06.11.2012 08:11 AM

308, still getting it wrong

Remember 308.com?  They were the ones who, repeatedly and without qualification, declared that Wildrose would win a huge majority in Alberta.

Now, they’re making similarly bold predictions about Ontario, here.

Why is this bullshit still happening? One, because some (not all) reporters are lazy, and they couldn’t be bothered to cease doing what they have always done.

Two, because many (not all) pollsters, like the media, have made a series of bad decisions over the years that are now coming back to haunt them.  They’re getting things wrong, as in Alberta, all the time.

Caveat emptor.

18 Comments

  1. Paul says:

    Three, many households don’t have landlines and most don’t answer calls from people they don’t know. Hence polling results are skewed by the subset they can reach.

    Only wackos like me answer and we love to give wildly divergent answers based on the day of the week to screw with the machinery then giggle while they scratch their empty skulls.

  2. sj says:

    Yes let’s remember that at one point in the federal election 308 had the NDP almost wiped off the map.

    • Warren says:

      Bingo. It’s junk science.

      • Olmanhall says:

        I still don’t beleive science is involved. I think these obviously wrong polls in Alberta were propaganda. Everyone likes to back a winner so these ‘polls’ were designed to tell people that don’t follow polotics closely Wild Rose was the most popular. Everyone will vote for them so you better too. Now the pollsters blame people for lying to them when in fact too many of us were paying attention. I’m betting this will happen in other places too. Maybe my theory is junk but can you tell me for sure this has never happened Warren? I beleive you have more experience with these things than I do.

      • Corey says:

        Warren: do you actually think Hudak and Howarth could come first and second if an election were held? I don’t think there’s a chance in hell that could happen. Polls can be wrong. VERY VERY wrong, as we all know. (For one thing, pollsters last fall made everyone underestimate how effective Dalton McGuinty can be in an election campaign).

      • Sean says:

        I wouldn’t even qualify it as “Junk Science”. Its nonsense – its guessing.

  3. Dan says:

    In that site’s defence… he really doesn’t do much other than add up all the polls that are out there and average them out. Sometimes he does little more than analyze one poll.

    He doesn’t do the polls. But garbage in, garbage out. Blame Nanos/Ipsos/etc.

  4. JamesHalifax says:

    Perhaps pollsters would be more reliable if they had more incentive. Currently, they get paid regardless of the accuracy of their results…perhaps a caveat would remedy the shoddiness of pollsters. Something along the lines of, ” No payment if actual results are more than 2.5 times the margin of error reported by the polling company.

    • Dan says:

      Supposedly competition would weed that out. The “best” pollsters would get hired to do more polls.

      But that assumes that accuracy is what they get paid for.

    • GFMD says:

      It would be impossible to tell if the poll had been wrong or if opinion had actually shifted. Hudak and Wildrose probably WERE ahead by a wide margin…at one point.

  5. Linda white says:

    It kinda looks like they are after the Liberals and Justin. I know you like him but Ipsos and Global seem to the culprits. I think they are really scared of him. I hope he runs and beats the pants off them. Decima seems to be the only trustworthy pollster.

  6. Greg from Calgary says:

    In heard one of the challanges that was clearly demonstrated in Alberta was a lot of polling relies on phone calls to land lines as cell #s aren’t published. Since more young and progressive voters have cells and no land lines the end result is these voters are not factored into the polls. Anyone else here something similar?

  7. Chris says:

    I think Canada is too small/regional for the american prediction model to work well here.

  8. Volkov says:

    “Without qualification”? That’s a laugh – the polls are qualification, and they showed what they were. It is amazing that you attack Eric over that, when it is hardly his fault that the polls were incorrect. You can disagree with the assessment, to practically attack someone’s credibility when any rational mind would qualify his words, is just sad.

    Furthermore, in the case that pollsters do get it wrong, they’ll figure out why and change their methods to ensure it doesn’t happen again. That IS science. It is inexact and not perfect and will not get it right every time, nevermind the inherent randomness of it all.

  9. Richard Besserer says:

    Actually, 308.com did nothing of the sort. At the end the probable range for BOTH Tories and Wildrose were from 20 to 62—a tossup. The Tories won the toss, with their seat count at the upper bound of the probable range. 308 at least allowed for the possibility to the Tories could save the day—which is more than many newspapers bothered to do, filing their obituaries for Redford before the vote count had even started.

    308.com can hardly be faulted for not predicting a 15-point swing back to the Tories and accounting for it in their calculations. Hardly anyone did. With the “correct” numbers plugged in, the 308.com model does much better matching the seat count of the Tories and Wildrose.

  10. Tiger says:

    It isn’t junk science. Their results are as good as polls are — which is to say, results vary.

    People change their minds over the course of a campaign, people change their minds in the car on the way to the polling location.

    As we saw over the course of the last campaigns in Ontario and Alberta.

  11. bluegreenblogger says:

    Polls are polls, and when done properly they tell you what they tell you. There are margins of error, and type II error tells you the chances that you have of getting a representative sample. Methodology matters, like how you reach people and whether you asked the wrong question. Above all, they are about the exact question you ask. Since most of 308 source polls ask `Ìf there were an election today….`they are starting with a false premise. and since that is precisely the question that political junkies want answered, you all get what you asked for when you look at a voting intention poll. If you expect such a poll to predict election results, then you are bound to be disappointed much of the time. Polls are good at tracking movements in opinion though. Polls are MUCH better at telling you about what people think about things other than the highly subjective and often impulsive decision about who they gonna vote for.
    308 is pretty well junk science, but it is shiny and catches the attention of a great many political junkies. I think the comments section there is more interesting reading than the conclusions, but I do not waste my time commenting on pseudo polls. (except here of course, lol)

Leave a Reply to Greg from Calgary Cancel reply

Your email address will not be published.