Redstate

I am not happy with these news but it's quite a stretch to call the LA Times poll heavily in favor of the Dems. Polls bouce all over. Rasmussen's poll, which had Bush up by 9 in Election Day 2000, and is updated daily, has them tied and Bush has not led in weeks.

But it's hard to deny the Swift ads had an impact.

I'm saddened that the SVFT has had an impact.  I wish the negative attacks on both sides were ignored by the public.  As long as they seem effective, they will continue with or without approval from the elected leaders.

of politics.

And I know both sides do it.

But I do think these attacks, particularly on the medals, were completely over the top. I understand the anger on Kerry's 1971 declarations and I think the second ad is a plausibl ne, even if I disagree. But not the rest.

I agree with you on the SVFT being over the top on their first ads and the second ones being more understandable.  However, I don't think they are as bad as Fahrenheit 9/11 and they are on par with some of the early MoveOn.org ads.  I don't like it on either side, but I think both sides have gone over the top in this cycle.

so I can't compare directly.

But the difference, in my mind, is that the Swift ads has accusations supposedly backed by affidavits from supposedly direct witnesses. Many people will miss that the affidavits don't amount to much and that all the official documents and that a growing number of witnesses contradict the Swift allegations.

F911, on the other hand, is just a movie. There are no affidavits or anything like that. You have to go to a movie theater to see it, a place most people go to watch fiction.

Again, I don't know how much f911 twists the truth or even outright lies. Buit I do think it is presented in a very diferent way than the Swift ads were.

It seems we are seeing a bit of a backlash against the Swift ads, with O'Neill openly called a liar by the AP. So maybe there is hope (for me)!

I'm basing that assessment of the Times poll based on their record of usually being a statistical outlier for the Democrats. Historically the poll has shown higher support for Democrats than the actual results. It is entirely possible that they've changed their sampling methodology since then to be more accurate. Given that the Gallup Poll (which tends to be one of the most trustworthy) has similar results, I'd put good money on the Times poll being a good representation of the actual numbers at the moment.

What do you base it on?

Even the Pollkatz website, which had supposedly found that Fox News was more pro-Bush than the rest, had to admit that a correct statistical analysis dd not indicate that.

You may remember some LA Times outlier but ALL polling firms have outliers. It takes a statistical analysis to determine if there is something there.

You say that no President with an approval rating above 50% has ever lost reelection.

This is a bit misleading.

First the sample is pretty small since we have poll numbers only for about 12 elections.

But the more important point is that if you look at the data the cutoff point is not 50% but rather a band between 47% and 54% (working from memory so I may be off by a point). That is, the lowest approval rating a reelected president got was 54% or 53% while the highest a defeated incumbent got was 47%. Nobody knows where, exactly the cutoff point is.

I would also point out that the Economist poll, out today, while it shows the race tightening, has Kerry up by 3. And Zogby has a poll out two days ago with Kerry ahead in 14 of the 16 battleground states.

admitted they do not weight their sample according to party. If you don't do that, then why poll at all?

The LA Times historically polls in a Democrat direction. For instance, let's look at a subject the LA Times should hopefully know something about: California politics.

The LAT poll before the October 2003 recall election gave these results:

Yes on Recall 50 No--47

Schwartzenegger 30; Bustamante 25%; McClintock 18%

When the actual results were tabulated:

Yes on Recall 55 no--45

Schwartzenegger 49; Bustamante 32%; McClintock 14

So they polled the overall results in the Democrat direction by +7. The polled the candidates in the Democrat direction by +8.

the reason is that party identification is a very volatile measure.

And what you are offering is just one piece of data. To say that the LA Times or any other poll is biased you need to run a statistical analysis on a whole time series.  

I've worked as a pollster in the past and every single polling organization has polls that are outliers or elections that they call wrongly.

Using your logic of finding a couple of 'wrong' polls I can discredit the whole polling industry. Not a useful tactic, I'm afraid.

What was their MoE on the recall.  Since 3% were undecided it is probable that 2% went Yes and 1% went No.  Then the prediction was 52-48 within a 3% MoE.

The individual numbers leave 27% undecided.  Those broke to Schwatzenegger more than Bustamante but it is definitely within the MoE once undecideds are figured in (and they seem pretty high in this case which reflects the craziness of the recall).  The only number that could be evidence of being incorrect is the McClintock 18% turning out to be 14%.  Was the MoE 4% or higher, if so it is within the MoE.  Furthermore, it being a close election McClintock could have lost Repubs to Schwarzenegger in the last few days.

I don't see any evidence of bias there.

Link here. And there are several links in the story.

The LA times IS Dem biased and it has been proven.

If I were John Kerry, I would not be jumping up and down regarding the Zogby polls.  Most are within the margin of error and I don't believe that the majority of Americans have heard about the SBVT.

As far as the SBVT ads go, these are fine decorated Americans who have the right to speak their minds.  Where was the outrage about the millions spent by the LEFT on anti-Bush ads with hellacious accusations?

Kerry had nothing else to stand on..his Senate record is to the left of Stalin..so he drapes himself in the flag and thinks that will do it.  He is the one who was running on his Vietnam experience.

He disgraced America with his testimony, he committed treason by meeting the Viet Cong in Paris while still in the military and he gave comfort to the enemy who prolonged the war and kept our prisoners like John McCain incarcerated even longer!  Yeah...he's the man, right!!!  

The Dems can't stand the heat in the kitchen...get over it!!  What whiners!!!

Ted in Arizona

I don't know where you get your information but without footnotes it seems a bit on the exotic side.

Let's look at Gallup.

With Gallup we have approval ratings on Presidents going back to FDR. That is either 11 presidents, not 12. Those 12 presidents presided over 17 elections in which they were incumbents.

17 elections comprise a third of all presidential elections in the history of the republic, so it is a huge sample. Since we are so fond of the LA Times, they used 1597 adults to represent an electorate that exceeds 60 million. Something less than a third... or it was when I went to school anyway.

Now we get to the big breakdown. Your theorizing about a "band" may be right. But it isn't according to Gallup.

Roosevelt July 40-- 58%

Roosevelt no polling for 1944 by Gallup

Truman June 48-- 39%

Eisenhower August 56-- 67%

Johnson June 64-- 74%

Nixon June 72-- 57%

Ford June 76-- 45%

Carter August 80-- 32%

Reagan August 84-- 54%

Bush August 92-- 39%

Clinton August 96-- 53%

I agree that my original statement is a bit misleading, so I've modified it to be more accurate.

As for the Zogby polls, they're worthless as far as I'm concerned. Zogby's state polls have been extreme statistical outliers in nearly every race, and their national ones have tended to be just as bad. I personally believe Zogby's polling is significantly skewed towards the Democrats, and is not an accurate assessment of the state of the race.

The Economist seems to be more credible, but I've not reviewed their methodology, so I can't vouch for it.

Basic polling requires you weight your sample.

If you've worked as a pollster and you don't weight your sample by party affiliation, race, age, and income you aren't polling. You might as well be sacrificing a goat because you certainly aren't matching your sample with the population. And from a business standpoint you are cheating your client. I recommend a basic course in stats and a basic course in survey methodology.

Party affiliation (more R than D), race (blacks are least likely to vote), age (over 45 have very high rates of voting), and income (lower the income, lower voting propensity) are the key predictors of propensity to vote.

Telephone surveys have an inherent self-selection bias (who answers the phone, when do you make the calls). If you don't weight for the above factors, and I recommend weighting for education also, you have no idea what you are measuring.

And I'm not discrediting an industry. I'm saying that you just don't understand polling, that's all.

The second point I'm making is that to use polls you have to understand their methodology.

Nearly every poll that has been conducted recently shows Bush closing the gap.  The trend is clearly in his favor.  I fully expect him to open a 3-5% lead by the end of next week after the GOP convention ends.  The impact of the SBVT controversy could not have come at a worse time for Kerry as it, in conjunction with the GOP convention, will have effectively muzzled Kerry's message for 2-3 weeks.  

You weigh by measures that have a standard you can refer to. For example if you know, through the census, that blacks are 10% of the population you don't want sample with 5% or 20% blacks. Same with sex, age and income.

The problem is that there is no equivalent for party identification. Many states don't even allow you to register by party, for example. And people many times call themselves one thing even if they are registerd as another.

That's why many pollsters do not weigh by party.

May I ask that you tone down the rethoric? From your post (One word: footnotes) it seems you don't understand what a sample is or how it's drawn. So there is no need for insults.

was one of only two to correctly predict that Gore would win the popular vote in 2000.

That doesn't mean he will be right in the future though.

the recent trend is in Bush's favor.

I receive the Zogby poll survey online.  I believe this is what his firm uses when it releases its poll results. (Many of the questions he asks in the survey are posted in the results he sends out.)  Can anyone confirm this?  If so, how accurate can his poll results be if they rely on internet respondents?

You weigh by measures that have a standard you can refer to. For example if you know, through the census, that blacks are 10% of the population you don't want sample with 5% or 20% blacks. Same with sex, age and income.

You just don't understand weighting or survey methods.

It involves an arcane mathematical priciple called proportions. It also involves another mathematical principle called matrices.

The sample group you weight to is NOT the general population, it is the population that voted in the most recent elections.

The problem is that there is no equivalent for party identification. Many states don't even allow you to register by party, for example. And people many times call themselves one thing even if they are registerd as another. Booshwah and a non sequitur to boot. You ask the respondent for party identification. The technique, used pretty successfully by the Census Bureau is called self-identification.

State voter registration has nothing to do with it.

That's why many pollsters do not weigh by party.

Wrong yet again. They don't do it for one reason only: it is too expensive. You end up making a lot of calls that you can't use.

May I ask that you tone down the rethoric?

No.

it seems you don't understand what a sample is or how it's drawn.

Stop blowing smoke. You don't know what you are talking about and its really obvious. If you do poll I really hope you're doing it for Democrats. Give me the name of your agency and I'll recommend you to the locals.

you don't have a clue.

To the bozo file you go!

and I think we don't really know. He's not the only one doing it, though.

but you don't know what you don't know.

The fact is they consistently overrepresented Democrat strength on the main issue and the candidates by 7-8%.

That is outside their margin of error by nearly 200%.

This was a state-wide poll in their backyard.

Now we can agonize over undecideds until the cows come home but we know that most undecideds, especially in a poll of adults, stay home.

I don't know if they are push polling to create a story that matches their editorial bias, though I really suspect at least one national polling organization of cherry picking respondents. Personally I think they are just sloppy and use a database of phone numbers that have responded to previous surveys rather than paying for a fresh sample.

Let's say they asked 100 people who they were going to vote for and got the response:

Schwartzenegger 30

Bustamante 25

McClintock 18

Undecided 27

Then the vote happens... you can't have the same result as the poll.  If the 27% stays home, then the vote is:

Schwartzenegger 41% = (30/73)

Bustamante 34% = (25/73)

McClintock 25% = (18/73)

Compared to reality of

Schwartzenegger 49

Bustamante 32

McClintock 14

Thus, I have 3 main points.  First, assuming undecideds don't vote is untrue and means you're making an assumption not the LA Times.  Second, even if you assume they don't show up the poll gets Bustamante almost exactly right but overestimates the most conservative candidate, McClintock.  Third, the point is that one in twenty polls are outside the MoE and this is not even that one in twenty becuase undecideds do exist.  You can wish they didn't exist, but there's nothing pollsters can do to make people decide before they are ready and still be accurate.

Between two of my favorite people, too.

Believe it or not, I actually find this interesting.

I think that the LA slimes may be intentionally cooking the books in Bush's favor so that any 'bounce' from his convention will be lost in the statistical haze.