Monday, September 3, 2012

Québec 2012: My Two Cents

Although I haven't built a formal projection model for Québec, I couldn't resist playing around with the numbers. First, we run into the problem of projecting the popular vote. This is even trickier than usual, as I will argue below. Four firms have released polls in the last few days:

CROP: 8/27-29, sample size 1002
Léger: 8/29-31, sample size 1856
EKOS: 8/31-9/3, sample size 1749
Forum: 9/3, sample size 2781

Their numbers:
CROP: PQ 32,    CAQ 28,    LIB 26,    QS 9
Léger:   PQ 33,    CAQ 28,    LIB 27,    QS 7
EKOS: PQ 36.0, CAQ 24.5, LIB 23.2, QS 10.7
Forum:  PQ 36,    CAQ 25,    LIB 29,    QS 6

What to make of this? First, here are some things to keep in mind:

1. Forum also polled 8/27-28, and had numbers very close to those of Léger and CROP. So there may indeed have been movement from the CAQ and QS to the PQ.

2. Polling during Labour Day weekend can be good because it's closer to Election Day. It can also be bad because the sample may be different. This puts the movement posited in point 1 in doubt.

3. Forum also had a weird poll after the first debate showing the Liberals at 35%, completely out of whack with everything else we've seen.

4. In the 2008 election, the Québec firms (CROP and Léger) had the PQ at 29% and 32%. Then Angus Reid, who had been quiet all campaign, swooped in showing the PQ at 36%. The PQ finished with 35.15%. Moreover, in the 2007 election, the Québec firms had the ADQ at 25% and 26%. Then Angus Reid, who had been quiet all campaign, swooped in showing the ADQ at 30%. The ADQ finished with 30.84%.

5. The ADQ, the CAQ's predecessor, beat the average of the last CROP, Léger and Angus Reid by 3.8% in 2007 and 2% in 2008.

6. EKOS usually overestimates the Greens federally. The same may happen with QS.

7. I am not aware of any track record for EKOS and Forum in Québec provincial elections. Will they play the role of Angus Reid? Or will they miss badly, being newcomers to the Québec provincial scene?

As you can see, it's quite a bit harder to interpret these polls than in a "standard" election, since no pollster with a track record was in the field in the last 3 days of the campaign, which was a summer long weekend. My best guess would be:
PQ 34.5% (30.5-38)
CAQ 28% (23-32)
LIB 27% (23-31)
QS 6.5% (5-10)

My call for the seat count is:
PQ 68 (35-82)
CAQ 25 (16-50)
LIB 30 (22-50)
QS 2 (1-4)

You'll note that my ranges are usually just a little wider than 308's, except on the low end for the PQ and the high end for the CAQ. This is to account for the possibility that the CAQ gets a repeat of the 2007 ADQ ballot box bonus - not the most likely scenario, but also not a remote one. However, even though all the ranges overlap, the following scenarios are extremely unlikely:

- the PQ finishing third, since the CAQ and the Liberals likely won't both be high, as they share the pool of federalist voters;
- a Liberal win, since the PQ and the CAQ likely won't both be low, as they fight each other in many ridings.

These could only happen in a "perfect storm" of sorts, where all three parties end up with 30-31%. Interestingly, the Liberals' only faint hope at finishing first involves the CAQ doing well, since this would hurt the PQ much more than the Liberals. However, since the beginning of the campaign, the CAQ and the Liberals have not combined for over 60% in any poll (if I remember correctly), so this is very unlikely.

In fact, the scenarios that I find plausible as to the ranking of parties and majority/minority remain the same as last week. However, the latest two polls showing the PQ at 36% have reduced the chance of a CAQ win, and increased the chance of a PQ majority. I'd say that there's a ~60% chance of a PQ majority, a 30-35% chance of a PQ minority, and a 5-10% chance of a CAQ minority.

Whatever happens, it'll be a fascinating election night tomorrow. Have fun!

No comments: