CNN Opinion Research Poll finds 37% think Obama a failure

CNN ^ | Yesterday | CNN

[StAntKnee]

Linked from Real Clear Politics piece that says Obama's numbers are worse than Bush's at the same point in his presidency, but, from the poll intro . . .

"Interviews with 1,136 adult Americans, including an oversample of African-Americans, conducted by telephone by Opinion Research Corporation on July 31-August 3, 2009."

Why oversampling in that group? How much? And how bad are the numbers if the sampling is the normal 11%?

[StAntKnee]

Why oversampling in that group? How much? And how bad are the numbers if the sampling is the normal 11%?

If you look for poll demographics in the link, you won't find it. Perhaps this is fishy enough to forward to the White . . . er, Oval-Office-House.

[Steamburg]

If the last of the Kool-Aid (a real Jim Jones reference, kind of like Obammacare) drinkers is down by 37%. His remaining supporters could all be found in just a couple of blue city pockets. This is a time to be extra watchful. Rats are dangerous when cornered.

[kellynla]

and a PHONY!

[greatdefender]

Interviews with 1,136 adult Americans, including an oversample of African-Americans, conducted by telephone by Opinion Research Corporation on July 31-August 3, 2009.”

Despite a very small sample number, the MSM will get poll numbers anyway- even if it does not represent the true voice or the majority of the people.

[screaminsunshine]

Eliminate 200 favorable votes and you will arrive at the truth.

[rockinqsranch]

Know your history.

CNN=Clinton News Network
ORC=Friend of Clinton, Gupta

2+2=Gotcha!

Yeah it’s fishy all right.

[freespirited]

Oversampling of blacks in this type of poll is a standard and accepted statistical technique. It is needed due to what statisticians call “clustering,” the tendency of blacks to concentrate in a few geographical areas.

Weighting (adjustment of the numbers) is used to compensate for the oversampling. It’s kosher.

[USFRIENDINVICTORIA]

Oversampling of a sub-group is commonly done for statistical purposes — it's necessary to lower the margin of error for reporting results by sub-group. It's standard practice in survey research. The oversampling is factored out in the analysis. There may be lots wrong with this poll — oversampling of African-Americans isn't one of those things.

[USFRIENDINVICTORIA]

GMTA!

[throwback]

Interesting if the poll is authentic. Zero tracks pretty close to Clinton for overreach. We’ll see if zero gets the same backlash in 2010. Also interesting to see how many people thought the economy was terrible in August of ‘08. The media really shapes perception. Republicans still get most of the blame too, so zero will keep playing that. Guess not enough people have lost their jobs before zero gets full ownership.

[Osage Orange]

It ain't kosher.........

It's re-dick-U-lus

[USFRIENDINVICTORIA]

It’s not only kosher — it’s necessary.

Read up on “oversampling” here:

http://people-press.org/methodology/sampling/

(There are millions of other sites on survey methods and statistical analysis you could also research to confirm this.)

[Kartographer]

[Osage Orange]

Well....common sense say's to me...that people are people.

Although I suppose I understand..that if one is to "sample" correctly you have to pick some folks percentage wise, to the population.

The problem in this instance...is that something like 95-98% of black people in this country voted for Obama.

So...imo, that makes the sampling inaccurate.

[kjo]

Barry: Sinking like a stone.

[StAntKnee]

Thanks for the insight.

Still, I’d like to know how much it’s over-sampled, wouldn’t you?

The fact that I don’t see the demographics on the sample, as I do on most other surveys reports, makes me suspicious.

The cynicism in me, I suppose.

[USFRIENDINVICTORIA]

You're wise to be skeptical of the methods used in polling. There are many, many ways to load a poll (as in “loaded dice”).

One important thing in a statistical survey is the margin of error — this is usually reported along with the averages. The more people you survey, the smaller the margin of error.

If you want to report on differences between African American opinions, and those of the general population; you have to have to have enough African Americans in the sample to get an acceptable margin of error. Oversampling is, by far, the most economical and efficient approach.

If you over-sample a group, you're supposed to weight their responses, to neutralize the oversampling. E.g. over-sample by a factor of 5, weight each response by a factor of 1/5. If the oversampling weren't neutralized in the analysis — that would be a very serious error.

[2ndDivisionVet]

Related thread:

Matt Towery: Is Significant Drop In African-American Support In Obama’s Future?
http://www.freerepublic.com/focus/f-news/2310309/posts