this is probably my final analysis of the polling in the upcoming election.

first, let us acknowledge that there has been a deficit of quality polling, and i'm less confident than i was in previous elections as a consequence of it. the aggregates are badly polluted with bad data, and the models are using poorly thought out approaches. it appears that the primary idea in this election is going to be a 20 point shift between the liberals and the ndp, and near stasis in the conservative vote. a swing model based on aggregating data is designed for small proportional changes and is going to completely fail at understanding how this is going to work with such a large shift between two parties.

what the models are going to do is apply the shift everywhere; if they're very good, they may frame a little, but they're all going to miss the point. you should not expect twenty point swings from the liberals to the ndp in every riding. there are plenty of rural ridings where the liberals were barely running at 20 in the first place. rather, what you should expect is minimal swing in rural ridings and exaggerated swings in urban ridings. so, you should expect little change where it doesn't matter - seats conservatives always win anyways - and 30+ percent shifts in downtown toronto.

if the liberals & ndp were both running in the high 20s, we'd have a nightmare scenario where liberal seats turn blue on a 15-20 point swing in the urban centres. but, the propaganda was too effective: we're rather looking at 30+ percent swings, and that should be enough to prevent the conservatives from making gains much of anywhere at all.

so, while it's dangerous to look at the data from a distance and conclude the ndp & liberals will simply switch places in the legislature, the balance of evidence suggests pretty much that. at this point, i expect the ndp to win almost all of the seats in toronto and most of the seats in the 905, as well as the urban seats in the southwest and potentially a few upsets in the rural southwest. but, exact numbers are going to depend on turnout.

the liberals appear to be polling strongly in a few unexpected places, and it may be exposing a new base. they've always done well in ottawa and will likely keep a few seats there. but, i think they're going to keep a few seats around the waterloo area, as well. these are the places where the new economy has asserted itself strongest, and the places where the liberals will need to rebuild from if they want to retake toronto.

this is my prediction:

1) conservatives: i think they'll come in under 35% of the vote, and that should be good for about 35 seats - all the seats they have right now, and a couple more around brampton, + the odd fluke.
2) ndp: i think they'll come in pushing 40% of the vote, but not quite. that should be good for at least 70 seats - all the ones they have now, and at least 50 of the ones the liberals have now.
3) liberals: they're going to beat the models, but it's not clear by how much. it would be absurdly optimistic to suggest they'll get to 25%, but i do think they'll beat 20%.  and, i can come up with at least 10 seats they should be able to hang on to.
4) the greens could win a seat, too, but it doesn't look like their vote is going to hit 5% province wide. if they win the seat, they could be a factor in four years.

so, something like the following is plausible:

ndp: 38%, 76 seats
conservatives: 33%, 35 seats
liberals: 23%, 12 seats
greens: 4%, 1 seat