Failures of imagination and arithmetic

Colby Cosh's latest includes this explanation as to why he wants to write off the party which holds a strong lead in Alberta's polls:

The province-wide NDP numbers, whichever set you prefer, are conceptually hard to translate into large numbers of seats outside Edmonton. Former Calgary alderman Joe Ceci, running for the NDP, is thought to be strong in his old stomping ground of Calgary-Fort, as is Shannon Phillips in university-influenced Lethbridge West. There is bound to be a third name on this list—a name no one knows yet. Some shrewd, hard-working NDP candidate is knocking on the last door in his riding right now, scarcely suspecting he is about to beat some country-fried PC cabinet minister who got careless.

Alberta New Democrats’ dreams are much wilder than this. They envision a chaotic “wave” election: a butterfly farts in Banff and they unexpectedly quadruple or quintuple their support, as they would need to, in seats outside Edmonton. But unless things have changed very dramatically in places like Cold Lake or Drumheller, the New Democrat ceiling is a couple of dozen seats. That would leave the overall outcome to be decided by 60-plus PC-Wildrose fights in Calgary and the hinterland.

Needless to say, this type of analysis is rather familiar: in 2011, more than a few pollsters and pundits looked at the NDP's commanding lead in Quebec polls, then declared based on past results that the party couldn't hope to win more than five to ten seats in the province. This line of analysis did not prove prescient.

Granted, Cosh allows for the theoretical possibility of a "wave" while simultaneously writing it off. But in so doing, he misses the point that what he considers to be dramatic change is already playing out in the polls.

No, it's not preposterous - nor even unexpected - for a party to quadruple or quintuple its previous support when current polls already show more than that gain. The NDP's 2012 support by region included results ranging from 5% in Calgary and 5-8% in rural Alberta; the most recent polling shows the NDP in the 25-30% and 30-35% range, respectively. So if the right standard is to ask whether the NDP can increase its 2012 vote by five times or more, the answer is that it's already there or further across much of the province.

That said, looking at the NDP's historical vote is probably the wrong way of approaching the question.

It's doubtful that anybody can claim to have a clear, riding-by-riding picture as to how votes will shift. But the regional polls consistently show a tight three-party race in Calgary and "rest of Alberta", including some with the NDP running ahead in those regions.

Which brings us back to the Quebec problem: how can any reasonable observer look at a party which projects to win as many votes as its opponents over a substantial number of seats, and default to a presumption that it will win none of those individual seats (nor indeed contend for any of them)?

I'd argue that absent some coherent explanation as to how a region can be split up into a series of two-way races which render the three-party numbers unreliable, the best assumption has to be that the distribution of seats will be broadly similar to the distribution of votes (subject to distortion by the first-past-the-post system). And even that type of theory can't be based on Cosh's assumption that the NDP will simply be out of the picture over a set of ridings where its vote share is similar to that of the PCs and Wildrose.

That means the fundamentals of the Alberta race boil down to this: the NDP has a significant head start based on its whopping lead in Edmonton, while the rest of the province is a dead heat where any of the three parties could win a large number of seats, but the best baseline assumption is a relatively even split.

Of course, it's theoretically possible that a party could manage to lose dozens of ridings by small margins to two competitors without managing to get over the top in a single one. (In principle, that could even result in the party winning the most votes without taking a seat. Thanks, FPTP!) But surely that outcome should be seen as a remote possibility, not a baseline expectation.

Similarly, a small shift at the end of the campaign could easily change the three-party dynamic in many different directions. But here too, there's no obvious reason to think a shift can only operate in one direction: surely the NDP is at least as likely as its competitors to be the one to nose ahead at the end, particularly since it's been the only party consistently improving its popular perception throughout the campaign.

To be clear, it's entirely possible that Cosh's assumed outcome might come to pass - and that risk will hopefully serve as motivation for NDP activists as the campaign draws to a close. But that doesn't mean there's any reason for Cosh or anybody else to write off the very strong possibility that the NDP can win as matters stand now.

Update: As Cosh notes, the column was written before this weekend's polls strengthened the evidence as to the NDP's position across Alberta. So readers may want to look to the projections and polls as they stood earlier in evaluating it.

[Edit: fixed wording.]