Polls and Reporting III Chilenski Strategies
By popular demand*, I am picking on another pollster — Chilenski Stategies.
There is a poll which is getting a lot of attention. The Chilenski Strategies poll of the Missouri Senate race which shows Todd “legitimate rape” Akin actually one point ahead of Senator Claire McCaskill.
I have noted a large number of brand new pollsters whose polls have never faced the election night of truth. I have the impression that they have Republican house effects. I even speculated that one or more of these pollsters might be deliberately biased choosing a sampling strategy which gives good results for Republicans because they aim to influence the narrative. My original interest was how the relatively new focus on averages of polls (which I consider a great improvement) might have changed incentives. Back in the misty past (say 2004) an obscure pollster might be ignored. Now they are tossed in the average unless there is demonstrably a good reason to exclude them.
Disclaimer: the suspicion of deliberately bad polling is just speculation. Also even I don’t even speculate about each and every new pollster. I have a pretty strong suspicion that the integrity of Alex Castellanos CEO of Purple Strategies is not above reproach, but I have nothing new and based on solid evidence to say about anyone.
Now the rules from @robertwaldmann are
@mattyglesias mkt demand for R biased polls high indeed. Look at brand new robo-pollsters with “marketing” or “strategies” in their name. (Sept 27).
And we get a weird poll from Chilenski Strategies. Ok “marketing” or “Strategies” in the name.They are a bit brand new with 14 published polls and a mean squared forecast error of 0/0 = does not compute. yeah now that’s a link I like . Do they robopoll ? Hard to say from their main webpage (there should be a special category for pollsters whose web sites don’t even say if they employ human canvassers). Ah yes
Data Collection:The survey data was collected through Interactive Voice Response telephone interviewing. Interviews were conducted during the evening of August 8, 2012 between 5:30 to 9:00pm. The interviews averaged 180 seconds in length.
The interesting thing is that (like the dread Baydoun-Foster) they are listed (D) affiliated. I think this means that their client list (so far) is mostly Democratic. The joke about how Democrats have an extraordinary ability to choose poor strategies is too easy and unfair.
My current guess is that the new robopolling technology is doing to polling what twitter is doing to nuanced debate. It makes it very tempting to try a bit of polling and, who knows, with luck one might be the most accurate pollster (hint not Rasmussen).
* by popular demand I mean the large number of page views and comments on my post on Gravis Marketing.