Were CNN Poll Questions Weighted Against Surge Report?

CNN released a poll on the 16th that claims that 53% of Americans don't trust the U.S. Military assessment of what is going on in Iraq and that 72% won't have their mind changed on their view of the war no matter what General Petraeus says about the surge next month. But if one reviews the questions of the poll and its methodology is reviewed (at least the only hint of methodology released), it makes one suspicious that it was anywhere near a fair and balanced method. In fact, there are so many questions about how this poll was carried out that the results must be viewed with skepticism.

To start with, of course, the poll is conducted by Hillary Clinton supporter Vin Gupta's Opinion Research Corporation, the organization CNN has hired to run their political polling -- a convenient situation for the Clinton campaign, to be sure. This single fact alone is enough to inform that the poll could likely be weighted to skew toward the ideas that Hillary Clinton is propagating in her campaign.

According to the front page of the partial downloadable PDF file of the poll, it was compiled from "interviews with 1,029 adult Americans" by telephone between August 6th thru the 8th with a plus or minus 3 percentage points.

There is no indication what party the respondents claimed to be members of, there is no mention if they were voters, registered, or likely. No geographic region is identified, no age bracket and no gender info for the poll is offered. This also causes skepticism. After all, they could have asked all Democrats, or weighted the Democrats to be a higher percentage. Maybe more women than men were asked? Maybe all the respondents were in the environs of Washington D.C., or maybe they were all women in Austin, Texas!? We have no idea as no facts of the sample size are revealed.

Then we get to the odd choice of asking only half those interviewed some of the questions from the poll. What was the deal with this? At least questions 28 through 33 were only presented to half those interviewed. One of those questions pertained to how respondents viewed the report general Petraeus would be giving next month.

But which half were asked this question? Was the question asked of women and not men? Was the question asked of Democrats and not Republicans? How do we know how this question was weighted so that we might assess the legitimacy of the results?

Further, if many of the questions were only asked of half the respondents, doesn't that mean that the poll was not conducted among 1,029 adult Americans, but was really only conducted among some 514 adults, roughly half the claimed sample size?

As Duane Patterson of radioblogger says, "It's hard to take a poll seriously when on the one hand, 50% can support the war or say they're open minded to change their mind, and then in the next breath say 72% wouldn't change their mind on Iraq regardles of what General Petraeus might say, because most people don't trust him anyway."

All in all, what we have here is another questionable CNN poll by Clintonista, Gupta, that is possibly weighted toward the antiwar side and one that should be viewed with a healthy skepticism... not that any of CNN's viewers and readers would be aware of the problems here.

Here is a breakdown of the CNN slime job -

CNN released a new opinion poll today, and it appears that the slime campaign on General David Petraeus has now officially begun with a month to go before his report to Congress. CNN claimed on their Situation Room that only 28% of responders would be more likely to support the war if Petraeus reports the surge is showing signs of progress, 72% wouldn’t. And worse news, if one were to believe this poll, only 43% of those polled trusted Petraeus to give an accurate report in September, while 53% said they don’t trust the top U.S. military commander in Iraq.

Now let’s dive into the numbers a bit, to see how trusted this poll ought to be. The sample given by CNN is 1,029 Americans, with no political idenfitication given, taken over a three day period from the first part of August. Question 24 is the first question CNN released, which is:

24. Do you favor or oppose the U.S. war in Iraq? 24a. Is your mind made up about the war in Iraq or do you think you could change your mind?

Aug. 6-8

2007

Favor/mind made up 21%

Favor/could change mind 12%

Oppose/could change mind 17%

Oppose/mind made up 47%

No opinion 3%

So a total of 50%, if you include the supporters and those who state they could change their mind on Iraq are reported among the 1,029 sample. This actually might not be a bad number. But look what happens when you get to the stats against Petraeus and his upcoming report that CNN ran with.

32. As you may know, in September the top U.S. commander in Iraq will report to the President and Congress about how the war is going. If he reports that the U.S. is making progress, would that make you more likely to support the war, or would that have no effect on your view of the war? (ASKED OF HALF SAMPLE)

Aug. 6-8

2007

More likely to support the war 28%

No effect on your view of the war 72%

No opinion *

33. As you may know, in September the top U.S. commander in Iraq will report to the President and Congress about how the war is going. Do you trust him to report what’s really going on in Iraq without making the situation sound better than it actually is, or don’t you feel that way? (ASKED OF HALF SAMPLE)

Aug. 6-8

2007

Trust him to report what’s really going on 43%

Do not trust him to report what’s really going on 53%

No opinion 4%

Only half the sample was asked on each question, bringing the sample size of a U.S. population of 300 million people down to a little over 500 people. Which 500 were asked this question? Were only the left-leaning respondents asked the question on Petraeus? Why weren’t all respondents asked all the same questions? Would the results have been different? Why would CNN tout a 1,029 sample size when they really didn’t use a sample size that big to ask the juicy questions, the ones they’re manipulating to try and turn the tide against a positive Petraues report to Congress a month before he gives it?

I don’t trust CNN to not play around with the poll numbers. Until they cough up the rest of the methodology and questions, like which half they popped the full set of questions on, or if the full set were asked of the full sample size, why only half were used in certain questions, it’s a propaganda poll to push CNN’s agenda to declare defeat in Iraq no matter what the cost.

Show us the political ideology of who the half sample is, and the political ideology of the full sample size, and show a little transparency in the polling process if you want it to be believed. It’s hard to take a poll seriously when on the one hand, 50% can support the war or say they’re open minded to change their mind, and then in the next breath say 72% wouldn’t change their mind on Iraq regardles of what General Petraeus might say, because most people don’t trust him anyway. It’s a contradiction that busts this poll. CNN ought to have seen if it they wanted to be fair, but then again, CNN is not in the fair business.

As you may know, in September the top U.S. commander in Iraq will report to the President and Congress about how the war is going. Do you trust him to report what's really going on in Iraq without making the situation sound better than it actually is, or don't you feel that way?

This is my problem with polls. Basically even the existence of this question leads to a biased result. Because, it implies that it's a realistic possibility that the top U.S. commander somewhere would lie to Congress about the state of affairs there. But no such poll question exists about the top U.S. commander in South Korea. No such poll questions exist about some report the Surgeon General is going to give Congress about some health issue, or that the Sec. of the Interior is going to give Congress about Indians, or whatever. In other words, by putting this particular question on the table CNN is already taking a position as to which government officials (Petraeus in this case) are prone to possibly lie, and about which situations (Iraq).

In general, polling people about some question implies that both "yes" and "no" are equally valid answers. The poll question being asked means that the issue is on the table. And this constitutes a bias because: who says it's on the table? Who says it's even a reasonable possibility fit for mainstream conversation that Petraeus would somehow lie to Congress about this?

Implicitly, CNN does. And anyone taking this poll will see that, if only subconsciously. And if CNN says something, maybe there's something to it. Thus, the answers get biased.

I'm not even sure there's a way to resolve this bias, either. One possibility would have been to have asked the question in the inverse way (and strip out the "As you may know" part, which is unnecessary... why do pollsters always front-load with an "As you may know"? So they can color the result, right?). But then the question would look like this:

"Do you have confidence in the honor of General Petraeus & that he will report accurately to Congress about his field of command, or don't you feel that way?"

I doubt this question would have come out 43% yes - 53% no. I suspect the "yes" vote would have been much much higher. Which, if true, illustrates the problem right away. Of course CNN didn't choose to ask it this way. (Why not?)

There are subtler ways they could have colored the result: notice they didn't refer to him by name, let alone by his title "General Petraeus", in the question; they just referred to an unnamed "the top U.S. commander in Iraq". Would the result have been different?

Another possibile change would be to reconstruct the question so that it's at least perhaps comparing apples to apples:

"Of the following current or recent public servants, in whom do you have the most confidence to tell the truth and report honestly about their respective areas of responsibility: Bill Clinton, Dick Cheney, Ted Kennedy, Joe Wilson, Karl Rove, .......[etc etc etc], or General David Petraeus?"

But really, these would lead to only marginal improvements. And that's only my opinion. The point is that how to ask the question was a choice made by the pollster, and you can't avoid this. And once he's making a choice, it's fair to speculate why one choice was made and not another, and how the results would have changed. So the real problem is that you simply can't poll (on something substantial and nontrivial) without influencing and coloring the results in some way, because you're always choosing one question (and not others) and allowing for some answers (and not others).

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.