New CNN Poll Is A Total Sham

This from CNN has to be one of the most ridiculous polls I've ever looked at. One need only look inside. Why don't they simply change their name to DNC and be done with it. In essence, they nearly doubled the percentage of blacks versus the percentage that voted in 2008 and under sampled whites to get the result they're pushing.

CNN Poll: More Americans think Democratic policies better than Republican  Washington (CNN) - Despite the bruising battle over their health care reform proposals, congressional Democrats have maintained an advantage over their Republican counterparts on one key measure, according to a new national poll.

Poll here, what little they reveal of it. First we learn that they over-sampled blacks by design. So, by definition, they under sampled whites. And even those numbers are telling. If these are straight numbers, they only interviewed 15 hispanics, or otherwise unidentified. What was the point?

The number of blacks included equates to 22.3%. Yet, even with Obama running in 2008, they only comprised 13% of the electorate, up from 11% in 2004. In essence, they nearly doubled the percent of blacks with a still limited sample yielding a 6% MOE within that group. And it's only due to that number that they can claim the result they do. What a sham.

Interviews with 1,160 adult Americans, including an oversample of African-Americans, conducted by telephone by Opinion Research Corporation on December 16-20, 2009. The margin of sampling error for results based on the total sample is plus or minus 3 percentage points. The sample also includes 259 interviews among African-Americans (plus or minus 6 percentage points) and 786 non-Hispanic whites (plus or minus 3.5 percentage points).

They also suggest Lieberman is taking a hit for not supporting a public option. Perhaps, but his numbers slightly improved among the under sampled white sample. The hit is entirely among the inflated black sample. As for the Democrats going in the Right or Wrong direction. Even 16% of blacks say wrong direction. A majority of whites say wrong. Without the oversampling, the majority would likely be wrong direction.

10. Do you think the policies being proposed by the Democratic leaders in the U.S. House and Senate would move the country in the right direction or the wrong direction?

Total      Right - 51%  Wrong - 46%    No Opinion - 3%
Blacks     Right - 82%  Wrong - 16%    No Opinion - 3% - oversampled
Whites    Right - 42%  Wrong - 53%    No Opinion - 4% - undersampled

As for Republican policies, that's a freaking joke. None of them even get discussed as they aren't relevant given the large Democrat majority in Congress. After that, the only other question they disclose is basically do you trust the government. The answer is no, not that much. They should add CNN to that list. See for yourself. The whole think is a joke.

Comments

[PA]

Hide the decline.

[crosspatch]

There's a reason why people are turning away from CNN in droves.

[joe]

sorry, but you don't understand what oversampling is. The extra African-Americans are not included in the total numbers (they never are) - they are only used to make the sub-sample large enough for the black-white comparison. This is a very standard procedure when you went to compare sub-groups and one is too small for its portion of the overall sample to be able to make statistically significant claims. Again, the oversample of African-Americans *is not* included in the total, its only used for the comparison. This is also why you mistakenly thought they only polled 15 Hispanics, b/c maybe half of the blacks sampled were not included in the grand total.

All you have do is some simple math to figure this out - 786 white out of a sample of 1,160, or 68%. Compare this to the census number of a little over 65%, its right where it should be.

[Dan Riehl]

"All you have do is some simple math to figure this out - 786 white out of a sample of 1,160, or 68%"

I don't care about the census number, Joe. Use their numbers "The sample also includes 259 interviews among African-Americans"

That's 22% blacks and when it comes time to vote, they don't come close to that (11-13%). If they want to post it on a social sciences site, fine. But it's misleading on a political site. I could care less about what any group of people who sit home on their asses and don't even vote think. They elect to not participate in the process. They can take what they get and shut up.

[craig]

There's a poll called Rasmussen that is totally biased. I don't see you criticising it. If you like it you consider it accurate. If you don't like it it's biased. Who reads this anyway?

[joe]

Did you read what I wrote? When a poll oversamples it *does not mean* that the extra people they take with are included in the total, they are only used to make subgroup comparisons. CNN polls call people at random, so the total numbers most likely include something like 11-13% African-American and 13-15% Latino. Again, the total sample *is representative* of the population.

Once they have their representative sample and have calculated the numbers you see as 'total' in the poll, they go out and find some more African-Americans so they have a large enough group to make black-white comparisons. That is *the only place* the oversample is used.

You can read their own explanation:
http://www.cnn.com/ALLPOLITICS/1996/polls/FAQ/polling.FAQ.shtml#oversample
"The term comes from the fact that we wind up producing a sample of African-Americans over and above those we have already interviewed as part of the regular sample of respondents."

The poll is no sham, you just don't understand how it works.

[Dan Riehl]

I understand that, Joe. I was going by the language I quoted. If that number isn't valid, then they disclosed no percentage of respondents that are black, white or hispanic at all. The final breakdown would be 67W - 22B - 11H/O or something like that. Are you disagreeing?

"The sample also includes 259 interviews among African-Americans"

[joe]

Yes I am disagreeing. Oversampling is a two step process. First, CNN calls 1,160 people randomly. Because they call them randomly this sample is representative of the overall population, so is something like 65-67 white, 11-13 black, 13-15 Hispanic non-white and 5-10 Asian-American and other. When you look at the poll and see a number labeled 'total' it comes *only* from this representative sample of 1,1600.
The second step is that they want to compare just blacks and whites. They take out the 786 whites and look only at them. They don't have enough blacks from the original sample to make any statistically significant claims so they get some more black respondents *just for this comparison*. This is the oversample. The original 1,160 *does not* include 259 African-Americans. It's probably around half of them.
Again, this is really standard polling procedure when you want to look specifically at the attitudes of a sub-group that is a small part of the overall sample.

[the foundation]

Joe,

1) Riehl is claiming that the white number is still too low under the scenario that you have outlined. The percentage of the electorate that was comprised of whites in 2008 was 74%. Of course, CNN is polling all adults and that seems to be the big issue that many conservatives have with "adults" polls. The actual electorate is always more conservative than the "adults" population.

2) It's unclear from the way that CNN worded it that your scenario is how it played it out. I agree that with an "oversample," that probably meant they got too few blacks in the sample. But why wouldn't they "oversample" non-white Hispanics as well considering that they comprise a smaller percentage of the electorate than blacks?

[joe]

The Fountain,

1) I didn't see Riehl making this point anywhere, but I think its legitimate if the critique is about a preference for a 'likely voter' model. Again, though, Riehl's point as I read it was not about this but rather about the racial makeup of the sample and so far as I can tell it was based on his misunderstanding.

2) It's unclear only if you don't understand how polls work. Oversampling is very common and its easy to find out what it means. I don't think CNN or any other news organization ought to be obliged to explain fundamental and ubiquitous polling procedures every time they report some poll. To your point, they don't oversample Hispanics because they aren't looking at Hispanic specific attitudes. As I wrote several times above, the oversample of African-Americans *is not* part of the 'total' numbers. They are *only* used to make the black category large enough to produce the numbers in the rows labeled 'black.' There is no Hispanic comparison group, so no need to oversample.