Unlikely Voter

Poll Analysis and Election Projection

Daily Kos and Research 2000 settle

Last June the big shock in the polling world was when Daily Kos dropped Research 2000 as its pollster, and then convincingly demonstrated that Research 2000′s polls were faked. So it’s not surprising that in a settlement now reached, Research 2000 is paying Daily Kos cash installments, despite admitting no wrongdoing.

I certainly was convinced. The Daily Kos report, still online after the settlement, leaves no doubt in my mind. Too many coincidences. Too many convenient patterns. Randomness cannot be eyeballed, and the human mind is great at detecting patterns that don’t exist, but the analysis done on the Research 2000 polls was strong. That’s why Del Ali had to admit that he was fudging numbers, claiming that it was fine because it was within the margin of error.

It still baffles me that a supposed professional pollster could say such a thing, when it shows a lack of understanding of how the MoE works. The Margin of Error is not a uniform band. It’s a bell-shaped curve with a peak at the actual result. The numbers at the fringes of the MoE are not nearly as likely as the numbers immediately around the actual result!

But it’s good to see a fraud of a pollster being made to pay money to his victim, the paying customer who demanded scientific polling but got guesswork and opinions.

Comments

No Responses to “Daily Kos and Research 2000 settle”

Write a Comment

Comments are closed.