LA Times poll thread

[The Vorlon]

UPDATE

Article on LA Times "Poll"

"I feel sorry for her," said a pollster who asked not to be identified. "But she should have known better."

Article DEFENDING the LA Times

END UPDATE



The well respected Roll Call Magazine, and now "Drudge" are running articles attacking the LA Times "poll"

One more blow for the already battered media.

Link to Drudge Story

Link to Roll Call Story

[Sk]

    It's just unbelievable. Therefore, my new signature add-ons.

[WalterMitty]

the la times is a good newspaper.  their polling may suck, but their journalism is pretty solid.

[Mort from NewYawk]

Well of course the NY Times is the same way. You read it for in-depth coverage, but the placement of front page articles and size of headline fonts, and the choice of investigative articles is very biased, particularly in an election year.

[zachman]

I'll be honest, I've been in the LA Times newsroom and a lot of anti-Bush satire was hanging around. My favorite was the Ann Coulter Action Figure. The New York Times I think is pretty fair. There is a slight liberal bias, but there is a greater liberal bias throughout the New York area. Let's see, the op-ed contributors has only divided 4-3 or so. I don't consider that politically charged.

[Sk]

I'll be honest, I've been in the LA Times newsroom and a lot of anti-Bush satire was hanging around. My favorite was the Ann Coulter Action Figure. The New York Times I think is pretty fair. There is a slight liberal bias, but there is a greater liberal bias throughout the New York area. Let's see, the op-ed contributors has only divided 4-3 or so. I don't consider that politically charged.

    I never factor in the op-ed pieces when I judge the liberal bias of a paper. What I don't like is how the NYT has put Abu Garaib on its front page 47 (or so) times despite the fact that it has pretty much blown over news-wise. This is politics, pure and simple. Also, It really shouldn't matter where a paper is located. Every news source has a duty to its reader/viewers to provide them with honest, non-biased, and informative news. We need a non-partisan (not conservative) revolution in the journalism world.
    Anyway, thanks for that insight on the LA Times' newsroom. I'm not surprised!

[MODU]

Vorlon,

Found this interesting article regarding the LA poll which you might not have seen.  It's long, but it makes you question who really understands what they read on polls:

http://www.americanprogress.org/site/pp.asp?c=biJRJ8OVF&b=91077

[agcatter]

Wow.  That piece seems like a load of crap to me, but I look foward to Vorlen's take on it.

[MODU]

That was my basic summation as well.  It takes him almost the entire article to finally mention how many more Democrats were polled, or that they are registered voters and not likely voters.

[millwx]

http://www.americanprogress.org/site/pp.asp?c=biJRJ8OVF&b=91077

Very interesting article.  My own opinion is that this is a biased article (note that they also quote the Gallup and CBS results, but fail to mention the handful of polls showing Bush up by somewhere in the 1-3% range).  As such, the LA Times poll is still the outlier... so, I, personally, still don't buy it.

But there are a couple of good and pertinent points I think this article raises.  For example, whether you're a Rep or a Dem I think you'd find it reasonable to conclude that with Bush's favorability ratings for several months tending to be net-negative in most polls, odds are, any shift in party affiliation should be towards the Dems or away from the Reps.  I'm not trying to make a strong defense of the LA Times poll, as they seem WAY low on the Rep side.  But consider what most polls that do weight back to party affiliation use as their weights.  Then look at what the LA Times had.  The typical breakdown is something like 38D/35R/27I; LA Times had 38D/24R/38I.  This shows an 11% shift from R to I, using the LA Times numbers (and note they're not off on the D percentages).  Is that too much of a shift?  I definitely think so, yes.  So, I still contend that the poll is flawed.  But I could certainly conceive of a 5%-ish shift... especially the R->I shift instead of R->D or even I->D.

In short, this article and the LA Times poll still look biased and wrong to me.  But there were some good points raised and the LA Times poll may not be AS BAD as it looks at first glance.  Just my opinion.

[The Vorlon]

Wow.  That piece seems like a load of crap to me, but I look foward to Vorlen's take on it.

The LA Times polls are historically, structurally, so F^&%ed up, it is hard to know where to begin.

A few quick notes on party ID.

Firstly, both she and the author of the article acknowledge that 13% is high, their argument is that a combination of random error and a democratic surge make it high, but not absurd.

No where in the entire poll, anywhere in the news articles written, or anywhere in the publically available information, did they mention ANYWHERE that this poll contained 13% more Democrats than Republicans.

I am no news editor, but I think that was a relevant fact.  Especially when the result was totally at odds with a typical result.

They go into the many, many tiny details of the poll, the 3% change in Iraq approval, Kerry's 2%gain in "strong leader" etc.. etc...  The fact that Democrats outnumbers Republicans by a 1.58 to 1 margin slipped their minds...?  I think not.

She quotes ONE ABC news poll where the Dems were +10 in party ID. (which ABC weighted down in the reported result - another fact she forgot to mention)

This is a graphic of what ABC pollings says, as a trend, about party ID. (From pollingreport.com)



Doesn't look like an average of +10 to the Dems to me...

Gallup, over their last 40,000 interviews has found party ID to be within 0.3% of equality, PEW, over their last 15,000 interviews found the Dem's to have a 1.5% advantage.

The LARGEST democratic self identification advantage I could find over a large sample size was +5 from Harris

http://www.forrelease.com/D20040227/nyf126.P2.02272004181556.07849.html

Harris even notes that the reason they have a few more democrats is that they ask the question a bit differently..

Some other polls report that the Democrats and the Republicans are now virtually equal. We believe the small differences between their numbers and ours reflect the use of slightly different questions; the trends are very similar.

Regarding party ID, I guess you have to ask, are Gallup, ABC, Harris, and Pew all wrong and LA Times right, or the other way around.?"

I invite you to compare the polling record and reputation of the various organizations and make up your own mind

Structural Problems with LA Times Poll

They just simply draw their samples wrong - it is a fundemental flaw in the way they do things.

This is part of an email I got from the LA Times deputy Polling Director:

From: Richardson, Jill [mailto:Jill.Richardson@latimes.com]
Sent: Tuesday, June 15, 2004 2:17 PM
To: 'Ken Bosman'
Subject: RE: Request for additional information on recent LA Times poll

Ken,

- We purchased the state samples from Scientific Telephone Samples – www.stssamples.com, and we generated the national sample internally. This was accomplished by creating a series of replicates of 200 numbers each. These replicates are generated randomly from the pool of all exchanges in the nation. The only exclusions from the sample (post generation) are numbers we have called within the last year. We keep our database of area codes and exchanges updated via a subscription to Telechordia Technologies Active Code List.

- We do not adjust for marital status.

What she is saying is a fancy talk for having a computer take all the phone numbers in the US and generate a random sample of PHONE numbers.

A random sample of PHONE NUMBERS is NOT the same as a random sample of actual VOTERS

Here is just one HUGE systemic bias towards the Democrats this method creates.

Married couples (two voters) usually have 1 phone line.
Single people (one voter) ALSO usually usually have 1 phone line also.

(Not a perfect 2 to 1 ratio, some singles have roommates, some families have more than 1 phone line, but roughly speaking true - The ratio of phone lines to voters among single people is much higher than the ratio of phonelines to married people)

Result? - In the LA Times poll, which is a random sample of TELEPHONE numbers, a non-married person has a dramatically(about 65% according to Survey Sample Int, who does the phone lists for Gallup, Harris, Rasmussen, Survey USA, etc) higher chance of being contacted as a married person.

Single people vote VERY differently from married people.

According to the 2000 Exit polls:

Married people voted 53/44 for Bush
Single people voted 57/38 for Gore

Again according to exit polls form 2000, Single people made up 35% of the electorate in 2000.

Do you think a poll that systemically and structurally has a dramatically higher chance of reaching a single people (Who voted 28% more Democratic than Married people) versus a married person might just maybe be a tad skewed?

Bottom line:

35% of the electorate (single people) who voted 57/38 Democratic are structurally OVER sampled

65% of the electorate (married people) who voted 53/44 Republican are systemically UNDER sampled...

[Bandit3 the Worker]

The Los Angeles Times has a very strong conservative bias. For years it was owned by some rich right-winger. I don't know why everyone considers it a liberal paper.

If a poll interviews 10% more Democrats than Republicans, then it's probably undercounting Democrats, not Republicans. Among the general public, Republicans are clearly in the minority.

[Bandit3 the Worker]

Also, polls generally overcount Republicans, because Republicans are more likely to have more than one phone line.

[agcatter]

LA Times a conservative newspaper?  Besides the fact that has nothing to do with the actual sample which was taken (the actual issue here) that's just laughable.

[Bandit3 the Worker]

I think the phrase "election fraud" would explain that.

[Lunar]

Democrats have NEVER cheated in elections.

Oh wait.

[classical liberal]

Honestly, the Dems had the House locked up through gerrymandering for so long that the GOP has overcompensated in the opposite direction.  Without any gerrymandering, I'm fairly sure that the House would be D 218, R 216, I 1.  Most of the gerrymandering is in OH, IL and FL.  PA and NY cancel out.  TX and CA cancel out.  However if the redistricting plan works in TX then TX will probably tip into no longer merely compensating for CA but biasing the House as well.

[The Vorlon]

Honestly, the Dems had the House locked up through gerrymandering for so long that the GOP has overcompensated in the opposite direction.  Without any gerrymandering, I'm fairly sure that the House would be D 218, R 216, I 1.  Most of the gerrymandering is in OH, IL and FL.  PA and NY cancel out.  TX and CA cancel out.  However if the redistricting plan works in TX then TX will probably tip into no longer merely compensating for CA but biasing the House as well.

Actually the House is Gerrymandered in a bi-partisan way to massively protect incumbants of BOTH parties.

In states where no party could rig things for their own advantage (ie Texas) BOTH parties rigged it to protect the incumbants.

I did a House breakout about a week ago, and when I put on my GOP rose colored glassed (The ones where it looks like Bush has a chance in Vermont) - I found 21 potentially vulnerable Democratic seats.

When I put on my rose colored Democrat glasses (Georgia is a Tossup when viewed through these) I found 26 vulnerable GOP seats.

There are MAYBE 15 legitimately competitive House seats left in the whole natiion.

[Beet]

Guess which one is Vorlon, Mr. A, or Mr. B:

Mr. A:

Weighting a poll to party ID is something that very smart people can and do disagree on.

My own OPINION is that it is something you should NOT do,

Gallup, Harris, ABC, and TIPP, etc agree you should NOT

...

(The quality of the poll has FAR FAR more to do with factors totally unrealated to party ID weighting anyway, so in the end it is a fairly minor issue actually)

(blah blah, along the same lines)

Mr. B:

A few quick notes on party ID.

...

I am no news editor, but I think that was a relevant fact.  Especially when the result was totally at odds with a typical result.

(blah blah, along the same lines

[© Tweed the Younger]

It may be both, but I know he is Mr. A.

[ATFFL]

The two are NOT mutually exclusive.  You do not throw out peopel if you are close to a reasonable split in party ID, say a 5% Dem advantage.  Might be a bit off, but not so far teh poll needs tweaking.

Now, when assembling a representative sample a 10% or more Dem advantage is so far out of line as to rule out a true representative sample.  A 14% advantage is downright indefensible.  At that point you have clearly oversampled one side and should adjust your numbers for it, or throw out the poll as garbage.

If every other piece of methodology is good, a slight bias one way or the other will not make a huge difference.  If your difference is too big, the rest cannot compensate for it.

[Beet]

The two are NOT mutually exclusive.  You do not throw out peopel if you are close to a reasonable split in party ID, say a 5% Dem advantage.  Might be a bit off, but not so far teh poll needs tweaking.

Now, when assembling a representative sample a 10% or more Dem advantage is so far out of line as to rule out a true representative sample.  A 14% advantage is downright indefensible.  At that point you have clearly oversampled one side and should adjust your numbers for it, or throw out the poll as garbage.

If every other piece of methodology is good, a slight bias one way or the other will not make a huge difference.  If your difference is too big, the rest cannot compensate for it.

If you are saying its a subjective call, on how bad a poll is, I agree. The LA Times poll was bad, sure, but the biggest complaint against it was exactly of the same kind that millwx pointed out for the SurveyUSA poll in Florida. The latter poll was off by 8%, this one by 14%. Vorlon's attitude towards the two polls was not to treat the two polls similarly but say the LA Times poll is worse, or even significantly worse, but to treat the two polls in the exact opposite manner. He completely defended the latter poll in regard to the specific complaint made and completely panned the former in regard to that complaint.

[ATFFL]

In part because the SUSA poll used voter self identification, there is no registry to compare that to.  Survey USA may have a grievous error, or it may have found a trend for FL voters to self identify as Reps.  More polling is needed to check this, which is one of the very first thing Vorlon says.

[Beet]

All I said was that you defended the SurveyUSA poll against the complaint that it was not weight well for party ID and that you panned the LA Times poll using the fact that it was poorly weighted for party ID.

And you did.

Vorlon, I'm not complaining about all your analysis just the different ways you treat the same issue.

I wasn't trying to say you aren't critical of Survey USA (though you did conclude you were confident there was no bias one way or the other), but that you present two entirely opposing viewpoints on the importance of weighting party ID. This has NOTHING to do with the Reagan effect or California or Michigan polls, or the quality of SurveyUSA/LA Times polls in general. It also isn't about your different conclusions of these two polls. Your knowledge and analysis obviously superior to anyone else here. I was just surprised that the two diametrically opposite views of party ID weighting that you took between the two polls. Reading your post on this thread one would assume that party ID weighting is something you look at closely, seeing as you spent half your post on it. Believing the Harris numbers, the LA Times poll would only be 9% off (though I have no desire to defend the poll in general). The SurveyUSA poll was 8% off, according to what millwx said about the voter registrtion numbers. So not that big a difference, but a big reaction. That's ALL this is about. I think you're overreacting a bit.

I ALSO, in the very same thread (another post you seem to have perhaps "forgotten") addressed the party ID issue:

Some otherwise independents identifying as Republicans due to Reagan is a valid point, that's not what this is about. I really wish you wouldn't have to assume I'm being dishonest and deliberately "forgetting" or trying to misrepresent what you say. I've read dozens of your posts and had plenty of opportunity to do that before if I wanted. The descrepancy between your reactions on party ID in these two threads just struck me as especially surprising.

Well

I am a really nice guy and will accept your apology and retraction

I apologize if I implied that I was seeing contradictions in any other part of your posts other than what you implied about party ID in the different threads. When you type a lot and do a lot of analysis I realise its easy sometimes to come up with these seemingly different views. Especially since the LA Times has such a bad history, I would understand. So just admit your differing bias between the LA Times and SurveyUSA affected your analysis enough to make it seem contradictory, there's no shame in that.

So in response to all your other defenses of other parts of your post, I wasn't complaining about those.

[Beet]

Ok, thanks for changing back the name of the topic (no sarcasm this time).