Did you catch this line in the latest Leger poll?
“The numbers were reached after distribution of the 20 per cent of respondents who were undecided.”
Meaning that Leger divided up the “undecideds” in equal proportion to those who expressed a preference.
D.J. McGuire of China E-Lobby writes;
If I remember your home province correctly (Saskatchewan, right?), you folks actually lived through one of these.� Remember when Roy Romanow was cruising to his 3rd majority in 1999?� When he had a 20-point lead and the election was seen as a formality?� Well, come election day, the UNDs decided almost entirely for the Sask Party, which nearly beat Roy, and�the shock�pushed him into early retirement.� Germany just went through the same thing, where pollsters had the CDU up 10 points, after it wiped out (ahem, “distributed”) UND voters.� They broke largely away from the CDU, and now everyone is stunned.
�
Down here in the US, we just don’t poll that way.� The UNDs are not wiped out/”distributed”; their called what they are – undecided.� If an American firm was polling the survey Leger had, for example, it would look like this: Liberal 33.3%, Tories 20%, NDP 12%, Bloc 10.8%, Others 6.6%, Undecided 20%.
�
In other words, the Grits would look good, but there would obviously be enough voters out there to tip the balance.� The next�step would be to find out who the UNDs are, to figure out to whom they would lean.� The arguments and other hijinks would then ensue.� Instead, most Canadians who are�looking at that poll will see a Grit surge that just isn’t there.� This sort of polling is great for whomever is in first, but terrible for everyone else, and wrongly so.
�
Now, Martin et al will dream of a majority, more knives will be out for Harper, and western voters will get a good deal angrier, all because Canadian pollsters can’t seem to find the energy to examine, rather than dismiss, undecided voters.
The issue is discussed in more detail by Maurice Pinard, Professor Emeritus at McGill University in reference to Quebec polling.
I’ve never heard of a pollster with any integrity that would predispose of undecides by distributing them evenly over the other choices…this is the height of unprofessionalism and completely unscientific.
The insider money is ion the fact that the next parliament for the Liberals will, at best, look pretty much the way it is today.
All the desparate planted polls seem to telegraph the Liberal camp’s gloom.
Why the surprise? The same methodology was used by the Standard in their Western separation poll and those numbers were not considered juiced, were they?
Kate: Nice spill on a technique used by every major pollster in Canada. And it worked. WLM took the bait like a fat bass on a summer day. You really do play unmercifully with your people. What lemmings.
I took a look at the polls since the spring from Ipsos-Reid and Strategic Counsel – two firms I trust. The results will probably surprise you. Their polls are extremely stable and the gap between the Liberals and Conservatives is not really that wide.
Polls are cynically used as cattle prods for public opinion. They have been so abused by whatever interest groups that find them useful for their own ulterior motives that their once proper usefulness has been debased and their reliablitity increasingly viewed as falacious. As an old saying states, “Figures don’t lie, but liars figure.”
Don,
“Kate: Nice spill on a technique used by every major pollster in Canada.”
Yes, Don, it is used by every Canadian pollster.
That’s the problem.
So long as this remains the norm, not a single poll in Canada is credible. That’s the point I was trying to make when I e-mailed Kate. I’m sorry you missed it.
But let me guess, “distributing” the UNDs is the right way to go just because American pollsters don’t do it, right?
Votes don’t count- it’s who counts the votes.
(United Nations monitoring our next election- no thanks).
The ‘Big Lie’ in polling is calling uncooperative respondants “undecided”. Just because I don’t declare my intentions to a pollster doesn’t mean for a second that I don’t know exactly who or what I intend to vote for (against)!
However they dick around with the undecided, the important thing to me is that the two polls that I’ve seen -IPSO in August and this one- both have a high undecided – 19 or 20 %.
Well, I will grant that McGuire is an expert in the field of sophistry. However, if you really want to deal with polls you pay closer attention to the trend lines than the absolute numbers. Case in point: German election, Merkel had a twenty point lead a month ago. Polling had that lead being ruthlessly ground away. And that trend was confirmed in the election. Undecided in, undecided out, who cares? Watch the trend.
Leger on Saticfaction:
2. More than one Canadian out of two remains dissatisfied with the Martin Government:
In total, 53% of Canadians surveyed say they are dissatisfied with Paul Martin�s Liberal government as opposed to 40% who say they are satisfied. The highest dissatisfaction rates are recorded in Qu�bec (72%) and in Alberta (57%), while the highest satisfaction rates are recorded in the Atlantic provinces where the majority (52%) say they are satisfied with the Liberal government.
In Canada in the last fifteen years there has been an identifiable trend to declare oneself undecided rather than choose Reform, Alliance, and now the CPC.
My general feeling is that undecided at this point means leaning away from the liberals. Whether this translates into more votes for the NDP and CPC or just voters staying home, I feel safe saying the vast majority of these votes will not vote Liberal this time out.
If polls were considered accurate Trudeau could have stayed PM in perpetuity. Change is in the air but none of the vested interests want to acknowledge it.
Humm… 20% undecided. And WHY do you think they are undecided? Maybe it’s because the majority of them are torn between the Tories and the Grits? Or even the Grits and the Dippers? So, when Nov. 1 hits, the 20% makes their choice, and part of the Liberal base splits either way.
But, of course, the MSM delares “Liberals ahead, will form next Government”.
And the public perpetuates the myth that the Conservatives can’t win…
Yeah I’m too right wing for the CPC but I don’t want my undecided chp / western separatist vote going liberal.
Bring back Stock!
Nice post. It’s certainly true that not all polling companies follow this practice. But some – more than one – do. I personally don’t think it’s an optimal practice. It’s basically a form of crude imputation and should be treated with a grain of salt, as all polls should.
But, this should be said: Leger always does this. This is not something they have done this time to juice their numbers. So if you want to understand the significance of this poll, you’d best compare it to previous Leger polls to get a sense of the trend of Liberal support. Leger’s imputation technique will have a minimal effect, at best, on the trend.
Polls aren’t magic, and results are often influenced by the method. But, polling companies make their money from being accurate and making correct predictions. Indeed, most polling companies are involved in political polling for profile purposes – not because there’s money in it. They get in the papers to get work later on. So, ultimately, they rise and fall on their credibililty. It’s a point to remember before assuming they are either uncareful about their results or manipulative.
Christian conservative – if you follow the math of your statement through the Liberals still win.
I got worked up about this for a few seconds, then I remembered why I didn’t care. First of all, I got caught by a polling firm a few months ago and the questions were so Lib-loaded, it defied belief. And I also know that all the folks who have made the Tory coffers swell to $20 million know who’s leading the party, and are not being polled. The numbers are fallacies aimed to forcing the Tories into a stupid blunder like believing they’re losing under Harper, and misleading the sheeple of Ontario. I hope Harper does force an election for the dead of winter, because one thing is guaranteed and that is that every Tory worth his/her salt will drive through an ice storm to vote to dump the Liberals. The rest of the a-holes can stay on the couch bitching about Adscam, gun registry, EI misappropriation of funds, whatever. There’s likely to be a new scandal by the fall anyway.
Most pollsters do not make their money by being accurate. They’re paid to come up with a particular “truth” that’s then presented by whomever paid for the poll as “the voice of the people”. The hope is that the truly undecided will then have their minds made up for them by the artificially derived trends.
Don’t look at the results of the polls, look at their internals.
Iron Lady: here’s a challenge. Spend 10 minutes on Ipsos’, Leger’s, and Strategic Counsel’s websites and demonstrate how the questions they used in their latest polls are “lib-loaded”.
Jeff in Washington, explain to me why polls commissioned for the Toronto Sun and National Post – both conservative leaning papers – do not differ from Toronto Star or La Presse polls.
As a final note, I love how people take their own bias-driven inferences over polls which are at least theoretically unbiased. Taking everything we know about bias-confirmation and cognition, and everything we know about the power of simple random samples and inferential statistics, I can’t believe the arrogance of people who assert that they have better insight into a whole population than a poll – even if they do leave their living rooms.
“polling companies make their money from being accurate and making correct predictions. Indeed, most polling companies are involved in political polling for profile purposes – not because there’s money in it. They get in the papers to get work later on. So, ultimately, they rise and fall on their credibililty.”
I doubt it Peter. They work like all the other friends of the Librano$$$ .
Do the polling work for the Liberal Party Of Canada for free and they’ll give you stolen tax payer money like druken sailors after you help them steal the election.
What would make you believe that polling is different from advertising when it comes to the mobs,.. er uh liberal involvement?
Ok, I haven’t really followed the discussion on this topic yet, even the post itself. I wanted to write out my own pov without any influence of other peoples’ opinions affecting my own.
Here’s the post, just uploaded:
http://thecanadiansentinel.blogspot.com/2005/09/yet-another-loudly-trumpeted-poll.html
Now, I’ll read the post itself and the comments. Please feel free to make your own comments on my post right there on the Sentinel or here, whatever you wish.
Thanks Rich. Don’t let the facts get in the way. Tell me, when was the last time Ipsos-Reid or the Strategic Counsel did work for the Liberal party?
You’re willfully ignorant.
Peter, don’t you think a question like this on last week’s Strategic Counsel poll shows a very distinct bias:
“Should Alberta share its budget surplus with other provinces hurt by higher energy costs or keep it?”
Unfortunately, the other questions asked in this widely reported poll were not available for scrutiny but if they were anything like the one question that got out, I would be extremely skeptical about all of their other results.
Two things that should be reported with any published poll are the actual questions that were asked and who actually paid for the poll. I’m assuming any reputable pollster is competent at random sampling. I’m more interested in the biases revealed by sponsorship of the poll and how it affected the questions that were constructed. Almost any given week you will find some special interest group trumpeting their latest poll results that show the majority concurs with their views. And almost every one of these polls is pure garbage because the questions are biased from the start to obtain the results that they paid for.
I agree with Tim R. that when poll results are reported, the actual questions, verbatim, should be included along with them. Then the people can make a better-informed analysis thereof for themselves. Surely no one could rationally be opposed to this. I move that it become standard practice if the polling firms want to be taken seriously and get polling contracts.
Gentlemen:
it actually is pretty standard practice. You can find question wording for SC polls here: http://www.thestrategiccounsel.com/our_news/polls.asp
As for revealing the patron, here’s a little background on how most political questions get into polls. Typically they are just added in by the pollster on top of some survey of a non-political nature, most of which are usually for private consumption by the client. So, technically, it’s the pollster who is paying out of her profits. And they do it to get publicity.
None of it is as nefarious as any of you make it out to be.
Peter
Wow, the CPC gets less votes than the undecideds… KAte you shouldn’t expose this stuff.
That’s fine Peter and I was aware that most polling firms do that. Unfortunately, they don’t do it concurrent with the release of the results to the media. Therefore, they can spin the results just about any way they want and no one is the wiser.
For example, the last poll on the Strategic Counsel site is dated August 7th, a month and a half ago. Is that poll still in the news? Hardly. How many journalists will check back a month or more later to see if they’ve been snowed? And it doesn’t hurt that most journalists (Colby Cosh the exception) are innumerate.
Um, Peter, I meant that we should see the questions verbatim in the MSM reports, which isn’t really the case very often, as far as I remember, unless memory fails me. I’m sure the pollsters would want the media to do this. It doesn’t surprise me that the pollsters’ sites would provide the actual questions.
The MSM reporting on polls, IMHO, does leave something to be desired, particularly a more comprehensive explanation of how particular polls have been conducted. Maybe then the polls would have more credibility than they do today.
Still, a poll is merely a momentary snapshot of specific people surveyed. What of all those who weren’t surveyed? What of people who aren’t available at home at din-din time? Perhaps these people may differ for some reason in voting intention from those who’re always at home at five o’clock pm, perhaps for lifestyle or work reasons or whatever… gotta consider the possibility there…
Kate, great post. During last election, folks were wondering why the Libs were avoiding an election like the plague even though their “numbers” where going up. And now Harper seems pretty intent on pulling the trigger, notwithstanding the polls. The reason: “internal” polling is far more accurate as more money goes into gauging “undecideds” and polling only likely voters.
The parties aren’t stupid. Martin’s avoiding an election (still) and Harper’s pushing for one (still) for a reason.
Perhaps then the polling firms should endeavor to learn the techniques used for “internal polling”? Makes sense, doesn’t it, if internal polling tends to be more accurate…
The other factor that is often ignored is the percentage of people polled in Quebec. There is where the figures become skewed. If the Bloc is polling 10%, then that tells you something about the proportion of the sample taken from that province – perhaps 20% of the total. If the Liberals received 8% of their “vote” in Quebec, where the Conservatives are weak, then you have to set aside those numbers to see how the two parties compare, sans “undecided”, sans “Quebec”, to get a more accurate idea of where the seats might actually rattle out.
Kate:
The percentage in Quebec tells you nothing. The polls are weighted to account for sample distributions.
CS: polls are not a snapshot of who answers the phone only. The point of random samples and weighting is that you can make inference to a whole population – given some measurable uncertainty – without having to talk to more than 1000 or so people. This is why polls are so wonderful.
I’d suggest the reason why Harper may be itching for an election could also be because he knows things are unlikely to get any better for him.
Are we seeing the Flora Syndrome at work here?
Translation: ..named after Flora McDonald, who ran for a party leadership.
Her numbers were great, ’till the voting began. Her support dropped faster than Maggie T’s drawers.
“You can make inference” to a whole lot of bullshit Peter, given to some measure of certainty, and don’t let the door slam ya on the way out.
Good point, Kelly.
Continuing the thread, I received some quite thought-provoking questions from an anonymous commentor on my site about how polls are actually conducted, aside from the arcane numerical techniques used to crunch the final numbers:
http://thecanadiansentinel.blogspot.com/2005/09/questions-for-pollsters-if-they-care.html
Any pollster-type people are invited to respond.
“Kerry Wins!!!”
Brought to you by a lefty pollster, like Peter.
Rich: like I said, measurable uncertainty. And, anyway, I am neither a lefty pollster, nor did I make any predictions about election outcomes. Andm incidentally, every credible poll-based forecaster in the United States predicted a Bush win. You can see it in the January 2005 issue of PS, I believe.
Stephen: I’ll take a crack at answering your questions sooner than later. The person answering them has a wildly off idea of how polling works.
House to Fall This Fall?
Douglas Fisher thinks not, but he’s relying on the Liberal spin machine
…