Inkless Wells

Paul Wells on all the latest out of Ottawa—along with the occasional post about jazz. Follow Paul on Twitter: @InklessPW
He also offers his thoughtful perspective of Stephen Harper’s last 10 years in his recent eBook, The Harper Decade.

Categories: Blog Central, Capital Read, Inkless Wells

New poll! New confusion!! Every sentence in this headline has an added exclamation point!!! Really!!!!

by Paul Wells on Friday, September 12, 2008 11:36am - 62 Comments

CBC Newsworld is on air right now with a new Harris/Decima poll: Conservatives 41%, Liberals 26%, NDP 14%, Greens 9%. Link when it becomes available.

This poll rather profoundly contradicts yesterday’s Nanos numbers. Commenters should feel free to indulge the most convoluted conspiracy theories to explain the divergent results, because offhand I can’t think of any rational explanation.

Tags: Election Watch, Harris-Decima, polls are for dogs and people who always use the "polls, that's majority territory or Fantasyland but I'm not su

More by Paul Wells

Chris B

SF – this is a four day rolling poll, sampling 300 people on the 8th/9th/10th/11th. So over half of the people were polled before the “gaffes” (can we get an alternate word to use – gaffe is becoming so annoying)

The story is up on CBC’s website now
Paul Wells

sf: Only my closest friends get to call me Kady. I’m afraid you don’t know me well enough.
JB

Do the polls account for lying in their methodology?
I’m willing to bet more and more people are starting give false replies.
Anon

I’m watching Dion on TV now. There are really two Dions. The speechmaker, who is an absolute disaster. And the Q&A guy who could beat pretty much anybody in a debate. It’s like two different world – don’t know how it will play.
sf

Ha, my apologies Paul Wells and Kady O’M.
I was clicking around and I thought I was on Kady’s blog.
Pingback: A brief ITQ interlude: “Now for my favorite part of the show …” : Capital Read : Inside the Queensway : Macleans.ca Blog Central
Geiseric the Lame

comment by Paul Wells on Friday, September 12, 2008 at 12:15 pm:

“Canadians know there’s such a thing as a Conservative Turning the Corner plan? Really?”

You’d have to give me odds on that one.
Kenneth

What happened? Easy, it was a different 1000 people.
http://deleted Sandi

Dion in a group situation is good….then when on stage/press conference….has more trouble. I think it’s because the press are more interested in how he sounds then what he says so they can all write about his English the next…I wonder if he’s so self-conscious because of the constant media criticism day in, day out.

….just wondering

You know, because that’s the most important thing in this election – Dion’s english and boy oh boy we really do need to read about it every day.
madeyoulook

Wow, tell me more: what corner gets turned? Shouldn’t this be in the, you know, news or something?

PW wants an explanation from among the irregulars for the “contradiction” between polls showing Tories ahead vs. waay ahead. Ever your humble servant, sir, I offer the following:

1. Not so much a contradiction as a question of degree of Liberal implosion (carefully note I did most certainly not say Conservative well-deserved approval).

2. Pollsters just make stuff up.

3. Maybe it’s a little less than “nineteen times out of twenty” after all, eh you smarty-pantsed statistics nerd, wherever you are?

4. More sinister, perhaps: just like Harper allegedly spooked the country last time with premature talk of a majority, the CBC will do its utmost to spook Canadians this time. This only works as a theory because the CBC believes “Conservative majority” would be spooky indeed.
southernontarioan

Three words: Margin of Error

The margin of Error on each poll is roughly 3% 19/20 times.

That means that if the CPC is polling at 41 +- 3% then really they are anywhere between 38 and 44.

Nanos says that they are at 37 +-3 so that mean they are actually anywhere from 34 to 40.

So there is actually no contradiction between these two polls. If the CPC’s actual level of support was say.. 38.. both polls would be correct.
PolPundit

Check out this site before venturing any comments on polls.

Aggregating all the polls on a rolling basis is state-of-the-art analysis

http://www.trendlines.ca/electcanada.htm
Sean S.

JB wrote: “Do the polls account for lying in their methodology?
I’m willing to bet more and more people are starting give false replies.”

Not something that can be accounted for, and it’s unlikely it would be widespread enough to become an issue.

However, there are pitfalls aplenty in any such poll, including:

-applying national statistics to an election based upon local races. (for starters, it is of little use to know that the BQ polls at 9%nationally). But conducting polls of each riding would be too expensive.

-the variability of voter turnout: all the expressed support in the world won’t help a party if those supporters don’t cast ballots.

-faulty representation: it’s not just those without land lines who are missed by such polls. It’s hard to know just who is motivated/bored/polite enough to take the time to complete the telephone survey.

-longitudinal drift: which is a high-fallutin’ way of pointing out that “rolling surveys”, completed over several days, are prone to serious errors if news and/or events arise in the midst (puffin poop, etc…)

And those are just the broad problems with surveys that are conducted with professionalism and integrity. Poorly constructed surveys, mis-managed interviewers, fudged sampling criteria to meet deadlines, and a whole host of other issues can undermine things.
Laura

Sounds pretty much like the end of it …
http://bigcitylib.blogspot.com bigcitylib

Its because the whole ghastly disaster that’s the Tory campaign has yet to sink in yet.

Ohmmm! Ohmm!

Its the early days!

Ohmm! Ohmm!

You fools, if Harper gets a majority…there goes the metric system! Furthermore, downloading will be illegal! Creationism will be made mandatory in our classrooms!

Wells knows all this too! He wrote a book! Tell ‘em Paul!
David

(for starters, it is of little use to know that the BQ polls at 9%nationally).

Unless you can do math and multiply 9 by 4.
madeyoulook

No, David, you’re still nowhere with a four-times table. If a poll won’t tell us the party rankings “inside Quebec” (where one party limits its play) vs. “outside Quebec,” then a nationwide poll tells us very little.
Chris B

And rolling polls aren’t exactly stat of the art – the National Election Study has been using them for at least a decade.

I never commented that they are bad, merely that some of the bad news from the Conservative campaign wasn’t reflecte in this poll.
sf

What we really need is hourly polls. Daily is too slow.
Mike Duffy

Stay tuned to CTV tonight. We’ve got a doozie of a poll. This campaign is looking like 1984 all over again.
Anon

Geez Mike Duffy, are ratings that bad that you’ve to troll the Inkless well for viewership?

NANOS is out — Tories 38, Libs 31 basically flat from yesterday. Separately, upto 60% could change their votes.
John D

Simple, Canadians are becoming increasingly frightened that they will be the next target of Conservative attack ads, so they are telling pollsters they support Harper. No one wants to be on the receiving end of bad Flash.
http://arandomprocess.blogspot.com/ Andrew E

Mr. Duffy, are you saying that you’ve got footage of Stephane Dion patting Iona Campagnolo on the butt?
Wayne

Hey SF : you got my vote this political junkie needs an hourly fix – tolerance builds up after as many years as I have been hooked on the stuff. But I still get a kick out of how people like to adjust the numbers to fit into their pre-conceived political view rather than tracking the trends and trying to determine what is happening and then are always distressed and blame the polls or (old Liberal Habit – blame the intelligence of the voter)
Wayne

Oops I almost forgot but I see Andrew picked up the ball – the next campaign highlite will be Dion patting Ms May’s posterior.