It is increasingly clear that IPSOS, a French polling company with close ties to the French leadership is producing polls for AP designed solely to allow them to run editorials in the guise of news stories. The samplings are widely out of line; the questions off the mark, and the results not surprisingly advance a Democrat point of view obviously shared by AP.

But the last two  IPSOS polls seem to have broken even their usual records for untrustworthiness. Big lizards notes the likelihood of fraud:

The Associated Press is touting its own AP—Ipsos poll purporting to show a significant majority of Americans opposes the Bush administration on warrantless tapping of al—Qaeda phone calls originated abroad to their terrorist agents in the United States. But as typical these days, there is much less here than meets the eye.

The poll in today's story is very similar to the "generic congressional poll" results announced yesterday purporting to show a massive shift in Americans' support for Democratic control of Congress, a seismic shift that would signal that the Democrats are poised to seize back the House; that poll found that if the mid—term election were held today, 49% of Americans would want to see the Democrats win, while only 36% would want to see the Republicans win. Besides both polls showing a big surge towards Democrats, they have other similarities:

They have exactly the same age breakdown;

They have exactly the same employment breakdown;

Same educational breakdown;

Same marriage breakdown;

In fact, every, single demographic question is identical on the "two" surveys.

And oddly enough, they were even conducted over the same days, by the same pollsters, with the exact same number of respondents.

One might almost conclude that this was really the very same poll —— just mendaciously reported twice, in two different contexts, to make it appear as though there were a trend moving in the Democratic direction, a rising crescendo of criticism of George W. Bush. But of course, it would be dishonest for AP to do that without noting the fact, so it can't be true. It must simply be an eerie coincidence.

Another point the "two" polls share: they both wildly overpolled Democrats —— 52% of the respondents were Democrats, 40% were Republicans, and 8% were independents; that is, they polled almost a third more Democrats than Republicans.

Read the whole thing. I think Big Lizards nailed it.

Clarice Feldman  1 08 06

Reader Meg Kreikemeier writes:

Clarice Feldman posted a piece about the most recent AP poll and linked to Big Lizards.  Neither she nor Big Lizards have an e—mail so I'm writing to you about a concern I have with the analysis. 
 
I agree that the AP poll was heavily skewed to democrats, but the issue Clarice Feldman quotes from Big Lizards that these are two polls with identical information and that somehow this is a sign of fraud is wrong. 
 
Pollsters sometimes release data over many days.  As a result (and more than likely) the two headlines about GOP control of congress and the NSA intercepts were from the same poll.
 
If the point of the post is to draw attention to a faulty poll then it is important to focus on where the poll appears to be faulty —— in its demographics.
 
In May 2005, I wrote a piece for the Chicago Tribune (archived and only available if you pay $4.00) that showed how the media uses polls to frame issues. 
 
This AP poll is an example of the media running with a poll because it suits their viewpoint regardless of what the underlying data says.  And I believe it is highly likely that the AP did not review the top—line data before publishing the stories. 
 
More sinister, of course, would be that the AP knew that democrats were favored by 12 percentage points but published the articles anyway because it furthered their agenda. 
 
Given that the democrats were losing traction on the NSA intercept story, the timing of the poll and its results seem to be suspicious because it gave the AP and any other publication using the AP report a chance to reframe the issue.