CAQ gains in Forum poll

The first new poll of the final week of the campaign emerged yesterday from Forum Research via the Montreal Gazette, the first survey to be in the field and publicly released in over a week. The poll showed a minor drop in support for both the Liberals and Parti Québécois, but also a significant gain for the Coalition Avenir Québec, echoing the results of the CROP/CAQ poll leaked on Monday. But the Forum poll is, again, not without its issues.

The Liberals remain on top in the projection with 41.3% of the vote, or between 40% and 45%, assuming normal polling error. The PQ has dropped to 30.7% (or between 30% and 34%), while the CAQ has jumped to 17.8% (or between 16% and 19%). Québec Solidaire slipped to 7.7% (or between 7% and 8%).

That is a big jump for the CAQ, but it doesn't give them many more seats. They are currently projected to win just four, though their high likely seat haul is now six (and their maximum is now eight). A complete wipe out of the party is no longer considered plausible. The Liberals dropped just one seat to 64, while the PQ was unchanged at 55. The ranges tightened up, however, securing the Liberal spot as the likely winner: they should win between 60 and 70 seats at these levels of support, compared to between 52 and 57 seats for the PQ. So a minority government is still possible, though only the PLQ is likely to form it.

In addition to the Forum poll, two riding level polls were released for the ridings of Papineau in the Outaouais and Nicolet-Bécancour in the Centre-du-Québec. I'll go over them at the end of this post.

Now to the Forum poll. A few notes before we get into it. The report says that 854 Quebecers were surveyed, but if you look at the crosstabs it seems that 697 were actually surveyed instead, and that 687 of them responded to the vote intention question. That increases the margin of error to closer to 4%, rather than the reported 3%.

Also, the language numbers reported by Forum and repeated by numerous media outlets (CTV, at least, had the wherewithal to mention how odd they were) are unweighted. In other words, perhaps too many francophones in Quebec City were polled, or too many older Quebecers. Those language numbers weren't adjusted to make them reflective of the general population. So, while Forum reported that the Liberals had 39% support among francophones and the PQ 30%, what this means instead is that the Liberals had 39% support among the francophones polled by Forum. Unlike other firms, like CROP and Léger who report properly weighted results even at the sub-sample level, Forum is reporting their raw results here. 

As the numbers were unweighted, they are not necessarily reflective of the population. And since the numbers make no intuitive or mathematical sense, then we know that they are certainly not reflective of the population. The overall numbers should be weighted properly and are thus fine, but the language numbers as reported by Forum should be ignored.

Forum was last in the field on March 19, in a poll first reported by the Toronto Star. Since then, the Liberals dropped four points to 41%, the PQ dropped three points to 29%, and the CAQ increased by six points to 19%.

Considering the small sample size of this poll, only the last shift in support appears statistically significant.

Québec Solidaire was unchanged at 7%, while the Greens, Option Nationale, and other parties split the remaining 4% between them.

From what we can glean from the language numbers, it does seem that the CAQ has indeed made a jump among francophones, as suggested by the internal CROP/CAQ poll. Roughly speaking, for the provincial numbers to be what they are the PQ would need around 35% support among francophones, with the Liberals at 32% and the CAQ at 23% (give or take a point or two for each party). That is quite close to the CROP/CAQ results, but I'd prefer to see a public Léger or CROP poll before confirming that the CAQ is on the upswing among this demographic.

Regionally, the Liberals led in and around Montreal with 48%, followed by the PQ at 28% and the CAQ at 14%. That was a gain of five points for the CAQ, outside the margin of error. QS was at 7% in the region.

In Quebec City, the Liberals were ahead with 47%, followed by the CAQ at 27% and the PQ at just 17%, a drop of 10 points. QS was third with 7%.

And in the rest of Quebec, the PQ led with 35% against 32% for the Liberals (a drop of 10 points) and 21% for the CAQ (a gain of six points). QS had 8% support here.

This would suggest that the CAQ's gains among francophones came primarily from the PQ in Quebec City and the PLQ in the rurul regions of the province (which does make some sense).

Forum reported an approval/disapproval rating of 31% to 63% for Pauline Marois, virtually unchanged from March 19. Philippe Couillard's score (45% to 42%) was also generally unchanged from the previous poll, but François Legault's improved considerably. His approval rating jumped six points to 54%, while his disapproval rating fell five points to 27%. He could indeed be experiencing an uptick due to his debate performance last week.

But, again, we need to see some more polls. Forum's results align with what the CROP/CAQ poll implied, so the numbers are not unusual in that regard. Nevertheless, that their poll draws such a disproportionately Liberal-friendly francophone sample, as suggested by the raw data, puts up some red flags. On the other hand, Forum has been a bit of a leading indicator in this campaign. We'll see what the others show.

Riding polls

Now to the riding polls, which have been oddly absent from this campaign after nearly one in five ridings were individually polled in 2012.

The most interesting is from Somum Solutions for Le Courrier du Sud, polling the riding of Nicolet-Bécancour. This is a useful poll due to the departure of Jean-Martin Aussant of Option Nationale. Who fills that vacuum?

The IVR poll by Somum was conducted on March 26 and surveyed 1,908 residents. Among decided voters, it found the Liberal Denis Vallée in front with 36%, against 30% for the PQ's Jean-René Dubois and 22% for the CAQ's Donald Martel (the incumbent). The Québec Solidaire candidate had 8%, while the Option Nationale candidate replacing Aussant had just 3% - suggesting that voters in 2012 were voting for Aussant, and not ON at all.

The current projection for the riding gives a range of 37% to 42% for the PQ, 25% to 28% for the Liberals, and 22% to 26% for the CAQ. So, it will be necessary to add this poll to the projection in order to reflect the unexpected Liberal strength in the riding. This will be done for the next update.

The other riding poll was conducted by Segma Recherche for 104.7 FM, polling the very competitive (at least in 2012) riding of Papineau in the Outaouais region. It was conducted between March 28-31 and polled 624 residents.

It found the incumbent Liberal, Alexandre Iraca, well ahead with 48% to just 28% for the PQ's Jean-François Primeau. The CAQ was well behind at 13%, with QS at 9%.

The projection currently has the Liberals at 39% to 45% in the riding, against 31% to 35% for the PQ and 14% to 17% for the CAQ. Though the projected results are well within the margin of error of the riding poll, the poll will still be added to reflect the Liberals' strength, though it will have much less effect on the projection than the one in Nicolet-Bécancour.

About the CAQ poll...

Starve the beast long enough, and it will eat anything.

There hasn't been a public media poll released in the Quebec campaign since March 25, and that poll was out of the field on March 23. In other words, the last information we have about the state of the race is now nine days old. A lot has happened since then. The information is terribly out of date, and may no longer reflect the actual situation.

Note: A Forum poll was released at the same time as I wrote this. So, there's that.

So when the Coalition Avenier Québec leaked a poll, that they ordered from CROP, to the media, it was jumped on with desperation. The poll first appeared on the website of The Globe and Mail, and the frenzy started with articles in La Presse, Le Devoir and Le Journal de Montréal and by the Canadian Press.

ThreeHundredEight.com does not use internal polls, and so the poll leaked by the CAQ will not be included in the model. This is not about reliability, as the CAQ apparently does not have its own pollster and instead ordered a survey from CROP. That is a reliable firm, and the report I have seen is indeed a normal CROP report. (Were the poll released by the exclusive pollster of a political party, I wouldn't consider the results at all.) There is nothing untoward in the wording or order of the questions. CROP was asked for a proper poll, and by all appearances they provided it.

Of course, the CAQ did not provide all of the results to the media. A cursory look at the PDF shows that roughly 16 pages out of about 50 pages are missing, containing about eight of the roughly two dozen questions asked. Now why could that be?

This is the problem with internal polls. If the poll had been bad for the CAQ, we would never have seen it. This might be the first poll ordered by the CAQ in this campaign (they aren't flush with money, after all) but we don't know if that is the case or not. Did the CAQ get some worse polls earlier in the campaign and keep the results to itself? Will the CAQ release any new polls it might order if the results are less favourable? That we have to ask these questions shows just why internal polls are problematic.

That the leaked report is a scan rather than the actual PDF tells the story - those 16 pages had to be removed somehow. But why? Were the results on those 16 pages unfavourable to the CAQ? Were the questions asked politically embarrassing? Did those questions reveal strategy or concerns that the CAQ has? Likely it is a mix of all three, but for our purposes the first is the most important.

It is hard to analyse a poll that is only partially released, with all the good results for the CAQ still in it and all the bad results hidden from view. The poll certainly has plenty of good news for the CAQ, but it likely also has plenty of bad news. So, while we can say that there are some positive signs for the party based on this leaked poll, we do not know if there are also some very negative signs for the party as well. If this was a normal poll and all of the information was available, my analysis might end up being that despite some good news for the CAQ, overall the party still has no chance.

I don't want to spend too much time on the numbers themselves, as I would then be playing the CAQ's game. But some of the results are interesting.

The headlines focused on francophone support (the poll was only of francophones, implying the CAQ has given up on non-francophones), and showed the Parti Québécois at 36%, the Liberals at 29%, and the CAQ at 24%. That represents a big drop for the PQ and a big gain for the CAQ. It also implies that overall support would be roughly 40% for the Liberals, 30% for the PQ, and 20% for the CAQ (interestingly, that is about what the just-released Forum poll shows). Majority territory for Philippe Couillard.

Yet the poll shows a majority (52%) expect the next government to be a minority, with just 32% thinking that it will be a majority government.

The poll also showed that Pauline Marois and Couillard have relatively mixed numbers in terms of whether francophones have a good or bad opinion of them, while François Legault and Françoise David have very positive scores. But in an interesting reveal of the CAQ's curiosity, they polled two other figures as well: Pierre Karl Péladeau (49% good opinion, 40% bad) and Jacques Duchesneau (33% good, 36% bad). Why did the CAQ poll about Duchesneau, who isn't running for them anymore? To get an indication of Péladeau's relative worth? To see whether Duchesneau's loss has hurt the party?

The most interesting result - and in this we really do see the CAQ's thinking - is on support for the charter and a referendum. The CAQ had CROP ask whether voters were for both, for one and against the other, or against both.

The result: 21% of francophones said they were both for the charter and for a referendum. But 47% said they were for the charter and against a referendum. This is the CAQ's butter zone. Another 3% said they were against the charter but for a referendum (no one is trying to appeal to these rare voters, except possibly Québec Solidaire), while 29% said they were against both (the Liberal clientele).

The poll also showed the power of perception. While 51% of those who watched the debate on TVA last week thought Legault won it, 41% who did not watch it also thought that Legault had won it, based on post-debate coverage. 

So what to make of this poll? Some of the entrails are interesting, but it is skewed too much by reporting bias (only the results the CAQ deemed positive for them were released). Let's wait until we see the next public poll by CROP or Léger before reaching any conclusions.