I'm glad to introduce the first projections model for the next Ontrio provincial election. The model is similar to the ones currently used for Quebec and Canada (i.e: federal level). In particular, on top of the seat projections, you also have the probability of winning (for each party in each riding but also for the election in general).
However, pelase keep in mind this is a preliminary model. It'll likely be modified and updated in the future. In particular, the regional coefficients as currently estimated using only the 2007 and 2011 elections. I'd like to use at least the 2003 as well.
As usual, a simulator is also available, minus the probabilities (so just the seats as a function of the percentages).
For the first exercise, I used the latest Ontario numbers from Abacus. The poll is about a month old, but the two more recent polls either have a small sample size (Nanos) or are from a pollster I don't really know. Beside, Abacus just nailed the Nova Scotia election, so why not use their numbers.
Seat projections:
Ontario Liberals: 30%, 39 seats
PC: 33%, 39 seats
NDP: 30%, 29 seats
green: 5%, 0
What is really interesting is to see that the NDP, despite being tied for 2nd place with the Liberals in term of votes, would be around 10 seats behind. And as far as the probabilities are concerned, this party would have a low 7% chances of winning. Still, it means that based on this poll and accounting for the uncertainty due to the sample size as well as the electoral system, we currently have a three way race in Ontario. The main three parties could potentially form the next government.
I'll admit that I don't follow Ontario politics very closely. Still, I'm surprised by the advantage the Liberals have with the electoral system (or, if you prefer, their vote efficiency). It seems to be as significant as the well known advantage the PQ has in Quebec over the Liberals, due to the concentration of the votes of the latter in the non-francophone community. The PC needs at least a 3 points lead in order to really be expecting to win.
Let me know what you think. I'm more than open to suggestions and corrections. As I've said, it's a first for me and you should see the current model as a preliminary one.
