Why in the hell are B** approval numbers going up ????

[Laura PourMeADrink]

http://pollingreport.com/BushJob.htm

L.A. Times/Bloomberg 4/8-11/06 Approve 39 Disapprove 57 Uncertain 4
USA Today/Gallup 4/7-9/06 Approve 37 Disapprove 60 Uncertain 3
ABC/Washington Post 4/6-9/06 Approve 38 Disapprove 60 Uncertain 2
CBS 4/6-9/06 Approve 37 Disapprove 56 Uncertain 7
Cook/RT Strategies 4/6-9/06 Approve 37 Disapprove 56 Uncertain 7
FOX/Opinion Dynamics 4/4-5/06 Approve 36 Disapprove 53 Uncertain 10

[shenmue]

1. They're drunk?

:shrug:

[benburch]

2. Nothing really horrible has happened for a little while.

Some people prefer to grant their approval to anything that isn't an out-and-out disaster.

[skooooo]

3. cause they're bringing out the scare tactics again....

They still get a little milage out of that, but not as much as they used to.

[Laura PourMeADrink]

13. You're right. I saw that they dug up an old al-Qaida tape from

NOV - saying they will annihilate Israel. MSNBC said that there was not explanation on why the tape was held so long.

[rocktivity]

4. Because he's threatening to scam us all over again with Iran?

:shrug:
rocknation

[enid602]

9. flags

Yesterday on the freeway in Phoenix, I saw 6 or 7 cars and trucks with American flags (big ones on wooden poles, not the small ones on plastic arms) hanging out of car windows or flat beds. Looks like the call to patriotism is again ramping up.

[Atman]

5. It is a slow pattern they'll need in order to pull of Election Theft 06

The goal is to get them up to the upper 40's again, ideally 47-48%, within the MOE, so they can say that "ultimately, it appears Americans trusted Mr. Bush and the GOP to protect their security."

Just a replay of 2004, but they'll try to pull all the GOP numbers up with them in order explain the "stunning upsets" Republican will win all over America.

[Alamom]

6. Selective Polling......when all else fails, call your friends. eom

[tuvor]

7. Dead-cat bounce n/t

[trogdor]

14. * has hit bottom.

Or close to it. There is more resistance at 35% than there is at 48%. Simple as that.

[Laura PourMeADrink]

20. Interesting thought. I asked a republican why, amid all the

horror and evidence B** lied, that people would still support him. He said that what you have left in support are the stubborn people who know he lied and know he has really fucked everything up, but they will not change because they can not or refuse to admit they were wrong to support him. A pride thing.

[mwb970]

31. It's more that they can't admit *we* were right. (n/t)

[ProfessorGAC]

8. Those Aren't Increases

Everyone one of those is within the bounds of precision of the polls. No need to be alarmed about this. Statistics aren't exact, when one only polls one of every 300,000 people.
The Professor

[LisaLynne]

18. Exactly.

What matters, IMHO, in polls is that you look at the over all trend and * is in free-fall.

[Alamom]

19. Agreed. I've always thought 1000 people or less IS NOT a good

representation of the voting public. It's really pathetic

[ProfessorGAC]

I Know Folks At Three Of Those Companies

The folks at Pew are friends, at one company is an former grad student, and i know others through professional organizations. We all agree that a good sample is between 22,000 and 45,000, depending upon the confidence interval and statistical power chosen, and which equations are used.

But, they tell me that the cost of doing that weekly or biweekly would be prohibitive, and these are sharp people. I believe them.

However, one of them and i have figured out a way to "rerandomize" the totals for all those polls. This gets us a distribution based upon averages that, of course, has a lesser margin of error, and a more normalized distribution. Using that technique, Silverspoon's numbers have been unchanged for about a month.
The Professor

[hlthe2b]

25. Have you published that method, Prof GAC?

If so, would you PM me the reference?

[ProfessorGAC]

29. No

The fella at Pew is trying to get them to use it, publicly. So far, he hasn't convinced the management of its value. But, we're still tryin'!

The problem is twofold:

There is an intellectual property issue, in that the responses and results of each company's polling belongs to them, airtight, and legally. So, using several sets of results and then publishing it could tread on copyrights.

Second: We can't be 100% sure of the value of individual polling, unless we can actually analyze the question itself, against the responses over a time series, to be sure that the "rerandomizing" normalizes push and pull type questions. Of course, Gallup isn't going to send their questions to Pew or ARI for review.

So, it's still sort of experimental. Mathematical valid, and seems to be quite accurately predictive, but not firm enough to convince the polling firm to consider its use.
The Professor

[hlthe2b]

33. Seems if you could replicate methods of leading polling firms

to the best of your current knowledge and apply this method in an academic analysis, get it published in a leading statistical journal, the polling companies would have to take notice. Of course the cost of replicating poll methodology would require grant funding$$. Seems there ought to be some source of funding for this? :shrug: Certainly make for a great Masters thesis or even PhD dissertation for a student?

[ProfessorGAC]

35. That's A Very Good Idea

I might just do that. That would be a good backdoor method of getting it published, even if my name isn't on it! Thanks.
The Professor

[Alamom]

41. Thank you for explaining how this is done with these groups. All

of these polls lately are driving me nuts and it's hard to know what to believe. I appreciate the information.

Would you know if the newspapers who poll 1000 or so people and come up with some of these stats & percentages go to these lengths to be as reliable?

[ProfessorGAC]

50. The Newspapers Commission Studies

They don't do them on their own. If they do, it's highly doubtful they have the stratification methodology to make them statistically valid. Most newspapers don't higher statisticians!

Many of the chains, and the big papers, co-comission studies with electronic media (like CBS/NYT or NBC/WSJ). They share the costs so they can be done weekly instead of monthly, and the big media companies (Chicago Trib, NYT, Time, and the like) publish the same results in all their papers.

So, with the Trib, for instance, the Tampa, Orlando, LA, Chicago, et al papers are all publishing the same survey.

Most of these surveys are done by the big 5 that we usually see, and there are a few smaller companies that are doing them independently or are subcontracted. Then there are a couple of purely partisan firms that work specifically for one party of the other. (See Rasmussen.)
The Professor

[ProfessorGAC]

51. Doggone It!

Double posted again.
The Professor

[ProfessorGAC]

24. Stupid Multipost

..

[taught_me_patience]

46. Its plenty enough to get a decent statistical result

if the sample is unbiased. This is a very common misperception of statistics.

taught.

[jsamuel]

36. It's never a good idea to compare different polls to get a trend

you should compare the past polls with the current ones of the same company, like Gallup vs Gallup last month

[ProfessorGAC]

37. If You Say So

I guess all that education and experience was for nothing. Thanks for the good advice. I guess we'll just trash that three years worth the work.
The Professor

[jsamuel]

39. i accidentally replied to you, I meant to reply to the op

I was saying that you could look at Gallup and say 37% a week ago, then look at Rasmussen today and see it at 43%. Does that mean he went up 6% in a week? NO! It means that Rasmussen is biased toward Bush as Ras was 43% a week ago and Gallup was 39% a month ago.

[ProfessorGAC]

47. No Harm Done

Actually, i've been working on this type of poll compilation concept, as a consultant, for about 6 years. The trick is applying it to different types of social statistics, from polls like these, to marketing surveys, to customer response polls, to pyschosocial studies. But, if it works with any of them, it appears to be most precise on simple approval polls.

Another poster provided an idea on how i might get these fleshed out and published, by assigning it as a graduate degree project. (Thesis/dissertation work.) That, of course, would require a complete validation of the theoretical underpinnings of the "re-randomization" work and the normalization methods. But, as long as i don't have to do it!

The Professor

[ProfessorGAC]

48. Double Post

...

[AX10]

38. true

[babylonsister]

10. Because this Moussaoui case is bringing back, for those who

forgot :eyes:, all the terror on 9/11, even though Moussaoui was in jail at the time?

[Christa]

12. My thoughts exactly. nt.

[npincus]

16. but you'd think it would remind people about OSAMA what's-his-name.

[izzybeans]

11. Those are different samples. Each sample will have a different

proportion. There is no way to say his poll numbers are going up by looking at these because (and I'm guessing) the differences all fall within the range of the error term.

And the Fox poll is not surprising for how low it is compared to the others. It is surprising that it is so high for Fox.

Do not despair. They are not going back to sleep.

[chat_noir]

15. Fox poll went down

Previous poll had bush at 39%

[IselaB]

17. I don't think they're going up

Different polling organizations will come up with different results even when taking the poll at the same time. To identify a trend, it's better to compare polls over time from the same organization.

I'm sure the three-point range there is within the margin of error of all of those polls. The only possible trend I see is that fewer people seem "uncertain."

[Laura PourMeADrink]

22. Yes, about the uncertains. Guess they split between approval

and disapproval - But disapprovals are trending up.

It just seems to me that every single week some new, horrible, thing comes out that Assclown did, and that it would reach a critical mass - enough to tip him into the 20's - but he's holding his own, amazingly enough.

[0007]

21. In reality and barring the fake world, junior is in the twenties.

[Laura PourMeADrink]

23. That's good news - means only 20 something percent of the

country is stupid.

[wordpix]

26. maybe because his Bushisms are so brilliant---see this thread

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=364x921666

[wordpix]

28. * neocon & fundie base think he's the 2nd coming, no matter what he does

Luckily, my bro who's big into repug politics in DC says repukes are VERY disorganized. Time to hit them hard before they regroup!

[mattclearing]

27. Margin of error. n/t

[hlthe2b]

30. 1. not statistically different from previous; 2. uneducable 35%

I agree with Professor GAC, that these are not statistically different from the previous.

In addition, I tend to agree with those who believe the bottom 35% or so are extremes in terms of being able to change their mind/opinions (or at least admit it).

As for the determined, uneducable 35%, add to this the (apparent) constant barrage of both the Moussaoui trial and the constant barrage of ads for the Flight 93. BTW, does anyone feel a tug of tinfoil hat, when they hear about the determination of the Flight 93 producers to get this film out and to show the previews in theaters, despite the negative public reaction? Just asking?

[RB TexLa]

32. because for anyone there are 25 to 30 that support no matter what

the other 6 to 9%, are people that are very apolitical and unless the sky is falling or troops are outside their city and they have a job are going to say things are ok, and get off the phone.

[WiseButAngrySara]

34. These have to be bogus numbers! 86% of about 250,000 voted for

IMPEACHMENT on a recent MSNBC Poll!

http://www.democraticunderground.com/discuss/duboard.php?az=view_all&address=364x919932

[Jamison]

40. Don't worry.

Gas prices are soaring to all-time record highs, so I think his approval rating is going to take another small tumble soon.

[tjwash]

42. He's not, he is still on a nose dive.

Check it out...

The latest Cook Political Report/RT Strategies poll shows "that the dismal numbers for President Bush and national Republicans have not only jelled, but hardened."

Key findings:

* President Bush's approval rate is now 37%, down three points from the previous month.
* In the generic congressional ballot, Democrats beat Republicans 46% to 36%.
* The approval rate for congressional Republicans is now just 38%.
* In a 2008 presidential matchup, John McCain beats Hillary Clinton, 44% to 39%.

[porphyrian]

43. Because most of those polls are bullshit? - n/t

[leftofthedial]

44. people are beginning to disapprove of disapproval

more than they are of the turd at the top.

[Zynx]

45. Those are all different polls with insignificant differences.

[WI_DEM]

49. well what was the LA Times poll before this one?

that is what you look at. Was it less than 39% or more? They actually all are in the same margin of error.