Grading the polls

While we like to say — accurately — that polls are merely a snapshot in time, let’s face it: We view them as the predictors they aren’t. So we can judge who nailed it, and who didn’t. Keep the margin-of-error in mind.

As I type this, it’s Dayton 44%, Emmer 43% and Horner 12%. Compare:

Poll/Date Dayton Emmer Horner Grade
Pub.Policy -10/29 43 40 15 A-
Survey USA (10/27) 39 38 13 B+
MPR/Humphrey (10/25) 41 29 11 D
Star Tribune 41 34 13 D
St. Cloud State(10/21)Likely voters 40 30 19 D
St. Cloud State(10/21)All respondents 37 27 18 D
Rasmussen (10/20) 44 41 10 B+
  • Todd Bergeth

    This shows very simply and elegantly. Don’t trust the very liberal pollsters like MPR, they skew their questions and their poll subjects toward their candidates. They do this to try and fix elections, it didn’t work MPR, nice try.

  • kennedy

    Seems to me the most recent poll was also the most accurate poll. Go figure.

  • Anonymous

    I think you’re shortchanging Survey USA. Yes, its estimate of the actual percentages for each candidate aren’t as close to what happened, but that’s likely the result of listing about 10% as still undecided, which does not seem unreasonable. It seems to me that the predicted margin is more important to a poll than the percentages, and SurveyUSA was was the only poll that got the narrow margin between Dayton and Emmer right.

  • JackU

    I picked nits with numbers yesterday on “what was important” from Select a Candidate, I guess I’ll do it again here.

    Bob, this time I’m going to be blunt, your grades stink. The last poll on the list gets a high mark because only 2% of the respondents are not accounted for in the result. At the other end of the spectrum is the MPR/Humphrey poll and the SCS (All voters) poll with 19% and 18% of the respondents unaccounted for in the numbers. All this proves is that in the last days of the campaign more undecideds went for Emmer than for Dayton.