I can't say I'm a big fan of the Léger Marketing group, a company that in my opinion, skates awfully close to the offside line in producing opinion polls that are tailored to provide results that clients demand.
Sneaky? Dishonest?
How about the way the company cleverly hides the French accent aigu (é) in their logo, preserving its Quebec pedigree here, while obscuring the fact that it is a Franco-Quebec company to clients in English North America.
There must be some research out there that holds that French diacritical marks on names are a turn-off to Anglophones, even the mighty
At any rate, Léger struck out rather badly in the last Quebec election, woefully underestimating Liberal party support, highlighting the fact that the four-percent margin error that they reserve for themselves, is a lot more critical than they intimate.
"With two days until Quebec's provincial election, the separatist Parti Quebecois remained poised to win a majority government, according to an exclusive Leger Marketing poll conducted for QMI Agency. The last QMI Agency poll before Quebecers vote on Sept. 4 placed the PQ in the lead at 33%, with the CAQ in second at 28% and the Liberals at 27%.We all know that the results of the election were much closer than Léger predicted and that the six percentage point difference between support for the PQ and the Liberals as predicted, was a far cry from the less than one percent (PQ-31.95%.....Lib -31.20%) which was borne out on election day.
The percentages for all three major parties remained unchanged since the last Leger poll, which was published on Aug. 24." Link
This incidentally was outside the margin of error, although I'm sure Léger would argue that it wasn't because the Liberal party strength was under-reported by about 4% and that the PQ strength was over-reported by 2%, both within the margin of error, when considered individually.
Adding the two numbers together the difference was indeed about 5%-6%, outside the margin, but I guess it's a case on interpretation.
You say ta-may-toe, I say ta-mah-to
Interestingly, when I spoke to Liberal party insiders, they told me that one of the reasons that Jean Charest called the last provincial election was because their own internal party polling numbers showed the Liberals and the PQ neck and neck.
Sadly, I'm pretty sure that the Léger poll actually affected the razor-thin outcome of the election, as some voters, considering the poll in the closing days of the campaign were motivated or demotivated to go out and vote.
There is also a minority of fence-sitters who actually are swayed to vote for the predicted winner.
Had Léger not published the inaccurate polling projections, we might very well have elected a minority Liberal government!
To be fair, other polling companies have done as abysmally poor of late, wrongly predicting the outcomes in both the recent Alberta and British Columbia provincial elections, in both cases botching the predictions so badly that they actually projected the wrong party as winners and this, not by a tiny margin. Read a story about these polling failures
All these polling disasters confirm the fact that polling has actually become less and less accurate over the years, something surprising in a world where technological breakthroughs has led to better and more accurate results in just about every field of human endeavor.
There are many reasons for this decline in polling accuracy. The cell phone age makes it harder to contact a targeted sample, where landlines of the past, pinpointed exactly where people lived. People today have also become less and less predictable, often changing their minds often before an election. Also, there is a dramatic drop in the number of people willing to participate in opinion polls, with privacy a very real and modern concern.
And so today, interpreting the shrinking numbers is as important as the numbers themselves.
Given the significantly higher numbers of undecided and non-responders, pollsters can no longer just ignore them and have to 'weigh' their impact, determining if they share a statistically significant disposition, substantially different form those who responded to the poll.
It isn't an easy task and we've seen pollsters in all three provincial elections badly misinterpret what information they had, where the undecided and unresponsive skewed widely towards the party that held power before the election.
Of all types of opinion polls, political polls should be the most accurate, after all, those being questioned are simply asked whether they will vote for party or candidate A, B or C.
Nothing complicated there, or so it seems and yet pollsters are getting it wrong at an alarming and expanding rate.
All this takes me to question the results of a poll that asks much more complicated questions, public perceptions on policy, where the nature of the questions may well determine or 'skew' the outcome itself, which is exactly what happened in the poll commissioned by the PQ and conducted by Léger, to 'determine' Quebecers' position in relation to religious accommodation.
This poll was commissioned by a PQ government with an agenda, wherein the poll is meant to validate and bolster support of their particular point of view over accommodations.
These types of polls are never be published if the conclusions aren't supportive and quite frankly, therein lies the rub.
Pollsters are expected to deliver supporting results and those that do are rewarded with repeat business, those that don't, are cast aside.
More often than not, the actual polling questions are dishonestly skewed to favour an expected result, it is the dirty secret of the industry.
Let us take for example this question;
"Do you agree that immigrants have a negative influence on our economy"
..and lets us compare it to this question;
"Do you agree that immigrants have a positive influence on our economy"
It is basically the same question, asked from the opposite point of view, but in a perfect world, we could expect that if 60% of respondents agreed with the first question , then only 40% of respondents should agree with the second.
But it just doesn't work that way at all, the nature of the question may very well result in 60% of the sample agreeing to both questions!
To be fair, the pollster should actually have asked this question;
"Do you believe that immigrants have a positive or negative influence on our economy"
In polling, the devil is in the question, and asking the right question is the crux of fairness, while asking a loaded question, an exercise in spin.
By the way, I'm not sure that even the third question above is indeed fair as well, because the question implies that immigrants are different.
We can reference Werner Heisenberg's Uncertainty Principle which tells us that the very act of observing, changes the result.
In fact a polling question that reflects bias, can never yield meaningful results.
- Do you agree that Jewish people are by nature controlling?
- Do you agree that devout Christians are intolerant?
- Do you agree that Pro-Life activists are dangerous?
- Do you agree that Native Canadians are lazy?
- Do you agree that Canadian Muslims pose a threat?
Can these loaded questions ever yield statistically valid data, or does the very nature of the question skew towards statistical manipulation?
And so we come to the poll commissioned by the Quebec government in respect to the debate over accommodations.
I must say, gentle readers, that I am outraged and furious at the crass manipulation and dishonesty of the entire Léger/Quebec government poll.
The very first question and the keynote of the entire poll is this gem.
Do you agree that:
"Putting an end to unreasonable accommodations favours social cohesion and integration"
Are you kidding me?
I haven't seen a more loaded and dishonest query, since the last referendum question.
If I was to answer the question, I would have to respond affirmatively because I don't believe in unreasonable accommodations, in fact who actually does?
Let me rephrase the question a little differently;
"Refusing to give into unreasonable demands by your children builds familial cohesion."
Who would disagree with that statement?
We all define 'unreasonable accommodations" differently, what is reasonable or unreasonable to one person, may or may not be to another.
But however we define 'unreasonable', we certainly would not be in favour of it!
I'm surprised anybody at all answered the question with a NO!
The rest of the poll builds on the same negative theme and cascades down to the point of ludicrousness where by the end, 60% of those queried, agreed that private schools should be included in provisions of a potential law that limits religious accommodations.
Think of the implications...
It would mean that private Catholic and Jewish schools would not be allowed to teach religion and that a Crucifix or Star of David could not be displayed nor worn by teachers! In a Muslim school, teachers would be barred from wearing a hijab and teaching the Koran!
And by the way, according to the poll, 62% of Francophones and 28% of anglophones believe that doctors should not be allowed to treat patients while wearing a kippa (skullcap) or hijab.
I could only imagine the international scorn such an interdiction would raise. Bizarrely, it would put Quebec in the same class as Iran and Saudi Arabia when it comes to religious tolerance!
How have we arrived at this intolerant juncture.
Well, the PQ has harped on the subject so long and so loudly, that Quebecers are actually now frightened by immigrants, egged on by a public debate over what good Quebecers should or should not tolerate.
Polls like this help fan the flames of intolerance.
Public discussions by politicians about how they are going to come down on Muslims, Jews, Sikhs and other minorities adds fuel to the already toxic level of xenophobia.
It's sad and disgusting.
The entire poll is crafted to yield certain results that the government wants in order to push ahead.
By the way, you can view the entire poll HERE, but it is in French.
......One last comment on the poll.
There appears to be a serious statistical error.
When the responses of 'Anglophones' and 'Allophones' are added to yield a combined percentage called 'Non-Francophones,' the results appear to be erroneous.
Simply put, Léger pollsters forget to put more statistical weight on the more numerous Anglophone responses.
As indicated in the poll, 58% of Anglophones and 70% of Allophones were in agreement with the statement. Combining the two yields, according to the Léger, yields 65%, which is statistically incorrect.
Consider two groups of people.
Group A consists of ten fat people averaging 200 pounds each.
Group B consists of six thin people averaging 150 pounds each.
In order to find the average weight for everybody, you'd have to add up all the pounds and divide by the total number of people.
Group A (10 x 200lbs,=2,000lbs.) + Group B (6x 150lbs=900lbs) = 2,900lbs divided by 16 people=181 lbs. average.
What Léger appears to have done is to just take the average between the 200lbs average of Group A and the 150lbs average of group B, yielding a 175 lbs., an error a first semester statistician would never make!
For those mathematically inclined here's a deeper explanation.
All the following numbers are extrapolated directly from the diagram.
Of the 324 Anglophones who were queried, 12% declined to answer, leaving 285 who did respond, of which 58% or 165 were in agreement with the statement.
Of the 179 Allophones that were queried, 6% declined to answer, leaving 165 who did respond, of which 70% or 116 were in agreement with the statement.
Between the two groups, 450 people responded, of whom 281 agreed with the statement.
That yields 62% NOT the 65% indicated in the poll.
With the error corrected, the grand total also changes from 70% to 68.2% By the way, this error is reproduced in every single question.
Readers, I promise you this.
Somebody at Léger is going to read this post and if I am wrong, I will hear about it immediately.
If I am wrong I will apologize.
If I am right, there will be a deathly silence.