Monday, February 14, 2011

Frank Graves Accuses Tories of "Byzantine Scheme"

Frank Graves seems to think he's under attack by the Tories because he released a poll that was not consistent with the Tories own internal polling and Frankenstein isn't happy. I'll admit that it is rare for a political party to come out and disagree with a poll that shows them with a substantial lead in majority territory, but nothing in the Conservative statement actually accused Graves of doctoring his results. What they said was “In the past pollsters have sometimes reported support for our Party that is unusually high relative to the prevailing data, only to have the anomaly corrected in a subsequent poll, giving the artificial impression of negative momentum”.

I don't think that EKOS manipulates its results, but I do question the randomness of their poll solicitation methods. While EKOS does have a tendency to produce outliers, their most consistent overvaluation is of support for the Green Party. The CBC pays Graves to produce weekly polling, and I'd like to know more about how they collect votes. Does every Canadian have an equal probability of getting a phone call from EKOS? Do they rely on voluntary internet polling? Nanos has the Green Party consistently at about 30% less support than EKOS does. What is the reason for that systemic difference between Nanos and EKOS? It has to be in the method. So what's different about their method? From Sept 1st 2008 to Nov 2008 EKOS averaged the Green Party at 10.8%. Nanos averaged them at 7.7%. In the October 2008 election, the Greens scored 6.8% of the popular vote.

Where I most often find myself disagreeing with Graves is not necessarily his numbers, but his analysis of his own data on Canadian political news shows. Anytime he reports a drop in Conservative support on the Soloman Show, his face lights up like a Christmas tree of excitement. I disagree with his punditry more than his actual numbers, so frankly I don't have a problem with the Tories sending out an internal memo telling their people to curb their enthusiasm. They didn't say this in a press release, they circulated an internal memo which was pounced on by Taber and friends in the media. This is the guy who advised the Liberals to start a culture war against old men in Alberta. He is as much a pundit as a pollster, and there is nothing wrong with disagreeing with a pundit.

I could run a poll on my website that asks which party you will vote for, and the Tories will score 90%. Does that mean that the Tories will win 90% support in the next election? No, it means that the people contributing to my polls is not a random sample of the Canadian population.

6 comments:

  1. I have always thought of poles a just a bunch of figures. As such everyone knows that "figures don`t lie" but to many forget the second half of the saying "but liars can figure".

    People that read poles to me are people that lack confidence in themselves and need to see witch way the wind is blowing to help them make up their minds.

    ReplyDelete
  2. To a degree polls can be helpful, but my take is how the reader perceives them and precisely what action should be taken once they have had time to disseminate the information. There is an unknown inasmuch as the margin of error becomes a critical though small number, which hovers between success or failure (the razors edge).
    Should this "unknown" number become spot on, the pollster would become a hero but the odds are this will never happen. There will consistently be that minute margin of error, nothing is 100% efficient or correct.
    No amount of scrutiny, homework or studying will ensure a guaranteed triumph and in most cases close is not good enough.
    Like Maxwell Smart said; "missed it by that much!" (:>)=

    E Mac

    ReplyDelete
  3. I once heard Nanos explain that a big difference between his method and that of most other pollsters (including Graves), is that Nanos does NOT prompt the person with a list of parties to pick from. The person must come up with a name--Conservative, Liberal, whatever--ON THEIR OWN. This probably explains Nanos's better record on pegging Green support.

    Calgary Junkie

    ReplyDelete
  4. Game on Iceman,
    the Bloc 'formalized' their participation in a coalition yesterday, over to you Jack and Iffy!

    http://www.cyberpresse.ca/actualites/quebec-canada/politique-canadienne/201102/14/01-4370300-le-bloc-relance-lidee-dune-coalition.php

    ReplyDelete
  5. goto 27 min mark on CPAC to watch the discussion of the Bloc's coalition move

    http://cpac.ca/forms/index.asp?dsp=template&act=view3&pagetype=vod&hl=e&clipID=5087

    ReplyDelete
  6. But wait, aren't all voluntary non-coercive surveys automatically skewed?

    Obviously Frank Graves should be given absolute legal authority to conduct mandatory polling which would alleviate any random sampling problems.

    ReplyDelete