Friday, December 16, 2011

How to Project Quebec?

I'm coming across a bit of a quandry when attempting to project the province of Quebec's seats that takes into account a proper, regional balance of the Coalition Avenir du Quebec's new-found success.

My original model followed a basic average of the amount of voter drift between the older parties and the CAQ in three "super-regions" - Montreal, Capitale-Nationale, and the rest of Quebec - which of course broke down further in the smaller regions which make up those regions, modeled somewhat after the patterns of the electorate following Action Democratique's breakthrough success in 2007 (since it follows that the CAQ and ADQ, essentially the same parties both ideologically and now physically, would see breakthroughs in similar fashions as the ADQ did in 2007).

This, of course, lead to a somewhat patchwork-quilt style projection system that gave the CAQ huge victories of 90 to over 100 seats. Something of a probable conclusion when you're leading 35% to 22%, but still, given how abstract the calculation actually was, it was essentially theoretical and nothing more.

Now, with the ADQ merger/subversion, I have the opportunity to basically use the ADQ's 2008 base of support as the CAQ's base, which gives the model something of a more reality-based calculation to go off of. After all, what better way to model regional patterns of vote except through the results of an election? And given the CAQ's lack of existence in 2007/2008, that's kind of hard to do.

However, the difference was, to say the least, fairly noticeable when I put in the newest CROP poll (39-28-18-9, just fyi) into the models:

I probably don't need to point out how much of a discrepancy the difference is. For one, it's the difference between partial, if horrible, existence for the PQ, and nothingness.

Now you might be wondering - why such the big difference? Its come the fact that the regional "vote drifting" is not uniform in the province of Quebec, at least according the polls I've seen. According to this model, essentially there are a lot more votes in Montreal that will go to the CAQ than you'd expect if the old ADQ base simply rose up (about an 8% difference), which makes a big difference. Meanwhile, voters in the Capitale-Nationale region are supposedly more willing to stick with the Liberals and the PQ according to the polls, than they would if the ADQ rose again (a 10% difference for the PLQ). Then, if course, in the rest of Quebec, though there isn't a hilarious difference in terms of actual vote levels, vote drift is enough to hand the CAQ a huge amount of seats.

So, I'm not really sure what to do here. The ADQ model is based somewhat in reality, but to me, it would just penalize the CAQ too much in Montreal if there truly was some sort of shift between pequistes and caquistes on the magnitude that the polls keep showing.

And given the overbearing results that the federal election showed when Bloquiste and Liberal support shifted to the NDP in Quebec, I'm not convinced that an ADQ based model is the right way to go.

Anyone else have an opinion on this? Not being from Quebec (despite being an avid fan), I don't know the regional intricacies and local voting patterns. I think both models would predict well, but who knows right?


  1. CAQ needs part of the PQ vote to be properly projected. I would use 100% ADQ + 33% PQ

  2. That's somewhat similar to what the "original" model uses, though I forget the factual amount - I think it was closer to maybe 35-40%. It also includes a bit of Liberal and QS support as well.

  3. I think we can safely drop the PLQ and QS support. Maybe even raise the PQ support up to 35%

  4. Five seats for the Quebec Solidaire I don't know where they would get that. Even with the PQ down that much I know of only two riding were the QS can get seats in Montreal with 15%. Mercier of course and Gouin.

  5. Vanillaman,

    Mercier (which they hold), and Gouin, yes.

    But also Hochelaga-Maissoneuve (they got 13% in 2008), Laurier-Dorion (also 13%), and Sainte-Marie-Saint-Jacques (15%).

    Those are the five best ridings for the QS in 2008, when they got just 3.7%. Now imagine if they increase their vote count in these ridings nearly three times over. They become quickly competitive against weakened incumbents, especially pequiste incumbents.