What went wrong with the projections? - Macleans.ca

What went wrong with the projections?


Eric Grenier explains where his system failed.

ThreeHundredEight.com projected two things: the popular vote and how that would transform into seats. It failed at the first, which means it failed at the second. 


What went wrong with the projections?

  1. Hashtag fail?

  2. What was the cause of the airliner crash?

    Well, Sir, first everyone assumed they were on a cruise ship. After that flawed assumption, everything else kinda took on a life of its own…

    • lol

    • #winner

  3. Does he have three hundred and seven more excuses?

  4. Well when even pollsters warn you not to believe polls, you know the cards are still in the air.

  5. The polls were all over the place too because nobody could correctly model the left-right split in Ontario. In the end, the fear of an NDP government was just too much for some on the centre and the right, and Iggy was too right wing for those on the left.

    • Or the pollsters are biased against or don't account for conservative support. The Conservatives have outperformed the polls in just about every election they've been in. Rob Ford decimated the pollsters' predictions in the Toronto mayoral race. Funny that.

      • I think part of that is "please the pollster" bias. In some circles in this country, it's not considered trendy, fashionable or socially acceptable to be, or to be seen to be, conservative. It's well-documented in polling studies and literature that some people are reluctant to tell pollsters that they're going to vote a certain way if those people perceive that that's not what the pollster wants to hear.

        It reminds me of university, where left-of-centre students are often very vocal about their political views, constantly speak up in class about their views, and tend to be visible and active in campus politics. Right-wing students, not so much.

        • But shouldn't pollsters then take that bias into account in their sampling? Instead, they seem to bask in feeding into it, or even use it as a form of push polling.

          • Social desirability bias (or the Bradley effect as it is sometimes called) is really not a measurable bias because it changes according to the current climate of public opinion about the subject at hand. Some pollsters do count this as one reason to use robo-calling services instead of live operators, or for the use of internet polls. I actually doubt that it is very important in this case, more likely is a failure to capture turnout effects.

          • To clarify I don't mean its existence can't be measured in other kinds of studies, I just mean that it can't be measured in such a way as to fix it on a poll to poll basis.

          • But wouldn't any pollster worth their money, or the profession in general, be interested in dealing with this anti-conservative bias? Instead, they seem intent on leaving it alone, or even use it as a form of push polling. Reason #1488 to despise pollsters?

        • It reminds me of when I was in BC when WAC Bennett was in power. It was not considered proper to support the Socreds, who were too populist for the elites but Bennett kept winning.

      • But the 'split' means that the Conservatives benefit from the vote divided up amongst the opposition. He's talking about the modeling of that. So just measuring Con support doesn't help to predict. Are you paying attention at all?

        • No. He's saying the polling was wrong. Graves had the Tories at 35% and weak minority. They got 40% and a strong majority.

          Do you care about the truth? Of course not. Next.

  6. the question isn't why such a mechanism was wrong … BUT … has it ever been right ? – if one goes back and actually look at pollsters predictions it is amazing at how many times they are wrong – the poll will always say 19 out of 20 or something BUT you actually look at the history and invariably it's wrong more oftent than not – maybe some things that have too many variables are quite simply beyond predicting

    • The one thing I would say about Harper is that he has had some amazing luck with externals factors that have helped change the channel. The death of Osama the evening before voting without a doubt helped the Conservative cause (though to be clear, by no means am I suggesting that they would have lost without that event). The luck is at times so uncanny, I have to naturally suspect witchcraft… ;)

      • I think what clinched it was when Harper claimed to have killed Bin Laden himself, flying to Pakistan, breaking into his compound, strangling him with his own bare hands, dumping him in the ocean before finally flying back in time to cast his vote in his own Calgary riding. By the time people realized he was making it up, they had already voted.

        • Brilliant !

        • The only problem was Obama beat him to air – b*stard!

      • External factors such as the lety media and bloggers here also helped Jack , or is it John, by diverting attention from his massage (which he and his wife both admit happened), by shooting the messenger.

  7. The only thing predictable about polls is the act of pollsters explaining why their polls were wrong the day after the election.

  8. Three thoughts…

    1. FPP will always make it difficult to precisely predict seat outcomes on the basis of national – or even regional – polls of voter intentions. Percentages and absolute ballot counts will never perfectly tell the tale.

    2. Almost a sub-point to the first: In tight races, where the vote is being split two or three ways, the MOE rears its ugly head. We tend to look at the precise numbers and forget the potential permutations the margins can provide. I think it's human nature to simply the picture painted by polls (at least in a culture that often treats the quantitative as less worthy of skeptical assessment), but we need to remember there's always some messiness and imprecision beneath the top line figures.

    3. While I'm not thrilled to see a Con majority (though I laud them for their success, quite sincerely), four years without the constant noise of polls and their potential sparking of an election might be worth the trade-off.

    • EDIT: "percentages will never perfectly tell the tale of absolute ballot counts" sorry….

    • I acknowledge the 2011 Canadian election results with a glance back towards 1993 when the Progressive Conservatives, after winning back-to back majorities in the Commons (one of those by an historic mega-majority) engaged in a public act of political disappearance. Brian Mulroney, a well spoken bilingual politician who held not only Ontario but Quebec in his majority, not only retired but took the party with him. It was the last time the PC's held a majority wrapped in a personality. The personality left the stage drowning in a sea of corruption and the party ,bereft of any other raison d'etre folded up and died. The conservative right enjoyed a time in the political wilderness to reconstitute itself until early 2006. Look at them now.
      The Liberals, in a very different and yet eerily similar way, are entering their own reckoning. The era of party leader's personalities dominating Canadian politics is gone. Not that it couldn't be revived. But just look at the competing party leaders from this election and remember back to Diefenbaker, Trudeau, and Mulroney (OK, Pearson doesn't show up here). The only old- time "personality" is Layton. One of the differences between the Liberals of 2011 and the PC's of 1993 is that the Liberals still exist. They must address this question to the satistaction of the 2015 electorate:" Why do we exist and what will we do that differentiates us from the Conservatives?" If enough of the Canadian middle class is satisfied with the state of the economy four years hence and all the ramifications that implies, the Conservatives will form another government. If the Liberals engage in the same soul-searching as the conservatives did, either by themselves or with some or all of the NDP, and come up with a compelling narrative for a campaign, Stephen Harper will be seen as the exception rather than the rule.

      • And I suspect that the nature of the Quebec win is "just passing through" just as the PC sweep under Mulroney was a "just passing through" deviation from a constantly liberal Quebec.

        "Too soon ve get old, too late ve get schmart."

        Seriously, I think the sponsorship scandal soured a lot of former Liberal voters. It sure did me.

    • Except the pollsters got the percentages wrong too. Which meant even in a pure PR system, they'd have been off, though not quite so wildly off.

      • Not so, NANOS was 3% out on the conservatives b0ut otherwise pretty accurate. The problem I think is transferring popular vote on a gross scale to seats, which operates on a micro scale.

    • If they can't get it right stop publishing the reports, which I think have a tendency to push people in a certain direction.
      In this election they couldn;t face up to the fact that PM Harper was on the way to a majority.

  9. I can only hope that Frank Graves, in particular, is laughed out of the industry.

    • Didn't he use the term "undeniably accurate" for his last laugher of a poll?

      • That's the funniest part.

        His words:

        "it appears that what was previously thought unimaginable has become a point of consensus."

        "As shocking as those numbers would have seemed just a few weeks ago, they are now more or less stable and undoubtedly accurate. The only real question remaining — and, frankly, this is more a matter of anecdotal curiosity at this stage — is how these translate into the new seat distribution in Canada's next Parliament."

        He sounded absolutely certain. No talk about his shoestring budget. Just absolute undeniable certainty about the accuracy. Sort of like the global warming zealots.


        Now he says:

        "We were almost spot on about the popular vote, but where we really screwed up was with the Tory majority"

        as if that makes any sense whatsoever, and also:

        "This is done on a shoestring budget, and while we do our best to try and understand our failures it requires analysis and time and nobody is really interested in that"

        That's exactly the worst thing you could hear from a pollster.

  10. ThreeHundredEight.com – Powered by Oracle.

  11. Asking bias leading questions for one.

  12. Garbage in, garbage out.

  13. probably some conservative vote tampering! considering the past behavior of s h…with all respect to the elected persons who now have giganticus and outrageous egos

    • Of course. If you don't like the election results, complain about the legitimacy of the voting system. Vote mobs, on the other hand, are perfectly legitimate. But the regular old-fashioned polling stations? My God, they're hornet's nests of corruption and graft!!

      • Vote mobs – do you even understand what they are? Doesn't sound like it. But there were voter suppression robocalls – an old US tactic. We had bogus calls last time in Saanich and the Gulf Islands.

    • @delford t louis,
      You are correct about vote tampering, but your in the wrong era!
      And yes, due to the poor I.D. process on the municipal level, yes, some ppl were sneaking back in to double vote, as we now see in Toronto & with court approval, certain wards will re-elect MP's.
      But the serious vote tampering that you speak of is more or less a thing of the past.

    • Aw com'on!!

  14. To be fair to 308.com, it's worth pointing out that inputing the actual regional vote percentages into his model produced fairly close seat projections. So seat projection is possible, if the polls are accurate. Which they are usually not.

  15. From now on, I'm going to forget looking at polls and just listen to what Chantal Hebert is predicting.

    • Colby Cosh nailed it too. He predicted 160 seats. Only 7 seats out.

      • Chet was closer!

        • Really, must you force us all to eat humble pie? I actually like Chet, I just tire of his repetitive posting on the same subjects. Mind you, I'm tired of my own repetitive posting on the same subjects, and no doubt others are too. Perhaps time for a few months off. :)

          • You like chet? I used to like Emily.

            Spoiler alert: the novelty wears off.

          • I used to like Emily.

            Never had that problem. :)

          • Nor me;<)

          • I can't say I've ever had that problem either.

          • Stop the predictalbe polling results……..marked with an x

          • hahahah, yes, we all need to change the channel, take a deep breath & relax knowing that PM Harper will get the job done. :)

            We've all had quite the adrenaline pump these past couple of weeks. Royal Wedding, bin Laden's death,
            Conservative Majority, NDP as opposition & the death of Libs & Bloc.
            Yep, we sure have had a bumpy fun ride, haven't we! hahahah

  16. I don't know how much money the various polling firms spend in an election, but wouldn't it make more sense to do one, or a few polls with an accurate sample of each of the 308 ridings than a bajillion national polls which are then inaccurately mapped to the 308?

    • People change who they want to vote for over the course of the election (see: NDP, Quebec). you'd have to do polls of ridings only near the end of the election period to be accurate. But if you did that, what would we talk do for the other 5 weeks of the campaign? Debate policy? Not bloody likely.

    • The margin of error of polls is based upon the number of people sampled. Most national polls call 1,000 people. To get equally reliable results for each riding you would need to call 308,000 people. Obviously this is much more expensive than producing national polls.

  17. One of the pollsters interviewed (I think it was Frank Graves) cited the cost issue, and I get that. I appreciate him being honest about that, at least. Having said that, Nanos' popular vote numbers were reasonably accurate. I think some of the high Conservative number was definitely the result of older people voting and centrist and free enterprise voters strategically voting Tory to keep the NDP from getting the keys to the car.

    • I wouldn't want some of the wacko NDP elected in Quebec getting the keys to the car, such as a candidate being away on vacation, and not being able to speak French, waitresses, masseuses,etc.
      As far as Jacko winning all those seats is concerned it only means more money for the NDP and less power.
      With the majority, when Jacko and his band of weirds start making noise, Harper can simply tell Jack to go get a 'massage'

  18. Re 308.com, I paid particular attention to their projections for two ridings, Vancouver South and Vancouver Quadra. And in both cases, their local projections throughout the campaign seemed off to me. They had Joyce Murray AND Ujjal Dossanjh both winning very comfortably (i.e., by 10 to 20+ point margins). Murray only won by a narrow margin, and Dossanjh lost. So I'd be interested in knowing how they came up with those numbers.

    • They lied.

      • Then post your real name.

    • 308 had CPC at 19% in Labrador, adn Yukon at 24% thats 24points behind lead, yet the CPC won both
      where did they get those # from?

      • Small sample sizes have ridiculous margins of error. Extrapolating national or provincial polls into ridings is a mug's game. We do it for the same reason we look at long term weather forecasts. Because we can. Not because the results are accurate.

  19. This is because the "pollsters" nowadays are nothing but propaganda machines. Trace to few elections back and you can't fail to notice the bias and misleading pollings. What is surprising is why the media keep wasting their money on them? Is it to to perpetuate their own biases?

    • Refreshing that after winning a majority government, largely succeeding in the secondary task of hobbling the Liberals, and getting the endorsements of ever major Canadian media source that offered one except the Star, you still haven't given up your devotion to playing the victim card.

      • Whinnig even in victory -must be in their DNA.

  20. Projections are nice and all, but why is anyone at all surprised they all got it so wrong last night? Basically all pollsters and predictors admitted that their models aren't set up to deal with the unprecedented happening.

    Really, if anyone actually manages to get a working crystal ball, please let me know. I'd like to borrow it for a bit.

    • Chet has one he might not be using anymore.

      • I think it would be safer to wait and see before asking to borrow it. I have a suspicion it'd tell me to buy IBM, or to bet on Winnipeg, or perhaps on the Jays based on belief rather than their actual prospects, you know?

  21. I see two problems. Firstly, seat projectors used old data (discounted somewhat), which simply didn't reflect the situation on election night. Grenier increased the polling decay rate, but even that wasn't enough. As a result, seat projectors are usually behind, and overestimate declining parties (eg. the Liberals in 2008 and 2011). We also lack a good mechanism for figuring out where a swing in the vote is going to happen – uniform swing is a clumsy substitute. Simulation models can help with this (calgary grit's simulation model was tied for second, as far as I know). Demographics may work as well, and of course, sub-regional polling would be of great help.

    Secondly, at the level of pollsters, I don't think we have good or consistent ways of screening for likely voters. Ekos was particularly harmed by this. While their topline results were way off, here is what you would get if you just looked at voters that indicated they would not change their minds:
    CPC: 38.64%
    NDP: 30.05%
    LPC: 18.94%
    Green: 5.1%
    Bloc: 5.7%
    Other: 1.5%

    This is almost exactly what we saw on election night. Of course one problem is that applying a likely voter screen of a particular level of strength may not work the same in every election. In a municipal election, where few people vote, you would want a very tight screen. In a high turnout election, you might want a loose one. The varying intensity of support among Canada's political parties are a more recent development, so the polling industry may not have had sufficient cause to adapt its assumptions.

    • Perhaps EKOS tried to be too fancy and applied too many fitlers.

  22. Angus-Reid beat Nanos, and was the most accurate pollster for the second time in a row. Of note, they are an online pollster (though they were way off in the UK election).

    • Really? I never paid much attention to their polls. Maybe the media never covered them much because they were the only ones predicting a Conservative majority.

    • Only because Angus Reid rounds his numbers & Nanos doesn't

  23. One other thing – Grenier did much better than the election prediction project (and 308 was on the low end of the math-based models).

  24. Why are people still surprised that pollsters under estimate Conservative support by 3-5% during elections?

    When any attempt to point this out was made during the election, the lefties went giddy with taunting and mocking about Con-Bot conspiracy. The usual rants about “Only 30% of Canadians support the evil Harper. He'll get another minority and he'll have to step down!” typically ensued.

    Yet again we see it proven. I predict the same stupidity during the next election.

    • I smile when I read that. You better believe it. The progressives went out of their way to rub those polling numbers in our faces. My cheeks are sore from grinning so hard for the past 18 hours or so.

      • Yup, now they are all saying Conservative supporters should be humble in victory.

        Funny, they weren't all that humble when they thought a coalition was a sure thing.

    • I don't recall anybody being so confident as to insist that Harper would for sure only win a minority, but in any event, I stand corrected. Only FORTY per cent of Canadians support the evil Harper.

      • In actuality forty percent of the sixty percent of the eligible voters supported Harper. We do not know who the other forty percent of the eligible voters who did not vote support. It is highly likely that some of them support Harper as well.

        • Oh no, I'm sure they're all progressive lefties. Every last one of them. And they didn't vote because of Harper's thuggery. It's Harper's fault. It's always Harper's fault.

          • Haha typical lefty responses.
            Maybe we should blame Mike Harris. Mikes only been gone for 8 years, or maybe George Bush.

  25. I'm amazed at the comments on this thread, and what is obviously being missed from a standpoint of basic statistical theory. A sample of voluntary responses is prone to bias compared to a mandatory sample.

    People are posting about some pollsters being more accurate than others – no one is fundamentally better at polling than anyone else. It is all just random variation and is prone to the biases of voluntary samples, and to the extent a pollster gets the results right with much accuracy it is mostly by luck, not their methods.

    This is why we used to have a mandatory long-form census. To ensure non-response bias does not occur.

    To get the data right.

    And yet some here will shout how the pollsters got it wrong and are a bunch of hacks, and in their next breath can say without shame a voluntary census will be accurate.

    Wake up.

    • Speaking of the census, I got it today. Handy reminder of damage Mr. Harper has already done to the country.

      • You're that disappointed you won't be getting a long form this year? Personally I think he made the wrong decision, but damaging the country is a little rich. Maybe he sees a diminished role for the federal government, so that extremely detailed census-taking is no longer necessary. At least that's what I'm hoping for. The provinces already manage anything that is important. Let the feds focus on immigration, the military, foreign affairs and criminal law, and let the provinces and munis do the rest. They can take their own censuses if they wish.

        • If he sees a diminished role for the federal government, extremely detailed census-taking is *more* necessary, because it would be good to know what type of effect that has on the people of the country. This is true with any signficant change.

          Unless, of course, you simply don't care.

        • Get a long form, Raging? The chance of getting one was once in 25 years. It's called sampling. I'm 60 and I've yet to get one. But I believe if I did get one, I could manage to fit it into my busy schedule. Every modern progressive country has a national census. I believe we're even obligated to conduct one in order to belong to the OECD. The idea that census data isn't needed for things like immigration and the justice system shows an appalling lack of knowledge of how census data is used. A serious fiscal conservative would want as much reliable data as possible in order to cut waste. Lots of other bodies conduct surveys, but they rely on census data to do it..

          • I'm 48 and I or the house I lived in with my parents has received the long form in every census since 1971. I'll find out in June when the NHS is sent out if I'm again one of the lucky "winners".

          • Had two of them so far. They were no big deal. But I was negligent in turning them in on time. A nice lady from StatsCan phoned me both times (not the same lady) and just interviewed me over the phone and had the whole thing filled out in half an hour. It was painless. My mom hated the long form with a passion, but she's like that. Personally I don't have a problem with it either way, but I don't see it as being all that vital either. The short form is still mandatory. Just got mine in the mail today in fact.

    • You are correct Matlock. If pollsters were guaranteed to be right, we could get rid of elections. Personally, I like the old fashioned election better than any polls, and the people, not the pollsters, always have the last word.

      • The larger point I was making was about the census rather than polls.

        What these election polls demonstrate is that with a self-selected sample (e.g. a phone survey), the data gathered may be prone to non-response bias, where as a sample that avoids self-selection (i.e. mandatory participation) is not prone to such biases and is more likely to reflect the true underlying distribution (e.g. election results).

        This is why we had a mandatory long-form census. This is why I continue to be amazed by this decision to this day.

        If census data is worth gathering, collect it properly. If the government deems it is not worth the intrusion, then cancel the long-form altogether. A voluntary census is the worst of both worlds – garbage data and all the costs of a mandatory census.

        • True enough. Perhaps they'll just can the long form census. That would be the best option.

    • That's not true. Some are better than others. Calling 700 numbers in downtown Toronto won't give you a good poll. Asking questions in a threatening manner won't either. There are plenty of variables. Obviously some pollsters do a bad job – just like any other industry.

    • Garbage in, garbage out. If you think people are dishonest with a voluntary poll, they are even more likely to lie on a mandatory poll, like a census. So forcing people fill out a long form doesn't make it any more reliable, any more than forcing people to answer a telephone survey. So either way, the census is not particularly reliable and should never be used as a decision making tool. Especially since governments at all levels have much more detailed and current information on hand.

  26. The conservatives benefit from having more "likely voters", for multiple reasons. So you could expect a few point bump, and most pollsters were within reasonable range. But Ekos was so far off (last poll 34 -32, election 40 -31) that they should be laughed out of the business. And his excuses of "shoestring" budget are a laugh, because obviously the other pollsters manage to do better. Graves is also the one that pretends his results are bang-on with statements like "undoubtedly accurate", while he goes on CBC and provides analysis that has no basis in fact.

    • I'm thinkin' Graves is polishing his resume right about now. And he's probably downplaying his polling history.

  27. Anyway, I'm still waiting for the long form census to be the nail in the coffin for the conservatives, as all the leftists on this site were saying for months. Still waiting. When I would tell them that it would have absolutely no effect, they would always laugh and toss insults.

    • I predicted it would have no effect… but if you want me to call getting rid of the long form census anything but stupid, well, I remain quite unrepentant. It was all kinds of stupid.

    • My short form census arrived in the mail today. What a delightful dessert.

      • I tried doing it online. It gets stuck on 'number in household'. Another Tony Clement project.

        • I did mine online today. Worked like a charm.

          • Mine worked fine as well.

    • I certainly was much more comfortable filling out this census than the last one.

  28. If pollsters were always bang on, we wouldn't need elections. That's why we have them. Otherwise we could just hire pollsters at a fraction of the cost and do away with this silly democracy stuff. I'm quite glad we actually still rely on the old fashioned method of marking an X on a paper ballot. And I'm quite glad that this primitive exercise always has the last word that proves most pollsters wrong.

    • Pollsters now on the Con enemies list? Suck it up – you won!

      • Who said pollsters were the enemy? Did you wake up retarded again?

        • You are the one going on and on and on about them. Do you find yourself having short term memory problems?

          • What were we talking about again? I forget.

  29. Pollsters are snake oil salesmen employed by corrupt media organizations in order to manipulate and deceive the electorate. Pollsters are always wrong in their predictions, but their use comes when they enable the media to give everyone their "expert" opinions on the reading of the polls, and the possible outcomes that the polls suggest. Polls are for dogs and strippers.

    • How about a simpler explanation? They were wrong. Some overestimated Conservative support. Most others under-estimated it. It's a very inexact science, and it is not surprising that some got it wrong. Some were close. They aren't likely corrupt or even biased. Just working in a field that isn't all the accurate.

      • But the parties use them – how do you explain that?

        • We also sometimes plan weekend events around long term weather forecasts, even though they aren't much good. Pollsters are the best tool available. They aren't perfect, but they aren't completely useless either. In the absence of crystal balls, we rely on pollsters. A little too much it seems, but there it is.

  30. Small 'c' conservatives, men, and older people are more private. They hang up on pollsters, or tell them to f$%k off. Lefties, young people, and chicks can't wait to tell the world how they are going to vote…and how much more wonderful this makes them. This is why polls always short-change the vote for conservative parties and inflate the vote of left-wing parties.

    • 'Lefties, young people and chicks' – the Harper majority has begun in earnest. Only 4 more years of this.

      • I believe they were saying when Harper first got in in 2006 that, "Thank God we've only got a few years of this." Read Paul Wells today. "The victims of Harper's resilience are piling up." There's the thing you want, and then there's reality. Harper won't last forever, but he's had some impressive staying power.

    • I spent a good portion of today listening to progressives talking with grave concern about the election results. Couldn't wait to share with each other. I remained silent – as I always do on political issues at work – until someone dragged me into it. When I said how I voted, one guy – who is usually quite friendly and reasonable – told me I was part of the "stupid 40".

      This exchange seems typical. Progressives are wonderfully warm and friendly as long as they think you're one of them. When they find out you're not, they call you stupid and worse. I merely smiled and told him that while I may be stupid, reminding me of that fact is not likely to make me come around to his point of view. Left him with nowhere to go. How does one respond to something like that? :) He blushed so red I could almost feel the heat coming from him.

      This is a big problem for progressives. They think everyone is like them, because most vocal people are like them. The conservatives tend to be very belligerent online, but socially very quiet about their views. There are exceptions of course. But I rarely express mine publicly. (British pollsters called this the "Shy Tory" effect after John Major's surprise majority victory in 1992). Unlike progressives, we don't eat, sleep and breathe politics and government. In fact, most of us go long periods of time without thinking of it at all. Unfortunately for progressives, we aren't shy about voting when the time comes. And we aren't likely to vote for those who dismiss us as stupid. You'd think they'd clue in sooner or later. Evidently, it will be later.

      • "Unlike progressives, we don't eat, sleep and breathe politics and government. In fact, most of us go long periods of time without thinking of it at all."

        You say that as if it's a good thing, though.. that's what confuses me. Isn't thinking about how our government is handled something that all of us should be concerned about.. and not just come elections?

        Interesting point about being belligerent online, but not socially. Why the difference? What makes those online less worthy of consideration?

  31. Comparison of projections here:

    Although none of the projections did well, ThreeHundredEight.com was significantly worse than all the others because it used a very slowly moving poll average, which necessarily fails when there's a large shift.

  32. Not quite the same spot as Bob Rae in 1990. Thankfully. Rae had a majority.

  33. Another reason is blind media leading blind Liberal insiders/handlers/hacks. Liberals listen only to but* lickers (pardon my language).

  34. I wonder why pollsters don't have a control sample? Each poll taken is with a different group of people and these groups are compared to one another. There is actually no common reference point between the groups. The pollsters could set up a large control group (say 3 times the usual sample size) and interview them on an ongoing basis during the campaign. Their regular polls can then be compared to the control group. Comparing how the control group changes versus the regular polling could highlight trends and such.

  35. CBC had all along predicted a majority for the Conservatives.
    so not all projections were wrong. Just yours.

  36. Pollsters and media and biased toward the liberal and NDP agenda…they do not understand the pulse of the Canadian people.

  37. I suspect if they other 40% supported any of the leaders they would've turned up at the polls.

  38. You do know that the 8-page version is the SHORT form, right?