Western Standard

The Shotgun Blog

« The global-warming emperor has no clothes | Main | Pakistan on the edge? »

Thursday, May 17, 2007

The problem with election polling in Canada (Manitoba edition)

Manitoba will have a provincial election on Tuesday.  Probe Research took a poll and predicted, based on the results as they reported it, that the NDP will be re-elected (CBC):

Forty-four per cent of the decided and leaning voters polled said they would support the New Democrats, while the Progressive Conservatives had the support of 37 per cent.

"It doesn't matter whether it's the Conservatives or the NDP, every time one of these parties gets into the 40s, they are rewarded with a majority government," Probe Research spokesman Scott McKay said Wednesday.

Pretty cut and dried, right?  Wrong.

Like nearly every pollster in Canada, Probe's poll "numbers" do not include voters who declare as undecided or simply refuse to declare altogether.  Airbrushing these voters out of existence (or "redistriubting" them, which is essentially the same thing) is routine among Canadian pollsters; I'm assuming this is in order to get numbers that can be used to project seats.  The problem should be obvious: no one can assume the UNDs and refuseniks will break down the same way "decided" voters do.

In some elections, where undecideds (UNDs) and refuseniks are low in number, this doesn't matter much.  In this case, however, the numbers are pretty substantial:

Thirteen per cent of those polled said they were undecided, and 14 per cent would not reveal their voting intentions.

That's more than one in four voters overall, which makes a difference between the reported numbers (44-NDP, 37-PC, 16-Lib, 3-Others) and the "raw" numbers (which I calculated to be 32-NDP, 27-PC, 12-Lib, 2-Others, and 27 percent undecided or refuse to answer).  More importantly, the critical question becomes not what the PC's can do to turn NDP voters at this late date, but rather what the UNDs and refuseniks will do on election day. 

Of course, we might know more about the UNDs if the survey would examine where they are (Winnipeg or the rest of the province) and who they are - as American pollsters do - rather than ignore them.  That said, we do have some semi-established rules of thumb on these voters down here.

Given the conventional wisdom that the NDP will cruise, odds are most of the 14% who refuse to answer support the PCs (or, for a few, the Libs), but don't want to tell the pollster.  As for undecideds, in legislative elections, they tend to break for the opposition - as high as 3 to 1 in some cases.  If these trends translate across the 49th, Manitobans and election watchers will be in for quite the surpise on Tuesday.

It wouldn't be the first time, either.  I've been following Canadian politics for about ten years, and I've seen numerous examples of pollsters caught flat-footed in provincial elections due to this strange (in my opinion) method of number reporting, such as:

Nova Scotia (1999) - An election that began with a majority of voters undecided saw polls bounce like Pacific seismographs, and they still missed a late Tory surge that turned a nip-and-tuck vote into a PC majority.

Saskatchewan (1999) - Roy Romanow had a 20-point lead in pre-election polls.  The Sask Party tried to counter the impression that the race was over by releasing its internal poll which showed a six-point gap.  The press and pundits howled with derision.  The Sask Party ended up winning more votes than the NDP and forcing Romanow into a minority (and a quick retirement).

Manitoba (2003) - Polls showed a crushing re-election for the NDP; while the NDP did win, the Tory strength stunned everybody and temporarily saved Stu Murray as PC leader.

Alberta (2004) - A Ralph Klein coronation turns into major heartburn when the Tories fall below 50% in the popular vote, a development that stunned the Alberta PCs and began the countdown to Klein's exit.

Quebec (2007) - Not a single poll put Mario Dumont's Action Democratic du Quebec in second place, but the voters did.

I would also note that not a single one of these surprises benefitted the incumbent party (The Nova Scotia Liberals were turfed from power in 1999, and of course, Jean Charest's Liberals fell to minority status in Quebec this year).

Back when the writ was dropped, I predicted a PC victory (in a now-largely-forgotten comment on another Canadian blog).  I stand by that prediction.  If anything, this poll, which supposedly paves the way for the Manitoba NDP, makes me even more confident that I am right.

Posted by D.J. McGuire on May 17, 2007 in Canadian Politics, Canadian Provincial Politics, Media | Permalink

TrackBack

TrackBack URL for this entry:
https://www.typepad.com/services/trackback/6a00d834515b5d69e200d8352e1ec553ef

Listed below are links to weblogs that reference The problem with election polling in Canada (Manitoba edition):

Comments

I think that I've said this before: unless a public opinion poll is for a single riding only, then it means nothing in a first-past-the-post electoral system. I'll just ignore these polls and vote for whomever I want.

Posted by: Cory Schreyer | 2007-05-17 11:39:58 AM


Polls are a stock-in-trade for the MSM, an easy story on a slow day.

That and speculating that an election is just around the corner.

Always takin' the easy way, huh guys?

Posted by: Set you free | 2007-05-17 11:53:05 AM


The media is lazy. Have you nseen this widely feported anywhere?

"Trudeau questions capitalism"


Justin Trudeau called on hundreds of high school students in Windsor Wednesday to rethink the capitalist system and reconsider Canada's reputation as a model nation.

http://www.canada.com/edmontonjournal/news/story.html?id=296460e7-f6ba-4e93-acff-81b053d20ab1

Posted by: obc | 2007-05-17 12:25:44 PM


In Canada most "pollsters" are motivated. In the
Atlantic Region all "polls" by Corporate Research Associates are politically motivated. Check out the Manitoba based Pollsters who appear to have undertaken a lot of Government (NDP) Projects and
view their current findings with a huge grain or grains of salt -Pollsters can and have been bought
particularly by the well honed but virtually defunct Liberal Party of Canada. Lot of Pollsters appear regularly with Newman CBC and Puffy Duffy Bell Media, for money. I am shocked and appalled. MacLeod

Posted by: Jack Macleod | 2007-05-17 2:17:18 PM


"I am shocked and appalled. MacLeod"

No your not.

Posted by: obc | 2007-05-17 2:24:10 PM


obc:

Justin Trudeau said that without reprimand from the MSM, but just let a conservative make such a statement and it's front page news.

Duffy and Newman have their favs when it comes to pollsters, it's anyone who has the Liberals in the lead.

This is one godforsaken country if the people of want to reward corruption and a totally disorganized and divided pack of people who are now making up the LPC.

They're broke, which tells us maybe the little people are not donating to their party for starters.

Election speculation is always out there with the boneheads in the media. They were the main ones talking about it. Harper kept saying he had no intention of arranging one but they kept on yapping. Now the all-knowing asswipes surmise with their polls showing the Conservatives falling, it's off the radar.
It is to weep.

Posted by: LizJ | 2007-05-17 2:36:15 PM


"Justin Trudeau said that without reprimand from the MSM"

Not just without reprimand - I sent it to CTV web site news - no response and no mention at all of his speech. Probably because they agree with him, and realize that publicizing it would be embarrassing to l'il Justin among average Canadians.

Posted by: obc | 2007-05-17 2:42:23 PM


A poll of Quebec Decided Voters that had the ADQ in second place: http://www.angusreidstrategies.com/index.cfm?fuseaction=news&newsid=35

Posted by: MCG | 2007-05-17 4:18:18 PM


Second place does not count - in politics winning is everything - Charest won the election and is unquestionably Premier of Quebec - All Canadian Pollsters are professional manipulators whose contribution to the politcal process in Canada
is disgusting - Pollsters should be forced to list their respondents by Elections Canada -because I think many of them are non existant. MacLeod

Posted by: Jack MacLeod | 2007-05-18 7:00:18 AM


I couldn't agree more with your analysis. If the Liberals get more than 5 seats in Manitoba, it'd be a miracle. The Tories and Dippers will be neck and neck come Tuesday evening - and I just hope to God that Gary Doer gets tossed!

Posted by: Natedawg | 2007-05-18 9:22:49 AM


We had years of Liebrals rethinking the capitalist system.

They never called it that. But I can tell you, I know many families who struggled like never before during those years--while their incomes were frozen, and their work-loads tripled.

If a conservative said that we should rethink the capitalist system, it would be to make sound improvements. In fact, that is what an honest forum for disccussion around the politics of Canada, is all about. And that IS what having a democratic government is all about.

And yes, that IS not what trudeau had in mind. He would like everyone to rethink the capitalist system, and push us back into the dark ages, where hungry people could eat grass, slugs and worms as their back-up food.

Perhaps what he said did not make news, because he is not newsworthy? Just because he is his father's boy does not mean he is going to be or is effective. He may have half the genes, but that is where the buck stops.

Posted by: Lady | 2007-05-18 11:53:20 AM


"Just because he is his father's boy does not mean he is going to be or is effective."

True, but he IS the Liebeeral candidate in a Montreal riding.

Posted by: obc | 2007-05-18 12:10:51 PM


His name may be Trudeau, but if you've listened to him speak & see how he thinks, he is really a Sinclair - his mother's son. No depth to him at all.

Posted by: obc | 2007-05-18 12:26:21 PM


Well call me undecided. I'm a right winger and there is no conservative party to vote for. The PC's here are as right wing as Joe Clark.

Maybe I'll go green, Maybe I'll decline my ballot......

Maybe stay at home and blog,

I miss the Manitoba party.

Posted by: DrWright | 2007-05-19 5:38:14 AM


. . . or maybe you'll help with Alberta separation so that we can have an actual conservative nation here in North America.

Posted by: obc | 2007-05-19 5:44:44 AM


Here's a copy of an email I just sent CTV after viewing their web site:

"NDP headed for third straight majority in Manitoba"


Is that so? You've already counted the votes, have you? What was the final tabulation?

Even the word "MAY" is omitted. The headline says it's a done deal - before one vote has been cast.

How about this headline for your web page?

CTV TRIES TO INFLUENCE MANITOBA VOTE


DRIVE-BY MEDIA! Sheeeeesh!

Posted by: obc | 2007-05-21 7:10:19 AM


DJ,
Thanks for the trenchant critique of the shortcomings of Canadian polling.

BTW, I see the results are in from Manitoba.
NDP 48% (Probe poll said 44%)
PC 38% (Probe poll said 37%)
LP 12% (Probe poll said 16%)

Sooooo, care to add to your comments? I think the following could probably use a bit of editing:

"Given the conventional wisdom that the NDP will cruise, odds are most of the 14% who refuse to answer support the PCs (or, for a few, the Libs), but don't want to tell the pollster. As for undecideds, in legislative elections, they tend to break for the opposition - as high as 3 to 1 in some cases. If these trends translate across the 49th, Manitobans and election watchers will be in for quite the surpise on Tuesday."

On second thought, maybe this paragraph needs more work:

Back when the writ was dropped, I predicted a PC victory (in a now-largely-forgotten comment on another Canadian blog). I stand by that prediction. If anything, this poll, which supposedly paves the way for the Manitoba NDP, makes me even more confident that I am right.

Posted by: truewest | 2007-05-22 10:34:01 PM



The comments to this entry are closed.