Sunday, June 28, 2009

The unholy Ipsos Reid + CanWest alliance; junk-polls and media bias shaping Canadian opinion.

Paul from Paulitics has revealed that Ipsos Reid has been releasing numbers with a pro-Conservative bias. Well, I hate to say it, but bias stemming from illegitimate polling practises and media bias is everywhere. Take one of the latest Ipsos Reid polls, just as an example. Respondents were asked the following question:

(a) Stephen Harper and the Conservatives are doing a good job of managing the issues that are most important to Canadians (b) and should continue to govern.

1. Junk Poll: Misleading question (read it here).

The above is a classic misleading question (actually, it's not even a question...). The technical term for a such a misleading polling question is double-barreled question, one of the "Top 10 Ways to Get Misleading Poll Results"

How it works

John Wright - Ipsos Reid CanadaJohn Wright and his conservative buddies (a.k.a Ipsos Reid Canada) know quite well that Canadians “don't want an election” (68% of respondents don't want an election, June 2-4 Ipsos Reid poll). It appears that Ipsos Reid Canada used this information to craft a double-barrelled question in order to confuse respondents and get a more positive result on an on itself perfectly legitimate question (that should have been asked in the first place):

Do you agree that Stephen Harper and the Conservatives are doing a good job of managing the issues that are most important to Canadians?

But by adding the b part ("and should continue to govern" = avert an election) Ipsos Reid seems to have been aiming at those 68% of respondents that don't want an election, hoping that at least some of them would respond more positive to the question as a whole

“I don't want an election (b) so therefore I still “somewhat agree” with the question as a whole (a+b).”

And by putting this part into the latter half of the question, Ipsos Reid is giving the b part some extra weight too. Bruce Grundy explains:

If you ask a double barrelled question [..] you will almost certainly not get answers to all the matters raised. Usually, only the last point will be dealt with.
- Bruce Grundy, "So you want to be a journalist?".

Perfect!

2. Junk-results: Misleading, again, and again.

The results of the misleading poll question were as follows:

Strongly agrees (20%)
Somewhat agrees (34%)

I don't have access to the full results of the poll (update below) but given the likeliness that this poll was in the 'Strongly Agree', 'Somewhat Agree/Somewhat Disagree', 'Strongly Disagree' format, it's probably safe to assume that the rest of the respondents (46%) either strongly disagreed with the statement or has no opinion (or were offended by the biased question and therefore not answering).

Let's assume the following:

31% strongly disagree
15% no opinion/offended

Logic also teaches us that if a person somewhat agrees (34%), they also somewhat disagree, right?

Then we can conclude that also the presentation of the results is misleading. From the results above, it would be just as well to conclude that an overwhelming majority of Canadians agree (strongly 31%, somewhat 34% - totalling 65% !!) that Stephen Harper is doing a bad job, right?

3. Junk-Journalism: Misleading, again, and again, and again.

While Ipsos Reid is happy to supply misleading results from a flawed polling practise, CanWest uses these faulty results to push an even more positive image onto Conservative Party.

Here's a good example how CanWest misrepresents the already flawed poll:

(1) Harper has the edge, however, with half (54%) of those polled saying he should continue to govern. (2) Five in 10 (54%) Canadians agreed Harper is doing a good job of managing the country while four in 10 Canadians agreed Ignatieff would do better. - italic inserts mine

Indeed, both (1) and (2) are first order fallacies because one cannot make these assumptions based on a single double-barrelled polling question.

Fallacy explained
Let me explain this with a simple example:

Jan (15) was asked to respond to the following statement:

(a) I dislike my current math teacher and (b) she can drop dead.

Yes, you've got it right, this is also a double-barrelled statement!

Let's assume that Jan doesn't like her math teacher. Let's also assume that Jan is not the hateful type either (even towards her math teacher that she really dislikes); then what is she supposed to answer?

Obviously she's NOT going to choose "strongly agree" because she doesn't dislike people in that (b) way . She's also NOT going to choose "strongly disagree" because she really dislikes her teacher. It's therefore quite likely she will choose "Somewhat agree" given the fact she really dislikes the teacher (a).

The example above shows that, although Jan does not have a deathwish for her math teacher, she might as well have chosen "somewhat agree". So how could someone take the "somewhat agree" as evidence that she has a deathwish for her teacher?

What we are dealing with here is fallacy of distribution. Jan's "somewhat agree" to the full answer can not be translated to a "somewhat agree" to the single parts (in this case the death-wish) of the statement, as shown above.

Back to our Ipsos Reid/CanWest poll and "analysis"
All that we can know for sure is that only 20 percent agreed to the full a+b question: we can only speculate on the other 80% due to the confusing and misleading double barrelled question.

It's very likely there are many people like Jan among the respondents of the Ipsos Reid poll who fully agreed with one part of the question (b part, to avert an election) , but not so much with the other part (a part, Harper and Cons do a "good job").

4. Conclusion
The unholy alliance of Ispos Reid and CanWest is good for some good laughs, but their polling (Ipsos Reid) and journalism (CanWest) practises presented in my examples are clearly and severely flawed. The confusing polling question (Ipsos Reid) in combination with the flawed analysis (CanWest) appear to have been designed to inflate support for the Conservative Party and deflate criticism of the current government: hard to fully proof, but I'll let you decide.

5. Recommendation
You think this example is an exception to the rule? Think again. I see this kind of junk polling in combination with junk-journalism EVERYWHERE in the Canadian press.

The best thing to do is to ignore all polls and stop reading the corporate media/press: it's junk. I've given up cable TV and newspaper subscription a long time ago, and I advice you to do the same. Please support the CBC for a more objective and unbiased presentation of Canadian news.

UPDATE

I noticed this post which is helping me out with the missing date:
A majority (53%) of Canadians ‘agree’ (20% strongly/34% somewhat) that ‘Stephen Harper and the Conservatives are doing a good job of managing the issues that are most important to Canadians and should continue to govern’. On the other hand, four in ten (43%) ‘disagree’ (23% strongly/21% somewhat) that this is the case.
It now looks like the poll was in the following format, and the numbers read as follows:

'Strongly Agree' (20%), 'Somewhat Agree (34%), Somewhat Disagree' (21%), 'Strongly Disagree' (23%)

Even if we accept these latest poll numbers as being legit results, then one can only conclude (and even this can be questioned) that there are more people disagreeing (23%) with both parts of the statement than agreeing (20%) ! All others respondents appear to be highly confused how to answer the question.

Furthermore, if Ipsos-Reid/CanWest is allowed to misrepresent the facts, then Getting it Right (GIR) feels fully entitled to do the same. The actual results with an explanation by GIR:
Strongly Agree (20%) - Okay, these people seem to support the Conservatives regardless of the stupid question.

Somewhat Agree (34%) - Logic will explain that people who somewhat agree (opposed to those who "strongly agree") also "somewhat disagree" with the statement, right? Right!

Somewhat Disagree (21%) - These bunch are as confused as the ones above, but they surely "somewhat disagree" too.

Strongly Disagree (23%) - The ones that clearly don't support the Cons and really don't mind election either.

6. Do the math and make your own headline:
Math:
20% agree, 78% disagree (23% strongly disagree, 52% somewhat disagree)

Headline:
Large majority (78%) of Canadians disapprove of Stephen Harper and his cronies and want elections now!
I feel great being so creative with flawed polling numbers!

Footnote:
Q: What's more flawed than an Ipsos Reid Poll?
A: Tons of them!

LINKS
- What is John Wright reading when he's not blogging, or, double-barrelled polling questions are misleading and should be avoided
- Ipsos junk poll using double-barrelled polling question - so much for credible polling
- Be warned if you dare to critisize us, because John Wright (vice president Ipsos) is ready to sue you too.
- National Post + Ipsos alliance: Canadians don't want an election right now
- Paulitics: "I'm John Wright from Ipsos Reid", or "why the recession has given me more time to piss off other bloggers"
- We at CanWest love the polls from Ipsos. They are so well crafted. We especially love results from double-barrelled question since those numbers can be used so easily to mislead our reader
- Wikipedia: Fallacy of distribution

FULL CANWEST ARTICLE

OTTAWA — Conservative and Liberal MPs united Friday to pass a government confidence vote — averting a tight summer election, which a new poll shows very few Canadians wanted.

Conservative and Liberal MPs voted 214-82 against the New Democratic Party and Bloc Quebecois to ensure passage of government budgetary spending estimates, and to set the stage for a Liberal confidence motion during the week of Sept. 28 that could trigger a November election.

An Ipsos Reid poll for Canwest News Service and Global National found 60 per cent of Canadians do not want a summer election, though half agree Parliament is not working. It found a statistical tie of 35 per cent for the Liberals and 34 per cent for the Conservatives, but a large gap in favour of Conservative Prime Minister Stephen Harper over Liberal Leader Michael Ignatieff in who can best manage the issues facing the country.

After a week of high political drama, ending with NDP and Bloc MPs accusing the Liberals of propping up Harper's government and abandoning the unemployed, MPs headed back to their constituencies for a three-month recess and preparations for a potential fall campaign.

The poll of 1,000 adults found a tie in vote-rich Ontario, too, with the Liberals at 40 per cent and the Conservatives at 38 per cent.

Harper has the edge, however, with half of those polled saying he should continue to govern. Five in 10 Canadians agreed Harper is doing a good job of managing the country, while four in 10 Canadians agreed Ignatieff would do better.

The survey was taken June 16-18 as Harper negotiated an agreement with Ignatieff to allay a Liberal threat to defeat the government Friday if he did not respond to concerns about unemployment, stimulus spending, the deficit and the isotopes shortage. The poll is accurate to within 3.1 percentage points, 19 times out of 20.

The poll found 53 per cent of Canadians agree Parliament is not working.

At the same time, Quebec stands as the only province where a slight majority of 52 per cent wants an election to clear the air.

In Quebec, the Bloc maintains a lead of 38 per cent, followed by 33 per cent for the Liberals, 16 per cent for the Conservatives, seven per cent for the New Democratic Party and four per cent for the Green Party.

The Harper-Ignatieff deal established a working group of Conservative and Liberal MPs and officials to study two potential employment-insurance reforms. One would allow self-employed Canadians to participate in the program. The other would establish a uniform eligibility rule to collect benefits, replacing a patchwork of requirements that are linked to regional unemployment rates.

During debate before the votes, NDP and Bloc Quebecois MPs blasted the Liberals for withdrawing their support for a long-standing NDP-Bloc call to lower the minimum number of hours required for EI benefits to 360 from 420. Ignatieff pulled that from the table on the eve of talks with Harper.

"I know we're going to meet people who've now run out of their EI insurance and are going to fall off into extreme poverty this summer, and the unemployment rates seem to be still rising, so it's going to be a tough summer," NDP Leader Jack Layton told reporters.



5 comments:

susansmith said...

well I don't know about the truthiness of this statement, "start supporting the CBC for a more objective and unbiased presentation of Canadian news."
Being a social democratic and all, Mansbridge had a distain for Layton which he just couldn't hide. Some folks often preceive the CBC as "liberal" friendly and not so friendly with the NDP. I much prefer independent media, in MVP.

susansmith said...

beyond that, it's a great post, which I forgot to add.

Erik said...

I fully get your poin, Jan, but the CBC is the best we've got.

Thanks for the comments - good to hear from you!

susansmith said...

thanks Eric.

David said...

Fantastic analysis of the garbage and drivel out there! I always tend to question statistics and polls, and look at them the other way around all the time. I tend to find simple yet logical 'deconstructions' like yours blow apart the supposed logic in many of these biased polls. Keep up the good work!

Post a Comment