Well, well, well - here are Gallup's internals - bias galore!!!

[annerevere]

From the great blog, The Left Coaster, which has provided excellent analysis on Gallup's internals, here's their report on Gallup's poll out today:

Poll of October 22-24, 2004
Reflected Bush Winning by 52%-46%

Total Sample = 1195
GOP: 466 (39%)
Dem: 406 (34%)
Ind: 321 (27%)

Note that Gallup has used a sample today that turns a 3% GOP advantage in their LV sample last week into a 5% advantage. Yet Bush's margin went down.

Link: http://www.theleftcoaster.com/

[THUNDER HANDS]

1. if you readjust those numbers based on likely voting patterns

I'd say we win 53-46, especially since I think more than 27% of the vote will be from independents, who are abandoning Bush in droves.

[leftchick]

5. I love your wolf puppies....

:)

[yardwork]

28. Those are vicious attack terrorist wolves - didn't you know?

;-)

[Anaxamander]

2. ah HA!

This is so ridiculous it's almost funny... in a sad, irresponsible, dishonest kind of way. :|

[wtmusic]

6. Love your melancholy smiley

;|

[montana500]

3. lol

nm

[Birthmark]

4. So, using my mad math skillz...

...I come up with a Kerry *lead* of three points, 50-47.

[swag]

18. Gee, that matches Democracy Corps - son of a gun.

[featherman]

34. this is about the likely spread IMO

[in_cog_ni_to]

7. Shall we let CNN know that WE know about this?...AGAIN!

Remember when they had to go on the defensive over their other polls? Gallup SUCKS!!!

I'm off to email CNN! Bastards.

[pabloseb]

8. Where did theleftcoaster get those numbers???????

There's no sampling by party breakdown even in the very detailed gallup internals for subscribers. For REGISTERED voters, the sampling can be determined from some of their internals, and it is: Dems 41%, Repugs 39%, indeps 20%. For LIKELY voters I guess theleftcoaster is close to the truth but Gallup is HIDING those numbers because they don't want their "mathod" exposed.

[HFishbine]

10. You're a subscriber to Gallup?

Just wondering how you know there is no party breakdown for subscribers?

[pabloseb]

12. It's over on the freepers site

Another DUer posted a link to the internals over in the freepers site. There were all kinds of breakdowns, but nothing for likely voters, just the total.

[HFishbine]

17. D'uh

Umm, pointing out that Free Republic does not have the party breakdown is meaningless. Because they don't have it nobody does?

[pabloseb]

22. No, the point is inconsistency

From the internals at http://www.freerepublic.com/focus/f-news/1256482/posts one can deduce that for *registered* voters the breakdown is: democrats 41%, republicans 39%, independents 20%. This is very different from The Left Coaster numbers.

Please note that I think The Left Coaster for likely voters can be accurate, if not slightly conservative (there's even more bias towards the gop).

[HFishbine]

23. Deduced? How?

I'd be curious to know how you deduced those numbers and why you consider them more valid than the numbers received by a subscriber. If you are observing "inconsisency" between the two, where is the likely point of fault?

[pabloseb]

24. There's inconsistency but i'm not blaming anyone

The internals posted at the freepers site give the following breakdown by party among registered voters:

Kerry Chimp
Total 48% 49%
Reps 5% 94%
Dems 88% 9%
Inds 49% 44%

So if x = proportion of repugs, y = proportion of dems and z = proportion of indeps, then

0.05 x + 0.88 y + .49 z = .48
0.94 x + 0.09 y + .44 z = .49
x + y + z = 1

Solving this system gives x = 39%, y = 41%, z = 20%. Very different from Left Coaster numbers (but doesn't make sense either... if true, they're undersampling independents, who favor Kerry).

[HFishbine]

27. It's been a long time

since I've been able to solve an algebraic equation with three variables. I'll leave the math to the expert.

I wonder if the proportion by party breakdown was among likely or registered voters. That could account for the inconsistency.

[lancdem]

11. He gets the numbers mailed to him by Gallup

[pabloseb]

13. If that's true, Gallup poll is totally made up

There are four possibilities:

* The super detailed report we've seen at freepersland is made up.
* The Left Coaster made the numbers up, or tried to deduce them from the gallup internals but made some math mistake.
* Gallup poll is entirely inconsistent with elementary linear algebra, and is therefore made up in a very obvious way.
* I made a math mistake when computing registered voters breakdown by party.

I double-checked to make sure it's not the last option. For the record, I'm a mathematician.

[pabloseb]

14. In any case, Gallup is totally wrong

when it comes to likely voters. We all agree about that.

[lancdem]

15. I vote for No. 1

[pabloseb]

16. Here's the link to the freepers internals

http://www.freerepublic.com/focus/f-news/1256482/posts

[lancdem]

9. It's almost as bad among RVs, too

BTW, Gallup's last poll comes out the day before the election. I almost hope they keep using these skewed samples so they look like fools when Kerry wins.

[unfrigginreal]

19. LOL- More crap from Gallup

I don't see how CNN or USA Today will be able to justify using these voodoo pollsters after this election.

[in_cog_ni_to]

20. I just emailed these Total Samples to CNN and MSNBC

and told them that we know the Gallup polls are BOGUS! Sent out 10 emails. They need to know that WE know what the hell they're up to. :grr:

[in_cog_ni_to]

21. dupe

[okieinpain]

25. from what I've heard on c-span is that they pollsters claim that

they are not calling folks by party, but that they are just cold calling folks at random and that this is how they are identifying themselfs. but what I don't understand is why they don't re-weigh the polls, to fit the last two elections.

dems have had higher voting turnout then repugs, remember that's how they said that reagan won. he got some dems to go and vote for him. Are the polls now trying to say that those dems are now indentifying as repugs.

[Beetwasher]

26. That Very Well May Be True, But It Means There's A Problem W/ Their Method

IOW, their methodology is resulting in more Repubs being called for some reason and they need to weight the results to reflect and correct for that...They're not weighting properly if they are not doing that...That's exactly what weighting is for, to correct that sort of sampling bias and for whatever reason they refuse to do it properly...

[yardwork]

29. Gallup has addressed this in the past

Their reason - which is highly unlikely, imo - is that people keep changing their party affiliations as they go back and forth on which candidate to support.

IOW, Gallup says that their numbers of registered Republicans, Democrats, and Independents change with the numbers of people polled who lean toward *, Kerry, or undecided.

I simply don't believe that. First, I don't believe that very many people keep changing their minds about who to support for president. A few, but not very many.

Second, even the undecided voters don't keep changing their party affiliation! That's ridiculous. The reasoning is that millions of Americans are changing their party affiliation on a day-to-day basis as they veer back and forth between Kerry and *?

"Uhhhhhh....I think I'll vote for Kerry. That means I'm a Democrat! Wait a minute, I guess I support * instead. Now I'm a Republican!"

I don't think so.

[Beetwasher]

31. Actually, That's True

That does happen, the question is, how often? I'm not sure there's a good way (or any way) to accurately identify that variable...I think they are in denial about their methodology and WAAAY overstating the incidence of party switching...

[Nicholas_J]

30. Of Course Gallop is reversing the average voter turnout

Over the last three elections, where Democrats came out and voted at 38 percent, Republicans came out at 35 percent.

[snippy]

32. Gallup's methodology for determining likely voters causes the bias.

To determine likely voters Gallup asks a series of seven questions of each respondent and based on the answers to those questions assigns a numerical score to each respondent. The seven questions include questions about voting history and certainty of voting.

All respondents then are ranked according to their score, from highest to lowest. Gallup then uses the expected turnout percentage as a cutoff point among the ranked respondents and those respondents above the cutoff point are called likely voters and those below it are not.

This methodology results in more "likely voters" from the party which is more energized at the time of the poll since voters from that party usually score higher on the seven questions. However, this methodology virtually excludes first time voters. It also is susceptible to measuring even a brief reaction to the latest news story at the time of the poll. Its biggest flaw though may be that it relies heavily on the turnout percentage from the last election in order to guess at the expected turnout for the current election.

[Mike L]

33. Gallup, Jr. is a fundie.