8 Comments
User's avatar
KDD's avatar

And Albertans praying the UCP will follow the CAQ!!!

Adam's avatar

Do you think you’ll go back to the old model’s safe/likely/lean framework? Speaking for myself, I found that easier to follow since it’s similar to the system most American handicappers (e.g. the Cook Political Report) use.

Éric Grenier's avatar

Possibly! I have to look into it.

Hamza's avatar

Hi Éric, you mentioned how the Liberals and QS blocked the PQ in Montreal, Liberals blocked them in the Montérégie region, Conservatives in central Quebec and all 3 blocked the PQ in Quebec City. Were there any specific ridings you were watching during the election coverage that hinted at those broader trends? Was it mainly the 13 ridings that the model missed? Thank you and enjoying your BC analysis!

Éric Grenier's avatar

For QS, it was their two seats in Quebec City and Sherbrooke, as well as Rosemont and Maurice-Richard in Montreal. For the Liberals, it was those seats in the Monteregie that they were able to win, like Vachon and Montarville, and for the Conservatives it was seats like Louis-Hebert in Quebec City, Champlain, Nicolet-Becancour. All those tight races that the PQ needed to get over 64.

Rene Couture's avatar

Great work and thanks for keeping us updated on elections.

Anthony Piscitelli's avatar

Will you test your model with the actual vote percentages by region to see how it holds up? I always feel that is a useful test as it shows what result perfect polling would predict.

Éric Grenier's avatar

I just put the numbers in, and the results are indeed a little better: 62 PQ, 37 LIB, 19 CON, 9 QS, 0 CAQ. And the median projection (which I've switched to from the mean projection) would have been 58 PQ, 38 LIB, 18 CON, 10 QS, 1 CAQ.

It would have increased the correct calls to 117 out of 127 (92%) with only a few misses that weren't tosses.

Pretty good!