i’d say you’ve got a last minute bloc surge, there. forum is picking it up as well. nanos has….let’s call it a bump rather than a surge. i would agree with anybody suggesting that you can’t honestly separate it from possible sampling error, but the thing is that i think this makes sense. with the ndp down a large chunk as it is, a lot of what’s left are going to be scrambling to find ways to make sure neither the conservatives nor the liberals (who are nearly equally despised in much of quebec) can win in their riding. with those numbers, i think it’s clear that the bloc win a lot of seats.

i don’t think a late swing to the conservatives in bc is unbelievable, either. but nobody else is picking it up yet. and, you have to ask “where?”, too, and “from who?”. i’m skeptical about a swing in the lower mainland. rather. what i’m willing to believe here is maybe counter-intuitive. outside of the lower mainland, the haunting sceptre of a liberal minority, even, could be enough to firm up traditional conservatives that were leaning ndp on the perception they had a chance of forming government. most of bc is going to vote on a populist/elitist axis, which makes the liberals broadly uncompetitive. it’s the one place in the country where you’ll get ndp voters voting strategically for the conservatives, just because the spectrum is completely different. nanos *does* have the ndp down and the conservatives up a bit relative to last week, but it’s the same thing with sampling and margins. i’m not overwhelmingly skeptical, anyways. it was maybe hard to believe they were down 15-20 points.

it’s clearly the uncertainty in bc and quebec that are warping your numbers. i’m still leaning towards it being a consequence, or artefact, of the method. but these have been trouble points for pollsters recently, too. really: you’re not telling me anything absurd, right now. i’m just left with same concerns regarding the question of what are truly reasonable sampling frames.

--

i’ve been posting photos of the 2014 ontario election for comparison, but the numbers have stabilized higher than that – closer to the 2004 federal election.

this is a very different map than the models are producing, but i suspect it will be closer to the final outcome.