Tuesday, October 16, 2012

Nanos Poll: 33.3% Con, 30.1% Lib, 27.9% NDP

As I said before the McGuinty bombshell, there has been some news over a recent poll by Nanos that has shown a big jump for the federal Liberals, most notably after Justin Trudeau made his announcement that he's running for the leadership of the party.

Now, correlation does not equal causation, and there has been a history of Nanos giving the Liberals a bit too much of a leg up. But this is the only federal poll since he announced, and while Nanos has some odd poll numbers from time to time, he's usually among the top pollsters in the run up to E-days.

Nanos Research (Federal - October 4th-11th, 2012)
Conservative: 33.3% (+1.1%) - 137 seats (-6  seats)
Liberal Party: 30.1% (+5.5%) - 108 seats (+34 seats)
New Democratic: 27.9% (-2.5%) - 93 seats (-25 seats)
Bloc Québécois (QC Only): 17.8% (-2.7%) - 0 seats (-2 seats)
Green Party: 2.9% (-2.1%) - 0 seats (-1 seat)
Changes are from the last Nanos poll


But, here's why I think this poll isn't as up to snuff as we should think it is:


I would be surprised if the Liberals managed to get above 20% in BC these days - now you're telling me they're above 40%? Come off it. The margin of error is 8.9%, a sample size of 126 respondents, so maybe we should expect this sort of thing... but that is a very skewed sample size, even taking the MOE into account. There is a similar MOE in the AB-SK-MN column (or Prairies), but those numbers are inherently more in line with the recent trends.

Ontario makes is also pretty high up there on odd results, but they aren't out of the atmosphere either. Quebec, meanwhile, barely changed from the previous Nanos poll.

What's worse about this poll is that if you shaved off 10% off the Liberal numbers in BC, the Liberal lead over the NDP in terms of numbers and seats is wiped out, though very close. Keep that in mind when you look at this poll.

Its what is called "voodoo polling" - or at least, one aspect of it. While Nanos isn't an actual voodoo pollster, like Campaign Research is for instance, they do tend to have odd samples at times. Maybe its a partial voodoo poll; not every region is mucked up, but there is definitely something strange about one or two of them.

No comments:

Post a Comment