The projection has moved rather wildly. Despite the Liberals actually dropping marginally to 40.7% (or between 39% and 45%), they have jumped to 72 seats from 64 yesterday. The PQ has slid to 28.8% (or between 28% and 32%) and just 46 seats.
The likely ranges for the two parties do not overlap, with the Liberals at between 62 and 79 seats and the PQ between 42 and 56 seats. The extremes do overlap, but even then the PQ is just projected to be capable of a maximum of 63 seats, the bare minimum for a majority government.
The CAQ has increased its projected support to 18.4% (or between 17% and 20%) but its seat haul has inched up one to just five (or between two and six). QS is up to 10.1% (or between 9% and 11%), and is still projected to win two seats. The maximum range now makes a third one possible, but we have to keep in mind that the Ipsos poll showing such big gains for QS on the island of Montreal is still just one poll.
Note that the projection uses Ipsos's likely voter numbers, rather than those of eligible voters. Those likely voter numbers match the most recent set of numbers from Forum very closely (except for QS).
I'd also like to take this opportunity to talk about probabilities, as I received a question concerning that on Twitter. The graph below spells it out:
In the case of the current projection, what does this mean? Well, it means that the Liberals have a 65% chance of ending up between 62 and 79 seats (there is a 25% chance of the result falling within the low-to-average projection, and a 40% chance of it falling within the average-to-high). For the PQ, there is a 65% chance that they will win between 46 and 56 seats, or an 80% that they will win between 42 and 56 seats. The CAQ has a 59% chance of winning between two and six seats, and QS has a 77% chance of winning between two and three.
There is a 95% chance that the results of the election will fall within the minimum and maximum numbers for each party, and a 5% that the results will fall outside of it.
What this means in practice is that, while the high and low ranges are the most likely results, there is still a decent probability that the results will fall outside of them. But, of course, it becomes less and less likely as you approach the minimum and maximum results. It also means that the PQ and PLQ are more likely to outperform the average result (77.5% chance for the PQ, 52.5% chance for the PLQ) while the CAQ and QS are more likely to under-perform the average result (55.5% chance). Option Nationale and the Greens have a 92.5% chance of ending up lower than the projected average.
Now let's move to the Ipsos Reid poll. The firm was last in the field March 14-18. Since then, the Liberals have held steady with 37% support, while the PQ dropped four points to 28%. Both the CAQ and QS increased by three points, to 19% and 13% respectively.
None of those shifts are outside the margin of error of a probabilistic sample of this size.
Among voters Ipsos considers the most likely to cast a ballot, the Liberals were again steady with 40% support. The PQ was down five points from the March 14-18 poll to 28%, while the CAQ was up four points to 18% and QS was up three points to 12%. Again, none of these shifts appear significant.
One shift that was, however, was the decrease of undecideds from 12% to just 7% of the sample.
Among francophones, the Liberals and PQ were tied at 31% apiece. That represents a seven point decrease for the PQ, outside of the MOE. The CAQ and QS were up to 21% and 14%, respectively.
A few notes, though, about the language numbers. According to the crosstabs of the Ipsos poll, French-speakers took up 87.3% of the weighted sample, whereas English-speakers took up the remaining 12.7%. This is borne out by the reported weighted sample sizes, as well as by a simple calculation of reported support by language (add them up at those weights, and you get the total result). This would seem to correspond with reported "first official language spoken" responses in the last census, but does not seem to be comparable to the francophone/non-francophone (combining allophones and anglophones) definition used by some other polling firms. That does not invalidate the numbers whatsoever, but would make them different from those reported by CROP and Léger, and so not comparable.
Practically speaking, this means that whereas Léger and CROP would place an allophone in the non-francophone category, Ipsos places them in either the French or English category depending on which language they did the survey in. So, for example, while Léger or CROP would put an Arabic-speaker whose second language is French and a Mandarin-speaker whose second language is English in the non-francophone category, Ipsos would put the first in the French category and the second in the English category. This is the effect of defining language by 'mother tongue' vs. 'language of use'. This may explain why the PQ and PLQ were tied among francophones, whereas among native French-speakers the PQ would probably still hold an advantage at a 37% to 28% provincial split.
And just to re-iterate, I am not questioning the reliability of the language results of the Ipsos poll. Readers just need to be aware that because the numbers are based on a different definition, it would be inappropriate to compare the results to those reported by CROP and Léger (and vice versa).
Back to the poll - the Liberals led on the island of Montreal with 45%, followed by the PQ at 23% and QS at 19%. That represents a big jump for QS worth nine points, outside of the MOE even at these small sub-sample sizes. But no other poll has shown QS that high in Montreal, so it is difficult to determine whether or not this is a wobble.
In the suburbs around Montreal, the PQ led with 32%, followed by the PLQ at 28% and the CAQ at 22%. Both the PQ and Liberals were down from Ipsos's previous poll, but not significantly so.
In Quebec City, the Liberals were up to 37%, followed by the PQ and CAQ, who were tied at 24%.
In the rest of the province, the Liberals were in front with 39%, with the PQ at 28% and the CAQ at 21%. That was an increase of six points for the CAQ in the region.
Ipsos found that the Liberals and PQ still have the most committed voters, with 81% and 78% respectively saying they would not change their minds. Just 57% of CAQ and QS voters said the same thing.
If those uncommitted CAQ and QS voters switched, both the PQ and PLQ stand to benefit. 36% of CAQ voters said the Liberals were their second choice, while 26% selected the PQ. Among QS voters, 33% said the PQ was their second choice - and 18% selected the CAQ. A bit of a leap for them.
On who would make the best premier, Philippe Couillard led with 33%, followed by Pauline Marois and François Legault at 25% apiece. That is bad news for Marois, as she has normally been in a close race with Couillard on this question, not Legault. He has certainly improved his party's standing in the last week.
But where will things go from here? There will be almost an entire week between this Ipsos poll and election day, with plenty of scope for intentions to change (they have been changing from Day 1). Unfortunately, we may only hear from Léger before Monday, so our understanding of the final days of the campaign may be limited.