Mark Palko points me to this article by Harry Enten about a possibly nonexistent poll that was promoted by an organization or group or website called Delphi Analytica. Enten conjectures that the reported data were not fabricated but they’re not a serious poll either but rather some raw undigested output from a Google poll.
This sort of thing is not new. Here’s an example I wrote about in 2008, a possibly nonexistent poll that was promoted by a consulting company called Prince & Associates and which got blurbed in the Wall Street Journal (yup, John Yoo’s newspaper). The data from that poll may be real, or maybe not, but no evidence was ever provided that the sample, if it existed, was representative of the claimed population in any way.
Eternal vigilance is the price of journalism.
P.S. Enten includes the following chart:
I disagree with this chart for two reasons. First, just cos a poll is real and it comes from a respected pollster, it doesn’t mean we should take it seriously. Remember Gallup in 2012? Second, what’s on the bottom right of the chart? Are there really and “respected pollsters” that do “fake polls”?
“Are there really and “respected pollsters” that do “fake polls”?”
Public Policy Polling still gets taken pretty seriously by the media. While they don’t do “fake polls” in the sense of just making up data, they do perform a lot of silly polls with loaded questions designed to make Group X look stupid, ignorant, etc.
+1 Samuel.
Once upon a time, I worked for a pollster, and they do mostly whatever it is their clients pay them to. They will refuse outright dishonesty, but if you hand them a script of loaded questions or a plan with weird demographics, they will execute it to the best of their ability. After (assuming you paid for it) they’ll give you glorious 4 color reports and analysis, most of which is usually circular-filed by the client.
It may just be a coincidental use of the word Delphi, but I am wondering if Delphi Analytica is called that because it uses the “Delphi method” of forecasting developed by RAND a few decades ago (https://en.wikipedia.org/wiki/Delphi_method).
It could be but I’d bet on the Oracle at Delphi which I would assume is where the RAND Corp. got the name for their technique.
By “coincidental” I was thinking of both being named after the oracle.
And the data base company.
My conclusion based on the evidence so far: “Delphi” has become an over-used word.