Monday, August 24, 2009

Conservatives Fade

It looks like the Conservatives prolonged surge in the polls has ended:
OTTAWA _ A new poll suggests the Conservatives and Liberals remain locked in a dead heat amid rumblings of a possible fall election. The Canadian Press Harris-Decima survey put the parties in a statistical tie, with 32 per cent support for the Liberals and 31 per cent for the Tories. The NDP were at 16 per cent, the Greens at 11, and the Bloc Quebecois at nine. The numbers have barely budged throughout the summer, a period in which voters are typically disengaged. Some Liberals have been pushing for an autumn election while the Tories have been trashing the idea, saying it would hurt the economic recovery. The survey of just over 2,000 respondents was conducted Aug. 13-23 and is considered accurate to within 2.2 percentage points 19 times in 20.

No internals yet, will update later.

I must say, I absolutely LOVE the timing of this poll, on the heels of that stench released yesterday. The "numbers have barely budged", reaffirming what everyone else has concluded. The fact HD gives the Liberals a lead, even if it's statistically irrelevant, is simply delicious on the heels of the Ipsos TURD.


Poll internals:

Harris Decima replicates recent Ontario results, Libs hovering around 40%, ahead of the Cons, NDP in trouble:
The Liberals lead in Ontario by six points with the NDP struggling in popular support in the province. The Liberals hold 40%, followed by 34% for the Conservatives, 13% for the NDP and 10% for the Greens.

Quebec shows a weakening Lib vote, and a very poor score for the Cons:
The BQ remains in front in Quebec with 37%, followed by 28% for the Liberals, 12% for the
Conservatives, 11% for the NDP and 10% for the Greens.

Might be a tad low for the Cons, other findings have shown a slight uptick. There's no question the Lib vote has sagged in the province, but it's fluid.

HD also replicates a pretty firm trend in Atlantic Canada, Libs out front:
In Atlantic Canada, the Liberals continue to maintain an advantage. Here the Liberals lead with 38%, followed by the NDP with 32%, the Conservatives 23% and the Greens with 5%

Kind of a goofy result for British Columbia, even with the relative manageable margin of error (two week sample):
The Conservatives lead with 28%, followed by the NDP with 26%, the Greens 24% and the Liberals with 20%.

That Green number is an "eye popper". Very low scores for the Cons and Libs, relative to other recent polls.


Mark McLaughlin said...

Pools, polls, polls. Still as useless as they've ever been.

The problem with them is that people tend to compare polls done by different companies with each other. Each has their own own sampling and statistical methods so comparing them that directly is folly.

The only semi-useful way to look at them is to follow the trends from a single company. That way the methods are identical.

Our lazy MSM doesn't get it though and just pay a polling firm to give them a headline for every 7th paper.


Anonymous said...

Note the differences in sample size and MOE. The Decima poll uses a sample size twice as large as the Ipsos poll, and has a smaller MOE.

Also interesting that the write-up for the Ipsos poll in the National Post casually mentions that the MOE is "significantly larger for the regional numbers." That would explain the Ispos poll's hard-to-believe Tory numbers in Quebec. 20%??? Please...

Steve V said...

Actually, why it's useful to compare pollsters, is so you can see which one's are the most useful. We have a whole bunch of firms, and most of them have been within the margin of error, relative to each other.

Just to add, I always get a kick out these "polls are useless" comments. They're so useless I feel compelled to comment ;)

Anonymous said...

The H-D poll was over two weeks. I'd wait for the polls week of Sept 14 onwards. The no-confidence motion isn't likely to be tabled until the week of Sept 28.

With the parliament back, and QP back, and all the talking heads back, that's when you'll get the real pulse of the country.

Ipsos-Reid hasn't been the same since CTV/Globe fired them as their pollster.

Steve V said...


I think we're all waiting for Parliament to reconvene to get an accurate read.

Anonymous said...

Iggy is saved! Iggy is saved!

One Decima poll and the Grits are excited. Then the Ipsos one had them biting their nails. Even though it was different from the rest.

The Decima one tend to have a bigger sample so it would be a more representative one. I don't know what Darrell Bricker was doing.

BTW, is Allan Gregg doing the Decima polls these days?

Steve V said...

"Then the Ipsos one had them biting their nails."

Who was biting their nails? About the only thing I considered, would the media demonstrate a critical eye, or just run with crap to support this anti-Iggy kick we've seen the past few weeks.

Here's the deal, this poll is just more of the same, that we've seen all summer.

Anonymous said...

This survey was conducted for the most part a week before the Ipsos poll.

Anonymous said...


I would hold my horses until the EKOS weekly tracking is released to counter the Ipsos one. Mind you, it seems that this is the first Decima poll for a long time. It is one that the NDPers and the Greens usually use to spin their message.

Anti-Iggy kick? Chantal had just written a good column on the good things that Iggy has done in Quebec. Good response for a party that hasn't done any mobilizing for almost two decades.

Steve V said...


I can't believe you havent' sensed a anti-Iggy kick out there. You cherry pick one column, I'm looking at totality.

Who's holding horses? Same old same old, until I see something different. Ekos won't move much.

Anonymous said...

The AR poll already served it's partisan purpose . . .

As I am sure is evidenced in other canwest papers across Canada on a Monday morning, there is a page 1 note (with a sharp pic of Harper) followed by a headline across the entire top of the Canada section:

"Conservate Lead Drops Odds of Fall Election"

with subtitle:

"Liberals under Ignatieff having trouble gaining traction on key issues, poll results suggest"

Ah, the payback for being saved from bankruptcy is sweet like the blueberries in August . . . ;)

And, Steve, I had to disagree with you on another blog today. Ignatieff turned things around at 2:15 yesterday. 2:15 and 26 seconds to be exact. You really should pay more attention to political events that unfold on Sunday afternoons in August ;).

Anonymous said...

Opinion Polls: Getting the results you want

I hope you enjoy.

Steve V said...

I stand corrected Joseph.

DL said...

The national numbers from Harris Decima seem reasonable and they are within range of recent polls we have seen from Ekos, Nanos and Strategic Counsel. Ipsos is clearly a total outlier. As was the case during the federal election, HD seems to take the cake when it comes to some really bizarre regional swings and numbers - those BC numbers are almost as ridiculous as that SC poll a while back that had the Greens at 26% in Quebec!

Mark McLaughlin said...

anon makes a good point. The HD poll was done over two weeks, the IR done over 2-3 days I think.

Tough to compare the two. a longer time frame is going to smooth out the curve. If polls are supposed to be 'snapshots', a really long exposure would lead to a more washed out picture.

On another note. I hate the opinions that say one pollster is better than another. Everyone was touting Nanos in 2006 as the de-facto pollster of record because he was so close to the actual results. Then in 2008 his last report wasn't even close.

He tried some shifty ass covering by saying his last day of the 3 day sample was on the mark, but fails to mention that he used the exact same methodology in 2006 to get his 'perfect result'.

Can't have it both ways. Sometimes these guys just get lucky.

Lies, damed lies and statistics.

Nobody seems to mention that Cons always outperform their polling due simply to their typical voting demographic and turnout numbers.

Good luck Liberals on getting all those 18-20 yrs olds out to the station.

DL said...

"Nobody seems to mention that Cons always outperform their polling due simply to their typical voting demographic and turnout numbers."

There is no evidence to support that. While the Tories did slightly better than the final polls in the 2008 election, in both 2004 and 2006, they did significantly WORSE than the final polls were suggesting.

I don't have a problem with a poll being in field over a ten day period as opposed to being done in two days - given that its summer and nothing much is happening anyways. Outside of an election campaign or some earth shattering even, having a long field period is a bit of a non-issue. If we were in the midst of an election campaign THEN I would say that a poll conducted over a 10 day period wasn't worth the paper it was written on - but we aren't.

Anonymous said...

As a final footnote on the Ipsos Reid poll, I had to share this comment (which I did also offer up to the ottlib, since he saw it coming days in advance) . . .

What bugs me most about Ipsos Reid is the way they defend their poll. Here's how Darrell Bricker "defended" the polls results for the poll he conducted, as cited in Jane Taber's G&M report today:

"Darrell Bricker, the Ipsos-Reid pollster who conducted the poll showing the huge Tory lead, defended his numbers yesterday, saying the Liberals are lacking momentum, in part because Mr. Ignatieff is a 'cipher.' Canadians don't know him or his policies."

Note he did NOT defend the poll by talking about the objective nature of the questions, their formula for determining normalizing the results, the breadth of their polling data across the nation, etc. You know, factors one would logically assume are relevant to accurate polling.

Instead, he defended it by citing the poll results! His logic is essentially, "Our poll results - which fly in the face of every other poll in the past 2 months - are accurate because, well, the poll findings show the results are logical."

Got it. I'll try that if I get pulled over for speeding.

"Well, you see, Officer, I couldn't be speeding because I was in the car, and from what I can tell, I was barely moving. My hair isn't even tussled, which it clearly would be if I were really going that fast. My assessment of the situation has to be correct. I WAS NOT SPEEDING!"

We'll see how that works out.

What credentials does one actually have to demonstrate to conduct a poll? Is logic included in the position description anywhere? If so, someone should explain to Darrell that circular logic doesn't count.

Steve V said...

"Nobody seems to mention that Cons always outperform their polling due simply to their typical voting demographic and turnout numbers."

Then how did NANOS nail it in 2006, you "talking in circles' boob?

NANOS was off the last election, but his final night of polling did capture a late break for the Cons, so I give him credit for the trend.

The reason people prefer one over the other is based on predictive record. You want to discard that, because WELL you're a Con, whereas I let the facts drive my ranking. For instance, feel free to reread all my posts lauding AR online polling and detailing their impressive record. Then see how friendly to the Libs they've been ;)

You read like a boring hack.


I suppose the other way to look at it, he's having to "defend" his polling, never a good thing.

Steve V said...

"Good luck Liberals on getting all those 18-20 yrs olds out to the station."

Further evidence that you know nothing. Just a mediocre blowhard with a keyboard. Get a clue.

Anonymous said...

More evidences that Ispos was way off:

EKOS (Aug 19-25, 2009): CPC 32.6, LPC 30.9, NDP 15.7, BQ 9.5, Green 11.3.

Results (for some reason) are not yet displayed on the CBC or Ekos website.