last night was eventful, but too cold to risk getting stuck outside in, so i did plan to get home, and did. i didn't get to sleep until close to 7:00. i'll do a review of the weekend up on monday.

for now: is it time to get predictive about the election, yet? if we're going to start seeing movement on the left, including in quebec, it's going to start soon, so take this analysis carefully. it could be out the window next week. but, based on what the polls are saying, what should we expect?

i'll start by pointing out that i think that the models that you see are fundamentally flawed in two ways.

first, they operate on aggregates, under the argument that it's a bigger sample size, but with no attempt to analyze the quality of the data. aggregates would be a good idea if all of the surveys were exactly the same. but, when you just throw everything into a giant pot and mix it up like they do, you're not "increasing the sample size", you're really just distorting the data. the bad surveys don't augment the good surveys by increasing the sample size, they just fuck up the sample. worse, you're doing apples and oranges comparisons across different sampling methods and modelling decisions. it's just not a good idea.

second, they smooth the data out, and they do it because they have to. so, they'll take a sampling frame called "alberta" and then calculate the results and then force them equally into calgary, edmonton and all of the rural areas. it's hard to take them seriously, for that reason.

in order to get good, predictive results, what we need are strong snapshots close to the election with far more detailed sampling frames than anybody wants to invest in. as it is, if you want to extrapolate the data well, you need to be knowledgeable about the different areas and essentially read into what the numbers are saying, by looking at what similar numbers provided for in the past.

the defenders of the models will claim i'm being irrigorous, and i don't deny that. i don't uphold this process as ideal. don't try this at home! but, the fact is that their models suck. this is less of a problem in the united states, where the states are smaller and more homogeneous, which is where these companies get their operating procedures from. until the industry decides to develop intrinsically canadian models from first principles, and revamps itself from scratch, starting with better sampling frames, this is the best that i can do.

so, what have we got?

the papers are focusing in on these day-to-day changes in polling leads, but the fact that both of the daily tracking firms that are publishing data are flipping every couple of days, and often contradict each other, demonstrates the point clearly enough: the parties are statistically tied in national polling numbers.

so, is it a coin toss? do the conservatives have a serious chance of winning?

no.

the regional polling has the liberals way ahead in ontario across the spectrum. if the liberals pull off ~45% in ontario, they're going to sweep the urban areas. if the ndp are really polling at 10% or lower, that sweep is going to be that much more pronounced. further, if the greens are really trending upwards in rural ontario, that's going to help the liberals, dramatically, as well. based on the data i'm seeing right now, it's reasonable to suggest that the liberals could win 100+ seats in ontario, by holding the gta and picking up a number of ndp strongholds.

if that happens, jagmeet singh should resign immediately.

there is a caveat, here: everybody has the ndp polling around 10-12%, except nanos, who is a bit of an outlier in this respect. he continues to have them pushing the top of the margin of error, that is closer to 15%, and maybe a little bit out of it. the signal appears to be significantly different. nanos is usually startlingly accurate, but he seems to be out of bounds on this. if he's right, the ndp may hold on to some of these seats. but, if everybody else is right, the ndp could be nearly wiped out of this province.

in the east, any increase in green support would appear to be cutting less into liberal support and more into ndp support. so, while it's tempting to think that the greens are just going to split the vote and help the consevatives, that doesn't seem to be actually true. i would expect the greens to poll higher than the ndp across the province, but that might just end up meaning a lot of second or third place finishes. it doesn't look like the picture is going to substantively change, here - the liberals should win most or all of these seats.

quebec is always messy, and in the end depends on turnout. with four or five major parties, polling data even a week before the vote is of minimal value, in quebec. but, the trends do appear to be towards the bloc and the greens, mostly at the expense of the ndp. at this point, it would be reasonable to assume that the bloc will steal most, if not all, of the existing ndp seats in the province. the liberals wanted to pick up setas here, but they'll be lucky if they manage to avoid losing any to the resurgent bloc, especially if the greens cut into their vote in any substantive way.

so, if the liberals end up losing 5-6% nationally, where is it, then? the answer is out west.

does it matter if the liberals go from 15% in alberta to 3%? broadly, no. they might lose a few seats. but, that will come out as an exaggerated drop in the national numbers. likewise, they may come in fourth place in saskatchewan, but they only have one seat there, anyways - and it's both the safest and least safe liberal seat in the country.

manitoba and british columbia are more important, with bc being more important than manitoba. the liberals do not compete provincially in manitoba, and the spectrum is shifted in bc. so, there are prominent provincial ndp parties that are overpowering the unelectability of jagmeet singh. the recent provincial results in ndp were not exactly inspiring, but they are better than the ndp is doing in most places, right now. so, the liberals could be in for some close fights with the ndp in winnipeg - but that's just a few seats. in bc, the liberals are going to get stung, but with the greens and ndp in a dogfight for the anti-pipeline vote, it's not yet clear who picks up the seats. if the greens get a majority of these seats, it's really "game over" for the ndp - even if they finish third, nationally.

when you add it up, then, the liberals are poised to win seats in ontario and lose seats pretty much every where else. so, it's going to be close - you're either looking at a very weak liberal majority or a very strong liberal minority. and, while it's not yet clear who the third party will be, it does not seem as thought it will be the ndp - it will either be the bloc or the greens.

the conservatives simply don't have the numbers in ontario right now to be looking at forming a minority government. gains made in the west may even be offset by losses in the east.

that's just the data, right now. we'll see what it says in a few weeks.

the liberals are supposed to do better than this