i think you're jumping the gun on claiming the conservatives are ahead in ontario (it's still within the margins and some firms are still putting the liberals ahead), recently released riding polls have suggested the ndp are still positioned to at least win several seats in bc and the same basic dynamic of the undecideds leaning heavily between the ndp and the liberals is still at play (so, i think the model is still biased). that said, i don't deny that the most likely scenario at this time is a conservative minority, even if i think your model is overstating that fact a little. but, i think what we've seen over the last few weeks is that this is becoming a two-way race between the liberals and the conservatives in seat count, if not at the riding level in most places outside ontario. it seems to me that that momentum is slowly leaning liberal, and may begin to pick up in the next few days.

i mean, on the most basic level, you just need to look at the detail in the costing they just released. the liberals released a nice document. the ndp released a spreadsheet, without comment. it's these sorts of little things that add up that make one party look more serious than the other; it's been a comedy of errors for the ndp for weeks. and, yes, the trend is discernible in some of the polling.

ontario will swing the election, no doubt. but the biggest question is actually where the liberals are *actually* running at in quebec. they're polling anywhere between 15 and 30. if it's closer to 15, this is still a three-way race just because the ndp sweep quebec. but, if it's closer to 30 then even a small movement from the ndp to the bloc could split the sovereigntist vote and put the liberals in a strong position to win urban and english seats that are currently not considered in play. the trends seem to be pointing towards the 2011 numbers, but with a 10-15% swing from the ndp back to the liberals - that is, a lot of federalist ndp support going back to the liberals. the significant swing from the liberals to the ndp in the last election was lost in the major shift from the bloc; the reversal of this seems to be the only thing that's really firming up in the province, as the bloc and conservatives are really just rebounding to where they were last time. if the ndp end up perceived as sovereigntists by federalists, and federalists by sovereigntists, as appears to be happening, then they don't just lose their advantage but could be in some trouble altogether. rather than having their cake and eating it, they could end up without a cake at all. speculative, but it's where the numbers seem to be going.

the bloc don't have to actually win any seats to be a major factor in this election. they just have to run at high enough levels to split the ndp vote in key ridings. and, yes, that will help the conservatives in quebec city, too. but, it's trying to figure out whether the liberals are in a distant second or not-so distant second that is the key in the modelling, as there's a good 25 seats in it.

rather than look at the king-byng affair, or the 2008 scenario, for precedent, or even the liberal-ndp accord in ontario, i might suggest looking at joe clark's tactic. he prorogued the parliament for months. he ultimately lost. but, harper is in a better position.

if harper can prolong the speech from the throne until some time next year, the political calculus may very well be completely different. and, if he can hold off long enough, the chances of an immediate election after the throne speech fails increase quite a bit.

i've been toying with the idea of the liberals abstaining from the throne speech to avoid an ndp government and give them space to rebuild, but that relied on them being clearly in third. with the ndp being down as far as they are, and the likelihood of them making choices that will upset their own voter base, the liberals are not in nearly as much danger in letting the ndp govern for a few months. considering that the ndp are likely going to have to immediately deal with the tpp, a quick election could even see them benefit from ndp support bleeding left.

so, i think tactics are likely something like:

1) conservatives want to drag their feet as much as they can. they'll put off the throne speech for ten months, if they can.
2) the ndp will vote against everything, regardless of the outcome. they will do whatever they can to get in, then deal with it later.
3) the liberals are going to want to ensure that the arrangements, whatever they are, are soluble on their own terms so they can force an election at the right time.

www.huffingtonpost.ca/2015/09/26/election-2015-seat-projections-conservatives_n_8200422.html

--

Alanna McKechnie
I'd really like to see how these polls are actually done, because it seems like you guys are reporting on every poll given, and giving wildly inconsistent news in doing so. Everytime I read one of these it's "______ is pulling ahead, and ________ is in trouble because of it" with the three party names swapped at random, and I truly can't imagine the debates are swinging the numbers as much these are implying.

Basically, we just need better, 21st Century polling in this country, because it's clearly making your jobs very difficult and is frustrating to read. Another thing that would make these less frustrating to read would be starting your autoplay ads on mute, you monsters. Don't contribute to that crappy way of bombastic advertising, it's awful.

jessica amber murray
the problem we're in right now is that the polling industry is doing a lot of experiments with techniques that don't involve random sampling. they're trying to get around low-response rates. but, the thing is that low response rates are a much lesser problem than eliminating randomness. we're getting results all over the place because they're using all kinds of weird, untested methods.

what we actually want to get back to is solid, 20th century polling - it's the 21st century "innovations" that are confusing the hell out of everybody. maybe this had to happen this election to demonstrate the problems. but, i think we'll see some back-to-basics movement in the polling industry, soon.

you're right to point out that aggregates aren't a good way to predict poll results, though. the idea is workable if you have equally good polling methodology with equally large sample sizes and they work over the same time period. but, averaging out an internet poll from monday-thursday with a phone poll from tuesday-friday is just creating muddied data.

you want to look at polls that have large sample sizes and use phones. the good firms are ekos, nanos (with a caveat - he uses a three day rolling average, which i think is less good), innovative (when they use phones), mainstreet and forum. the bad polling firms are angus reid, ipsos, abacus...

what you'll see if you look carefully is that the good firms are consistent: harper is clearly down considerably over his 2011 results, but he's still positioned to win a minority. that has been true since the writ dropped. it has also been true since the writ dropped that there's a large base of voters torn between the ndp and liberals that has the potential to swing the election.

--

Stephen Solyom
Your analysis weights EKOS, which appears more than ever to be an outlier, far too heavily, and does not jibe with anyone else's analysis. You started by giving the Liberals absolutely no chance, you began to grudgingly accept that they had a growing trend, and with one poll, you have decided they are absolutely sunk. A cynical person might suggest that you were attempting to manufacture results.

jessica amber murray
the thing is that the ekos polls are almost unique right now in their methodological superiority. if you consider the margins, and you look at the other polls, it's reasonable to suggest that it may be a slight exaggeration. but, that still puts the conservatives around 33. and, there's really little reason to question that result, given that this is within the margin of error of firms putting them at 31 or 32.

i don't like the way that he gives incumbency an advantage, but we'll see if he's right in due time. but, if i were to suggest a modification of the aggregate calculations, it would be to remove the online polling altogether, which would actually favour the conservatives in the models.

his basic conclusion that a conservative minority is the most likely circumstance is upheld by essentially all of the reputable polling agencies, even the ones that have the conservatives trending lower. and you have to understand that the vote is distributed in such a way that they could conceivably end up in third and still win a minority. i don't like it, either. but, it's what the numbers say.

the correct criticism of this is that it skews and exaggerates a correct conclusion, not that it's an incorrect conclusion, in itself.