Ipsos Watch 2

Posted by John Rosenthal

Having observed last month that the AP’s French polling company, Ipsos, seemed perhaps to be changing its highly dubious sampling practices in the face of bad publicity, I would be seriously remiss not to note that in the meanwhile Ipsos is back to its old tricks.

The latest AP-Ipsos presidential approval ratings, released in the first week of January, are based on a sample in which fully 52% of the registered voters sampled identify as Democrats, as opposed to merely 40% who identify as Republicans. As the data I presented in my last “Ipsos Watch” post shows, this split is not untypical for AP/Ipsos samples. However, last month – perhaps not coincidentally, following the publication of an article by me on TechCentralStation exposing the long history of cozy relations between Ipsos and the French political and economic elites – the party registration gap in the Ipsos sample suddenly shrunk to a far more plausible 47%-44% Democrat-Republican split. If this more plausible split – or even anything vaguely approaching it – had been maintained in the sample in the latest poll, then George Bush’s approval rating would undoubtedly have risen for a second straight month in the Ipsos poll and to a “respectable” figure somewhere in the mid-40% range to boot. We couldn’t have that, could we? (After all, just a month and a half ago, the Le Monde group – making use of the subtlety and sophistication that is its trademark – had carefully explained to the inquiring minds of its Parisian public that the alleged decline in the President’s fortunes was beyond repair.) In light of the inflated party-registration gap in the Ipsos sample, an associated AP-Ipsos report allegedly “showing” that Americans want a Democratic-controlled congress in 2006 is positively amusing. For AP and Ipsos, it is apparently a scoop that Democrats would prefer a Democratic-controlled Congress. Highly informative!

As I pointed out in my last “Ipsos Watch”, while it would not be unusual for presidential approval ratings to show a certain volatility in response to particular events, such volatility in party identification numbers makes no sense at all. I repeat my question: does anyone at Ipsos care to explain?

My suspicion is that the Ipsos people prefer instead to hide. Ipsos’s “topline” results – including the breakdown of the party identification figures – used to be available free on the Ipsos site for a month. The latest data disappeared from the site with remarkable speed.

[Note: Click here for the complete “Ipsos Does America” Trans-Int dossier. Other dossiers are available in the left sidebar.]

Comments are closed.