Thursday, November 11, 2010

The Wide, Wacky World of Ekos Polling

I've said it before: Ekos polling is just freakin' weird sometimes. Indeed, sometimes its more than weird, it's positively wacked. This week's polling round-up is no different.

TOPLINE: Con - 29.4, Lib - 28.6, NDP 19.3, Grn - 10.7, Bloc - 9.3

The Conservatives under 30%? Hm. I can see the Liberal, and even the Dipper numbers as believable, but that CPC number is just weird. The regionals might help explain why:

BC: Lib - 30.3, Con 28.2, NDP - 20.6, Grn - 16.5
AB: Con - 55.7, Lib - 21.0, NDP - 11.2, Grn - 10.2
M/S: Con - 39.4, NDP - 29.3, Lib - 19.1, Grn - 11.5
ON: Lib - 34.8, Con - 32.9, NDP - 19.6, Grn - 10.3
QC: Bloc - 37.2, Lib - 23.0, Con - 15.7, NDP - 11.0, Grn - 9.3
AT: NDP - 44.5, Lib - 30.6, Con - 17.0, Grn - 7.3

The Conservatives do poorly in every region, even in Alberta, where 55.7% could be considered too low for them if that were the actual general election result. Indeed, the only place they hold leads are in the three prairie provinces. If it weren't for those provinces, the Liberals would probably have a lead right now. Does that sound familiar to anyone?

The Atlantic Canada result is slightly hard to believe, but according to, even at 45% the NDP could only manage to elect 7 members from the region. That's a kick in the pants if there ever was one.

An interesting comparison by Ekos shows side-by-side the difference between this poll, and a poll done by Ekos the same time last year:

It's a little bit too wide, but you can see all the information. The big thing to take away from this is the huge drop in Conservative support, from 36.6% last year, to 29.4% this year. The Liberals, meanwhile, have a two-point bump, while the NDP are up 2.5, from 16.8% last year, to 19.3% this year. Both the Greens and the Bloc are more or less the same, down and up 0.5% respectively. "Others," who Ekos didn't prompt last year, sit at 2.7%, explaining away some of the Conservative drop.

I'd question, however, how good of an idea it is to compare this poll, with some very strange regionals, to this poll last year, which seemed relatively in line with others. So, I did my own little investigation, and did a basic average of all polls between October 1st and Nov 11., this year and last year, and this is what I got:

Lighter = 2009, Darker = 2010
As you can see, the biggest, most noticeable difference is the Conservative number, who compared to last year, dropped 5 points, from 38.9 to 33.9. The Liberals bumped themselves up 1.6% from 27.1 to 28.7, while the NDP managed a respectable jump of 1.1%, from 15.5 to 16.6. The Bloc jumped 0.5%, from 9.2 to 9.7, while the Greens didn't change a bit. Which means 1.8% of the Conservative drop went towards "Others," which makes sense, given that this time last year, not many prompted for "Others" anyways.

Which means, all in all, the Opposition parties have gained at the expense of the Conservatives since this time last year, which makes sense overall. Without the famous "Your time is up" speech, the Conservatives have no fear-of-an-election bump to boost them, so the Liberals and NDP make some modest gains at their expense. What was a nearly 12-point lead, fell back to a 5.2-point lead, simply because the Liberals haven't managed to screw themselves up again.

So, take it as you will. Ekos shows more or less something similar, just not as exaggerated.


  1. A 28% upswing for the NDP in Atlantic Canada. A 17% upswing for the NDP in Saskatchewan/Manitoba. A 12% upswing in BC for the Liberals post-Campbell. All in light of results by all the rest of the recent polls.

    I know what to make of EKOS - this poll is not just an outlier, it is an embarrassment.

  2. In general, Ekos polling is relatively OK. It actually follows a lot of trends, it just ends up exaggerating some of the results, especially in the regionals. The 15-point lead for the NDP in the Atlantic region is definitely not apart of any trend, but heh, it's a small sample.

    But, overall, yes, an outlier this is not - just bloody weird.

  3. But, overall, yes, an outlier this is not - just bloody weird.

    "Bloody weird?" Yeah, like I said - an outlier; especially considering every single other recent poll. There is no burning issue that would contribute to this large of a shift inside of a week when everything is considered. This is Frank Graves to the rescue; that is all this is.

  4. I wouldn't necessarily call it an "outlier," just small sampling. The results in other provinces, like Ontario, Alberta, Quebec, and even BC to an extent are not that far off. It's just the two regions with the smallest sample sizes - the Prairies and Atlantic Canada - vary wildly from poll to poll, especially for Ekos.