showing Bush up 10 per cent was unbelieveable, and the ABC poll showing him down several points was also not credible.

I believe that the Fox poll (on its face) is at the upper end of credibility and the Gallup Poll at the lower end of credibility.

Also, I must question any poll which shows the undecided vote below seven per cent.  Too many polls are pushing respondents too hard.

I will be interested in seeing the internals of the Opinion Dynamics poll.  Think the internals of the Pew poll was the best public poll I have seen recently.

If you are doing a "likely" voter poll (asuming you know what you are doing in the voter screen) I agree with pushing the leaners - People with a strong history of voting in the past will likely vote in the future so their opinion counts, even if it is just a "lean"

I also agree that pushing the leaners in a REGISTERED voter poll (where 1/3 of the sample will not vote) produces jibberish.

You get the slant of the news coverage in the last 24 hours, it has nothing to do with a "real" result.

[CARLHAYDEN]

You have a reasonable point except for the way in which 'likely voters' are determined.

At this point out in the campaign, if you rely on respondents stating how closely they are following the campaign to establish your likely voter screen, then I think pushing the voters is very defective.

If you were to use past voting habits as a screen, then I don't believe there would be a major problem with pushing the respondents to make a choice.

I guess this is one example of while polling is as much an art as a science.

[Smash255]

Their are problems with pushing the undecideds too hard, but their are also problems wwith barley pushing them at all (TIPP) especially in incumbent elections when undecideds tend to break heavily for the challenger.

[ATFFL]

Its best not to push them and mention that undecideds usually break for the challenger, but not always.  The trick is getting them to mention that.  Last thing you want to do is push them and then have them all stay home.

I prefer when they push the undecideds, but don;t figure that in the main number.  Say that there were 10% undecided, when pushed they would add 2% to Bush, 4% to Kerry 1% to Nader and 3% are truly undecided.

[○∙◄☻¥tπ[╪AV┼cVê└]

ARG has Kerry +1 in FL. I'm not saying that Kerry is leading in FL, but I think that Bush +10 is way off.  Looking at all of the recent FL polls here, FL remains a complete tossup.
http://www.realclearpolitics.com/bush_vs_kerry_sbys.html

[millwx]

Excellent post Vorlon.  I will disagree a bit, though...

1) While you make a good argument for all of the Fox/OD polls being "sane", they do almost all run on the high-ish side (except OH... I think that one's running right on).  Yes it's all within the MOE, but it just seems unlikely for a poll to err on the same side of the MOE four out of five times (and have the 5th time not be on the other side, but just dead-on in the middle).

2) As the data on Bush's favorability ratings shows (I posted it in the "Fox/OD sub-polls" thread) Fox/OD is clearly Bush-friendly.  I won't try to argue that they're doing anything sinister.  You know their methodology better than I and believe it to be A+.  I also don't know how any issues with their favorability ratings would translate into horse-race polls.  However, I can only go on the data we have.  Fox/OD's favorability gap runs an average of 4-5% pro-Bush compared to the poll median.  Now, they're running 3.5% pro-Bush compared to the national poll mean you posted here?  This is WAY too coincidental to be a coincidence, if ya know what I mean.

Given the typical MOEs in these polls, point #2 plays right into the hands of point #1 and allows someone to think the Fox/OD polls are "sane".  That is, their apparent bias is about the same size as the MOE.  So, rarely will they look outrageously bad, but they will always (well, I'm sure not ALWAYS, but almost) err on the side of Bush.  I'd take the Fox/OD polls, state and national, and shift them 3% (even that may not be quite enough, but I'll give them the benefit of the doubt, given their presumably excellent methodologies).

[ATFFL]

If what you are saying were true, millwx, then Fox would always have Bush up a few points on the national average.  So let’s look at some Fox/OD polls and the other polls that were conducted at the same time.

We’ll start with head to head.  All data is from http://www.realclearpolitics.com/bush_vs_kerry_hth.html

IBD/TIPP (6/8-6/13)           Bush +1
Pew Research (6/3-6/13)  Bush +2
NPR/GQR (6/6-6/10)          Bush +1
Fox News (6/8-6/9)          Kerry +2

Fox was the only poll conducted on those  dates to have Kerry up head to head.

Quinnipiac (5/18-24)              TIE
CNN/Gallup/USAT (5/21-23)   Kerry +2
WP/ABC News (5/20-23)       Kerry +2
CBS News (5/20-23)             Kerry +8
Fox News (5/18-19)              TIE
IBD/TIPP (5/12-18)                Bush +1


This is a tough one because few polls have the same dates.  We can safely assume the CBS poll is an outlier.  Fox is either a point biased to Bush or picked up the time when things really were dead even.

Pew Center (5/3-5/9)           Kerry +5
Ayres McHenry (5/3-5/6)*     Bush +3
IBD/TIPP (5/2-8)                   Bush +3
ARG (5/3-5/6)                       Kerry +3
Fox/OD (5/4-5/5)                 Bush +3
CNN/Gallup/USAT (5/2-4)     Kerry +1
NBC/WSJ (5/1-5/3)              Bush +3

Fox does have Bush at the high end of the spectrum, but a lot of polls agree with them (* including a Republican polling firm.)

Let’s do one more.

CBS News/NYT (4/23-27)          Kerry +2
Rasmussen (4/22 - 4/24)          Bush +3
Marist (4/20-23)                        Bush +3
Fox News/Op Dyn (4/21-22)     Bush +1
Democracy Corps (4/19-22)**  Bush +2

Fox/OD comes in a hair below the average of the others (** including a Democratic polling firm.)

While there may well be a bias toward Bush in the individual approval questions, it does not show up in the head to head poll question.  At least, not until this poll.

[The Vorlon]

Nice post Tredrick

IBD/TIPP (6/8-6/13)           Bush +1
Pew Research (6/3-6/13)  Bush +2
NPR/GQR (6/6-6/10)          Bush +1
Fox News (6/8-6/9)          Kerry +2

Fox 3.33% more pro-Kerry than average of other 3 polls

Quinnipiac (5/18-24)              TIE
CNN/Gallup/USAT (5/21-23)   Kerry +2
WP/ABC News (5/20-23)       Kerry +2
CBS News (5/20-23)             Kerry +8
Fox News (5/18-19)              TIE
IBD/TIPP (5/12-18)                Bush +1

Fox 0.75% more pro-Bush than average of other 4

Pew Center (5/3-5/9)           Kerry +5
Ayres McHenry (5/3-5/6)*     Bush +3
IBD/TIPP (5/2-8)                   Bush +3
ARG (5/3-5/6)                       Kerry +3
Fox/OD (5/4-5/5)                 Bush +3
CNN/Gallup/USAT (5/2-4)     Kerry +1
NBC/WSJ (5/1-5/3)              Bush +3

Even if we leave in the Pew Kerry +5

Fox 2.00% more pro-Bush than average of other 4

CBS News/NYT (4/23-27)          Kerry +2
Rasmussen (4/22 - 4/24)          Bush +3
Marist (4/20-23)                        Bush +3
Fox News/Op Dyn (4/21-22)     Bush +1
Democracy Corps (4/19-22)**  Bush +2

Fox 0.50% more pro-Kerry than average of other 4

Summary:

Fox 3.33% more pro-Kerry than average of other 3 polls

Fox 0.75% more pro-Bush than average of other 4

Fox 2.00% more pro-Bush than average of other 4

Fox 0.50% more pro-Kerry than average of other 4

Twice above the average of the others, twice below...

On average (0.41%) more pro-kerry than the average of the others...

I personally do not see a bias...

Their current poll ia about 3% above the national average, again not bizarre, and again not part of a pattern... They were once 2% below the average too...

Unless, of course, the 4 reasonable polls before this one were all part of an elaborate VRWC plan to build credibility before shamelessly fudging the data in a pro-bush way... :

Last poll "cycle" Gallip had Kerry +6, very much more pro-kerry than the average of the time..

Virtually nobody said Gallup had blown up, or was biased, or gone nuts - they had a half blip sample is all.

Fox has one poll of their last 7 or 8 that is a few points off.  They are a good firm, - I am cutting them some slack.





Millwx..

My memory is a bit fuzzy sometimes, but if I recall, your own analysis of approval ratings over an 18 month period showed that every single firms, on average, showed:

An average Bush approval rating of either 58% (Pew) or 59% (everybody else including Fox)

The only exception was Harris which showed 56%.  

Harris asks a fairly differently worded question than the others, so their result is not a direct comparable and can be discarded for comparison purposes.

Do I recall your own data correctly?

The reason everybody shows almost the exact same total approval, but different NET (Approve - Disapprove) is in how hard you push the undecided.

Non-partisan people don't like to say bad things about their President.  If given a "don't know/Undecided" option, a good chunk who actually mildly dis-approve will say "no opinion"

Fox does not push folks on Presidential approval (or anything else for that matter) hence a higher undecided/don't know and fewer "soft" disapprovals.

It is not "bias" per se, they just ask and push the question differently..

Fox has made a choice not to push the undecided, others have made a different choice...

Its that simple

[millwx]

Fox does not push folks on Presidential approval (or anything else for that matter) hence a higher undecided/don't know and fewer "soft" disapprovals.

It is not "bias" per se, they just ask and push the question differently..

Fox has made a choice not to push the undecided, others have made a different choice...

Its that simple

No.  By not pushing the undecideds the favorability gap should be lower... assuming those undecideds would "break even" (which MAY not be a safe assumption, but I can't see no reason to assume otherwise).  So, by that very fact, Fox/OD contains even MORE bias than their favorability gap shows.

And as for the raw favorability numbers, yes, there were one or two poll averages up there with Fox/OD.  But, Fox/OD was still right at the top.  Moreover, after going back further another six months, Fox/OD rose further.

So, the only way that "not pushing undecideds" is the answer is if nearly 100% of those who would decide when pushed (as in the other polls), came down on the unfavorable side.  And even then, like I said, Fox/OD remains on the "fringe" high end in the raw favorables.

[The Vorlon]

And as for the raw favorability numbers, yes, there were one or two poll averages up there with Fox/OD.  But, Fox/OD was still right at the top.

Actually:

In fact, let's just jump back a ways and randomly select a longer-term period (so you can see I'm not cherry picking) and average their polls.  For all of 2003 (averaging the polls per month, so multi-polled months don't skew the data too much), the average approval rating for Bush...

Fox/OD = 59%
Newsweek = 56%
Time/CNN = 56%
Zogby = 54%
CBS = 59%
Gallup = 59%
Pew = 58%

Re-arranging the Data:

Gallup = 59%
Fox/OD = 59%
CBS = 59%
Pew = 58%

Newsweek = 56%
Time/CNN = 56%

Zogby = 54%

Three at 59%
One at 58%

TimeCNNNewsweek at 56%

Zogby = Different Question = Doesn't count/not comparable

[The Vorlon]

I believe that the Fox poll (on its face) is at the upper end of credibility and the Gallup Poll at the lower end of credibility.

Yup.. - Fox is MOE high, Gallup is MOE low

The polls have actually been more in agreement this year than is typical.

The old rule of thumb..

"Toss out the low and the high - average the rest"

works pretty well.

Bush up a very few points is about right IMHO...

[© tweed]

This week has shown me that Kerry can queak out a win in the EV if he is down over 3% nationally

[millwx]

Gallup = 59%
Fox/OD = 59%
CBS = 59%
Pew = 58%
Newsweek = 56%
Time/CNN = 56%

Thanks for posting this old thread info.  While I suspect you disagree , since you took the time to find/post this, I think this proves my point.  Fox/OD is right up with the top numbers in terms of favorability... up with Gallup and CBS.

So what??  The other three are barely lower, so what's the big deal??  Well...

As I said, because of their large undecided pool, Fox/OD should be on the LOW end of this list.  In fact, they have the HIGHEST undecided pool and, so, all else being equal, they should be the LOWEST on this list.  They're just the opposite!

What's more, I only randomly selected 18 mos.  Go back further and it gets worse...  For the six months earlier...

Fox/OD = 68% Favorable
Gallup = 68%
CBS = 66%
Time/CNN = 65%
Newsweek = 64%
Pew = 63%

Only Gallup holds up there this time.  And even they were fractionally lower (rounding made them even).  And Fox/OD continued to have an average of DOUBLE the undecideds from Gallup.

And further proving my point that undecideds should drive the favorables (and unfavorables) down is Pew.  They are the only other one with an average of double-digit undecideds, and they came in at a low 63% favorable.  Moreover, two of the three other lower end ones, Newsweek and CBS, had relatively high undecideds compared to Gallup, but less than Pew (or Fox/OD).

In short, after accounting for the undecideds, all of these polls are very neatly and tightly clustered, with one exception... Fox/OD.  Dividing by the percent decided, all of the others are within 1.5% of 69% favorability (for the six additional months), except Fox/OD; dividing by the percent decided, they are at about 76.5%.

[ATFFL]

Vorlon already pointed out, the majority of undecideds to break to the negative side.  If you can't say something nice and all that.

Even if they do tend to poll more favorably (which is possible) it is not having an effect on the head to head horserace.

[Smash255]

If what you are saying were true, millwx, then Fox would always have Bush up a few points on the national average.  So let’s look at some Fox/OD polls and the other polls that were conducted at the same time.

We’ll start with head to head.  All data is from http://www.realclearpolitics.com/bush_vs_kerry_hth.html

IBD/TIPP (6/8-6/13)           Bush +1
Pew Research (6/3-6/13)  Bush +2
NPR/GQR (6/6-6/10)          Bush +1
Fox News (6/8-6/9)          Kerry +2

Fox was the only poll conducted on those  dates to have Kerry up head to head.

Quinnipiac (5/18-24)              TIE
CNN/Gallup/USAT (5/21-23)   Kerry +2
WP/ABC News (5/20-23)       Kerry +2
CBS News (5/20-23)             Kerry +8
Fox News (5/18-19)              TIE
IBD/TIPP (5/12-18)                Bush +1


This is a tough one because few polls have the same dates.  We can safely assume the CBS poll is an outlier.  Fox is either a point biased to Bush or picked up the time when things really were dead even.

Pew Center (5/3-5/9)           Kerry +5
Ayres McHenry (5/3-5/6)*     Bush +3
IBD/TIPP (5/2-8)                   Bush +3
ARG (5/3-5/6)                       Kerry +3
Fox/OD (5/4-5/5)                 Bush +3
CNN/Gallup/USAT (5/2-4)     Kerry +1
NBC/WSJ (5/1-5/3)              Bush +3

Fox does have Bush at the high end of the spectrum, but a lot of polls agree with them (* including a Republican polling firm.)

Let’s do one more.

CBS News/NYT (4/23-27)          Kerry +2
Rasmussen (4/22 - 4/24)          Bush +3
Marist (4/20-23)                        Bush +3
Fox News/Op Dyn (4/21-22)     Bush +1
Democracy Corps (4/19-22)**  Bush +2

Fox/OD comes in a hair below the average of the others (** including a Democratic polling firm.)

While there may well be a bias toward Bush in the individual approval questions, it does not show up in the head to head poll question.  At least, not until this poll.

Of course it depends on which polls you show, and in trying to show Fox as correct you takee a poll a couple days before the Fox poll on one grouping that shows it close to the Fox poll, but you take out another poll that goes against the Fox Poll.  

Gallup 6/3-6/6  Kerry up +6

Anyway the biggest reason in the poll Fox was the most Kerry favorable polls in the ones you mentioned was probably the Reagan factor bump that boosted the other polls, that Fox didn't have the same bump  because the bump seemed to come from the middle to end of the week

[ATFFL]

Of course it depends on which polls you show, and in trying to show Fox as correct you takee a poll a couple days before the Fox poll on one grouping that shows it close to the Fox poll, but you take out another poll that goes against the Fox Poll.  

Gallup 6/3-6/6  Kerry up +6

Anyway the biggest reason in the poll Fox was the most Kerry favorable polls in the ones you mentioned was probably the Reagan factor bump that boosted the other polls, that Fox didn't have the same bump  because the bump seemed to come from the middle to end of the week

I said from the beginning that I would be looking at polls on the same days, and broke thatonly when there were too few polls available on the same days.  But I will satisfy you by bringing in the three previous polls which ended a few days earlier.

IBD/TIPP (6/8-6/13)          Bush +1
Pew Research (6/3-6/13)  Bush +2
NPR/GQR (6/6-6/10)          Bush +1
Fox News (6/8-6/9)          Kerry +2
LA Times (6/5 - 6/8)         Kerry +7
Gallup (6/3-6)                  Kerry +6
IBD/TIPP (6/1-6)              Bush +1

We all know about the LA Times poll.  Looking at all that polling data which one do you think is the outlier?  The Fox one or the Gallup?

In my four groupings you can challenge only one, and that one is clearly either outside the MoE, or pushing the top with Fox pushing the bottom.

The NPR poll should have also missed any effects of the Reagan bounce.

Sorry, but there is no evidence of a systemic bias or error in the Fox News/Opinion Dynamics poll.  While it is likely that this poll is off a bit in Bush's favor, Fox even mentioned the possibility last night, it is not from a built in bias designed to slant the news (i.e. LA or NY Times.)

[Smash255]

Of course it depends on which polls you show, and in trying to show Fox as correct you takee a poll a couple days before the Fox poll on one grouping that shows it close to the Fox poll, but you take out another poll that goes against the Fox Poll.  

Gallup 6/3-6/6  Kerry up +6

Anyway the biggest reason in the poll Fox was the most Kerry favorable polls in the ones you mentioned was probably the Reagan factor bump that boosted the other polls, that Fox didn't have the same bump  because the bump seemed to come from the middle to end of the week

I said from the beginning that I would be looking at polls on the same days, and broke thatonly when there were too few polls available on the same days.  But I will satisfy you by bringing in the three previous polls which ended a few days earlier.

IBD/TIPP (6/8-6/13)          Bush +1
Pew Research (6/3-6/13)  Bush +2
NPR/GQR (6/6-6/10)          Bush +1
Fox News (6/8-6/9)          Kerry +2
LA Times (6/5 - 6/8)         Kerry +7
Gallup (6/3-6)                  Kerry +6
IBD/TIPP (6/1-6)              Bush +1

We all know about the LA Times poll.  Looking at all that polling data which one do you think is the outlier?  The Fox one or the Gallup?

In my four groupings you can challenge only one, and that one is clearly either outside the MoE, or pushing the top with Fox pushing the bottom.

The NPR poll should have also missed any effects of the Reagan bounce.

Sorry, but there is no evidence of a systemic bias or error in the Fox News/Opinion Dynamics poll.  While it is likely that this poll is off a bit in Bush's favor, Fox even mentioned the possibility last night, it is not from a built in bias designed to slant the news (i.e. LA or NY Times.)

I said their was some big errors with the latest poll, I didn't say the same errors existed in every fox poll, but I did find it interesting that in the polls with Nader Fox has NEVER had Kerry ahead amd that does lead me to believe their is ome inaccuracy.  Wheter its bias or not its hard to say Fox doesn't release their internals, but some things just don't add up especially with their last poll

[ATFFL]

I said their was some big errors with the latest poll, I didn't say the same errors existed in every fox poll, but I did find it interesting that in the polls with Nader Fox has NEVER had Kerry ahead amd that does lead me to believe their is ome inaccuracy.  Wheter its bias or not its hard to say Fox doesn't release their internals, but some things just don't add up especially with their last poll

Lets see if we can se a systemic bias in the 3 way race then.

IBD/TIPP (6/8-6/13)           Bush +3
Pew Research (6/3-6/13)  Bush +4
Ipsos-AP (6/7-6/9)             Bush +1
Fox News (6/8-6/9)           TIE
LA Times (6/5-8)                Kerry +6
Gallup (6/3-6)                    Kerry +6
TIPP/IBD (6/1-6)                Bush +2

Throwing out the tiems poll, because it is the Tiems poll, we are left with Gallup as a likely outlier at +6%.  Other than that Fox/OD has Kerry higher than anyone.


Quinnipiac (5/18-24)                     Bush +1
CNN/Gallup/USAT (5/21-23)          Kerry +1
Insider Advantage (5/21-22)        TIE
Wash Post/ABC News (5/20-23)   TIE
CBS News (5/20-23)                     Kerry +6
Fox News (5/18-19)                     TIE
TIPP/IBD (5/12-18)                       Bush +1
Newsweek (5/13-14)                   Kerry +1

Here we have an obvious outlier in CBS, and every other poll being a tie or a 1 point advantage.  Fox/OD seems to be right on.

CBS News/NYT (4/23-27)              Bush +2
Marist (4/20-23)                           Bush +4
Fox News/Op Dyn (4/21-22)        Bush +2
Democracy Corps (4/19-22)**     Bush +3
TIPP/IBD (4/14-4/19)                   Bush +4
CNN/Gallup/USAT (4/16-4/18)      Bush +6
ABC News/WP (4/15-4/18)          Bush +5

Here Fox/OD is the lowest poll of hte batch for Bush, tied withthe CBS/NYT poll, who seem to usually overpoll Kerry.  

Yes, the clear bias in the Fox News Opinion Dynamics poll is clear to me now.  They have either been right on or underpolled Bush at the time of the poll.

[millwx]

Vorlon already pointed out, the majority of undecideds to break to the negative side.  If you can't say something nice and all that.

Problem is, since Fox/OD's straight-up favorables run high themselves, the undecideds would need to break almost 100% unfavorable in order for this logic to work out.  Extraordinarily unlikely.  Sorry, but I'm very hard pressed to buy into Fox/OD.  Even the head-to-heads you posted are unconvincing, since it's a mere handful of polls (yes, you showed more than one cluster, but in a couple of them Fox was, indeed, on the high side with Bush... so, it is a mere handful that support your premise), while the data I'm examining is literally dozens of polls.

Let me be clear... I'm not saying Fox/OD is wretched.  In all, they seem to be off by about a few percent.  I'm also not being selectively partisan... I recognize the wretchedly flawed nature of the LA Times poll.  I also recognize the "less sinister" and far less extreme, but still flawed sample in the recent ABC/WP poll.  Plus, I also have no problem with overtly partisan pollsters from both sides (like POS or Democracy Corps) as long as they do the job right.  I've simply looked at the numbers... taking something statistically significant (i.e., not just a handful of polls)... and Fox/OD is "off", period... not off a lot, but, nonetheless, off.

I would also add that if undecideds are going to break so harshly in one direction (though there is no one-to-one correlation in the direction they'll break between favorability polls and head-to-head matchup polls... but almost certainly, there'll be some relation), these polls with high undecideds are almost useless.  Since the undecideds are not expected to break either evenly or with the same proportions as the decideds, the higher the undecided totals the worse a poll will be.  Sure, if people are TRULY undecided, and it's an RV poll, and you force them into a choice, that's bad (The Vorlon has talked about these issues of pushing in the past).  So, I'm not saying you force everyone to make up their minds.  But, the point is, the undecideds are, by your argument (implied in the favorability data, which should have at least a weak correlation to the horse-race data) distributed non-randomly and rather inversely correlated to the decideds.  As such, the higher the undecided total, the worse the results.  It's no coincidence that GENERALLY (certainly not 100% of the time - like with the recent Harris Poll) the polls with the lower undecideds poll Kerry higher.  Look at the burst of five that just came out... Fox/OD 8% undecided => Bush +6; Gallup 3% undecided => Kerry +4; ABC/WP 0% undecided => Kerry +8 (this one had a 5% sampling error, roughly... so this is, in reality, about Kerry +3); TIPP 13% undecided => tied; Democracy Corps 1% undecided => Kerry +1.  The two with undecideds over 5% have it tied and Bush +6.  The three with undecideds under 5% all have Kerry up.

High undecideds may protect you from "spoiling" your sample, but they may also cause you to seriously miss "real" data.  Perhaps it is philisophical.  And maybe that is 100% all that's going on here... pushing versus non-pushing... though the favorability data implies at least something SLIGHTLY more than that.  In which case, my harping on Fox/OD is unwarranted and I merely have an honest disagreement with the whole "push or don't push" issue.  Clearly, I prefer pushing... I think there's real data there, and I think recent polls (and historical data) bear that out.  So, I am more than willing and happy to concede that Fox/OD is not remotely biased.  Still, as I believe in the "pushing" philosophy, I don't agree with their methodology and I'll take Fox/OD with a grain of salt.

[The Vorlon]

Regarding Favorability ratings...

Also go back and factor in the following:

Register votes vs Adult population

Different populations => different answers

(Time and Newsweek are usually All adults vs Registered voters)

Also, factor in the order the question resides within the questionaire - is it the first question, or the last, or something in between - does if follow issue questions, or preceed them?

Just a few more monkey wrenches to toss into the debate...

[Smash255]

I said their was some big errors with the latest poll, I didn't say the same errors existed in every fox poll, but I did find it interesting that in the polls with Nader Fox has NEVER had Kerry ahead amd that does lead me to believe their is ome inaccuracy.  Wheter its bias or not its hard to say Fox doesn't release their internals, but some things just don't add up especially with their last poll

Lets see if we can se a systemic bias in the 3 way race then.

IBD/TIPP (6/8-6/13)           Bush +3
Pew Research (6/3-6/13)  Bush +4
Ipsos-AP (6/7-6/9)             Bush +1
Fox News (6/8-6/9)           TIE
LA Times (6/5-8)                Kerry +6
Gallup (6/3-6)                    Kerry +6
TIPP/IBD (6/1-6)                Bush +2

Throwing out the tiems poll, because it is the Tiems poll, we are left with Gallup as a likely outlier at +6%.  Other than that Fox/OD has Kerry higher than anyone.


Quinnipiac (5/18-24)                     Bush +1
CNN/Gallup/USAT (5/21-23)          Kerry +1
Insider Advantage (5/21-22)        TIE
Wash Post/ABC News (5/20-23)   TIE
CBS News (5/20-23)                     Kerry +6
Fox News (5/18-19)                     TIE
TIPP/IBD (5/12-18)                       Bush +1
Newsweek (5/13-14)                   Kerry +1

Here we have an obvious outlier in CBS, and every other poll being a tie or a 1 point advantage.  Fox/OD seems to be right on.

CBS News/NYT (4/23-27)              Bush +2
Marist (4/20-23)                           Bush +4
Fox News/Op Dyn (4/21-22)        Bush +2
Democracy Corps (4/19-22)**     Bush +3
TIPP/IBD (4/14-4/19)                   Bush +4
CNN/Gallup/USAT (4/16-4/18)      Bush +6
ABC News/WP (4/15-4/18)          Bush +5

Here Fox/OD is the lowest poll of hte batch for Bush, tied withthe CBS/NYT poll, who seem to usually overpoll Kerry.  

Yes, the clear bias in the Fox News Opinion Dynamics poll is clear to me now.  They have either been right on or underpolled Bush at the time of the poll.

The fact remains They are the ONLY Major pollster with more than a couple polls not to have Kerry ahead in the 3 way at any point.  Regardless even if their past polls haven't been far off their is some serious issues with this poll.  As I pointed out earlier Bush is up 43-42 in the "Blue States" in this poll.  That in itself shows this particular poll has serious issues.  If another poll showed Kerry leading in the red states you would be having a heart attack & be screaming bias.

[The Vorlon]

High undecideds may protect you from "spoiling" your sample, but they may also cause you to seriously miss "real" data.  Perhaps it is philisophical.  And maybe that is 100% all that's going on here... pushing versus non-pushing... though the favorability data implies at least something SLIGHTLY more than that.  In which case, my harping on Fox/OD is unwarranted and I merely have an honest disagreement with the whole "push or don't push" issue.  Clearly, I prefer pushing... I think there's real data there, and I think recent polls (and historical data) bear that out.  So, I am more than willing and happy to concede that Fox/OD is not remotely biased.  Still, as I believe in the "pushing" philosophy, I don't agree with their methodology and I'll take Fox/OD with a grain of salt.

For what it is worth, here is my opinion on pushing the leaners:

In a likely voter poll, where you have done a proper voter screen based on past voting behavior AND current level of interest...

Push the heck out of them... beat them with an ugly stick till you get an answer....

In a registered voters poll,  offer them a "don't know" option (or better yet, use an "unaided ballot") and don't bother pushing them, because you will get gibberish.

As an aside, what you and Tredrick are REALLY debating is if you should use a "likely" voter poll, or a registered voter poll.  

If you take the argument to it's logical end, what Millwx comes down to is you want to extract all the data you can from a sample, even if it is "soft" data, while Tredrick argues that 45% of the sample doesn't vote, so their opinion does not count...so why beat an answer out of folks who won't vote anyway...

You are both correct, and the only way to reconcile these opposite but both correct positions is to really really screen well (none of these bullsh*t 3 question likely voter screens) and then after you have boiled it down to the 55ish % who will actually vote you beat the stuffings out of then for every last ounce of information..

Just my opinion

[ATFFL]

That about sums it up for me, Vorlon.  THough I do favor pushing undecideds later in the campaign season and having a seperate listing of how they break.

My ideal is to have no undecided option, make the respondednt come up with that on their own.  Then push and describe how they break seperately.  

Undecided means undecided, especially this early in the campaign.  Most people are not engaged in the campaign right now.

Good job, Vorlon.  I'll have the UFO drop off a cookie tonight.