Tuesday, December 9, 2008

The National Projection vs. Results

The results last night confirm that my system works. Here are the results with the difference in my projection in brackets. I am quite proud in accurately projecting the two major parties so closely, and in opposition to how the other projectors saw things.

PLQ - 66 seats (-2)
PQ - 51 seats (-2)
ADQ - 7 seats (+3)
QS - 1 seat (+1)

And now the popular vote:

PLQ - 42.05% (-0.85)
PQ - 35.15% (+1.99)
ADQ - 16.35% (+0.89)
QS - 3.79% (-0.19)
PVQ - 2.18% (-1.65)

All of my projections were within 2% and the projections for the PLQ, ADQ, and QS were within 1%. The QS projection was even within 0.2%! The over-achievement of the PQ surprised me, both as a projector and a political junkie, while the under-achievement of the PVQ should not have come as a surprise considering they only nominated 80 candidates. In the future, I will have to take that into account. A lot of the people who told pollsters they would vote PVQ probably didn't even have a PVQ candidate in their riding.

A few words about ThreeHundredEight vs. the other projectors. As far as I am aware, there are only five who made regular projections using a projecting model. In terms of seat projections, ThreeHundredEight was the closest with an 8-seat margin. Next closest was that of the Geloso-Breguet blog, who had a 10-seat margin. They do have the distinction, however, of being the only one to predict 7 seats for the ADQ. Third place goes to DemocraticSpace with a 12-seat margin. Fourth goes to Quebecpol with a 14-seat margin, and last to HKDP with a 24-seat margin.

It should also be noted that ThreeHundredEight was the only one to project the Liberals with less than 70 seats and a small majority, and the only one to project the PQ with more than 50 seats and a strong opposition.

Some of those projectors have since shown that with the actual popular vote results, their model gave a very similar result to what happened last night. I'm not quite sure what that says. Any projecting model has to take into account the possibility that the polls are inaccurate, which is what my model did.

As for the popular vote, the title goes to DemocraticSpace with a 5.0-point margin. ThreeHundredEight is a close second with a 5.5-point margin, and Geloso-Breguet in third with a 10.3-point margin. The other two did not make popular vote projections.

In the coming days, I will take a look at my individual riding projections, region-by-region.

I made some predictions yesterday for stories to watch. How did they turn out?

I didn't think that the ADQ would become a recognised party in the National Assembly, giving him 4:1 odds. It didn't happen.

I gave Charest 5:4 odds that he'd form a government. He did.

I gave the PQ 3:1 odds of earning more than 34%. They did.

I gave QS 3:2 odds that they'd improve their vote from 2007. They did. I gave QS 3:1 odds that Khadir would be elected in Mercier. He was. And I have 4:1 odds that David would be elected in Gouin. She wasn't.

I gave the PVQ 2:1 odds that they'd improve on 2007. They didn't.

Finally, I gave a 65%-70% turnout 3:2 odds. I gave 3:1 odds for 60%-65%. The result? 57%. Quite a surprise.