Thursday, September 12, 2013

Stability in new Abacus poll

The latest numbers from Abacus Data, released to Sun Media on Tuesday night, caused a bit of a stir. They put the three parties in a tight three-way race and placed the Conservatives narrowly ahead for the first time in any poll since Justin Trudeau became Liberal leader. But though the poll differs from the consensus opinion on where the parties stand, it is in complete agreement in terms of which direction they are heading.
That consensus opinion is reflected by ThreeHundredEight's current polling averages, which still give the Liberals a four-point lead even after including Abacus's findings.

But Abacus was last in the field in mid-June, like the other firms that have recently reported. And like those firms, it shows that nothing of real consequence has happened over the summer months in the polls.

Since that June poll, Abacus finds the Conservatives gaining three points to sit atop the table with 30%, while the Liberals were unchanged at 29% and the New Democrats were up one point to 27%. All of these shifts in support appear insignificant, assuming a probabilistic sample.

The one change that was outside the theoretical margin of error was the fall of the Greens by six points to 4%, though that puts them at a far more reasonable level of support than June's 10%. The Bloc Québécois was up two points to 8%, while 1% said they would vote for another party and 22% were undecided.

It is clear that Abacus differs from what the other polls are showing in terms of actual support, but the trends are similar. That is, perhaps, more important than where the parties actually stand. Why Abacus has been consistently at odds with other pollsters since the beginning of 2013 is less clear.

Abacus found the Conservatives ahead by two points among men and the NDP ahead by five points (the Liberals were second) among women. The New Democrats had the edge among voters between the ages of 18 and 44, the Conservatives had a wide advantage among voters between the ages of 45 and 59, and the Liberals were narrowly in front among voters 60 and older.

The large national sample reported by Abacus is deceptive, as it over-sampled Ontario (note that Abacus releases complete weighted and unweighted results, including for all respondents and committed voters, and kudos to them for that). This was re-weighted accordingly, but it means that the samples outside of Ontario are quite small. This leaves those numbers vulnerable to wider variations, which did indeed occur.

In British Columbia, the Conservatives led with 35% while the Liberals were up 15 points since June to 32%. The NDP fell to 28%, while the Greens were down 17 points to 5%.

The Conservatives increased by 15 points to 65% in Alberta. The Liberals were up to 22% and the NDP was down 16 points to 5%.

There was less statistically significant movement in the Prairies (39% for the Tories, followed by the NDP at 28% and the Liberals at 27%) and Atlantic Canada (42% for the NDP, followed by the Liberals at 40% and the Conservatives at 17%, a result that is significantly at odds with other surveys).

Ontario, which had a large sample of 784 decided voters, also showed stability: the Conservatives were ahead with 33%, followed by the Liberals at 32% and the NDP at 28%. The over-sample meant Abacus was able to give good-sized regional breakdowns: the Liberals were narrowly ahead in Toronto, with 36% to 34% for the NDP, and eastern Ontario, with 36% to 34% for the Tories. The Conservatives were well in front in the GTA, Hamilton, and central Ontario regions with 42% to 30% for the Liberals, and were slightly ahead in southwestern Ontario with 31% to 30% for the Liberals and 29% for the NDP. The New Democrats led only in the north, with 44% to 28% for the Liberals.

But Quebec was the most interesting result. Being an important electoral battleground for the NDP and Liberals, the unusual result is worth noting. Abacus gave the New Democrats the lead with 32%, one of the few polls to do so since Trudeau became Liberal leader. But unlike those few other examples, it was the Bloc Québécois that placed second with 31%, followed by the Liberals at 25%, a drop of nine points. The Conservatives had just 10% support.

Is this about Quebec's Charter of Values? It is popular among francophones, and both Trudeau and Thomas Mulcair have come out against it (the Bloc's support has been tepid, but that would have only been known after this poll was conducted). But Trudeau's opposition was much stronger than Mulcair's (at least, it was when this poll was done) so it follows that the NDP may have regained some support. But it could just be a coincidence.

One thing that is interesting about Abacus's numbers in Quebec, however, is how they are consistent with the trends recorded in Léger Marketing's latest poll:
Both Léger and Abacus were in the field at similar times recently and in June, making it useful to compare their numbers. The numbers themselves are quite different for the Liberals and Bloc, with Léger having given the Liberals 46% in June and Abacus only 34%, while the Bloc was at 15% and 23% in the two polls.

The numbers are still different now, with the Liberals at 35% according to Léger and 25% according to Abacus, while the Bloc is at 23% and 31% respectively. But the trends are almost identical: the Liberals slipped 9 or 11 points, the NDP gained five or six, and the BQ was up eight. Are we looking at an emerging trend?
In terms of seats, the Abacus poll would deliver the Conservatives a greater advantage than their one-point lead would suggest, with 123 seats to 95 for the Liberals and 87 for the New Democrats.

With the heated language we've heard from the NDP concerning the Liberals lately, it is hard to imagine them working together. But their combined 182 seats would give them the ability to form a working majority government, despite the Bloc's 32 seats.

Of note, these poll numbers (and the new boundaries) would give the New Democrats five seats in Saskatchewan, generally considered the best case scenario for the party in the province (they held their caucus meeting there this week, and much was made of their potential for gains). Mulcair did mention hoping to win a majority of seats in Saskatchewan, a reasonable hope to express for a party wanting to form government, and perhaps not as unrealizable as some have suggested. By my reckoning, at these levels of support the NDP is within range of another two seats, which would give them half of the province's 14 seats. Winning an eighth seat, though, would be very difficult.

But though the poll does have some good news for the New Democrats (Within-the-margin three-way race! Leads in Quebec and the Atlantic! Strong numbers in Ontario!), they should be as reluctant to take this poll without a grain of salt (a phrase I have come to loathe) as the ones that have put them well behind the two other parties. Abacus is showing a consistent trend with the other pollsters, suggesting that they are not discovering a new political landscape that will be borne out in other surveys. And that suggests that, though there are a few dissenters, the Liberals are probably still ahead.

14 comments:

  1. I noticed that these numbers break with Abacus' previous numbers in that they no longer seem to factor in "leaners", only the decided voters after an initial posing of the ballot question. Evidence of this can be found in the crosstables of their separate report on Trudeau' pot admission. Where results are broken down by "federal vote intention (includes leaners)", you'll notice that the weighted totals add up to 1,388, whereas in the voting intentions report the weighted total for "committed voters" is listed as 1,248.

    Interestingly, if you reverse-engineer the numbers with leaners in the Trudeau report, you get:

    Lib 30 Con 28 NDP 28 BQ 8 Gr 5

    Which leads me to conclude that the most of the undecideds right now are torn between the Liberals, NDP and Greens.

    Dom

    ReplyDelete
    Replies
    1. Yes, that is one way that Abacus currently differs from most other pollsters. Even so, the inclusion of leaners does not drastically change the results.

      Delete
    2. After I noticed that I did a quick scan of other pollsters' reports to see if any others also refrain from including leaners. Environics, Harris-Decima and Nanos don't seem to state whether or not they include leaners. All the rest do.

      Dom

      Delete
    3. I'd be surprised if HD, ERG, or NR do not (I know Nanos did in their Bell/Telus poll).

      Delete
    4. I heard Abacus took the opposite approach to leaners for their Ontario numbers. Which seems odd... or downright biased...

      Delete
    5. I noticed that and asked, and was told it was because that is the approach that had in May in their last poll, and wanted to compare apples to apples since the numbers were part of a broader presentation.

      I think it makes more sense, particularly outside of a campaign, to include leaners like everyone else, but as long as they're internally consistent...

      Delete
    6. Yah, do one or the other IMHO. Both tell us useful information.

      Delete
  2. And since Abacus is so far off from the other majors I have to think it is an anomaly we can ignore.

    ReplyDelete
    Replies
    1. 'Cuz, you know how reliable all those firms have been... :~)

      As Éric said, it's the trend that's important, not the individual results. And this poll largely supports the others in seeing nothing remarkable happen over the summer...

      Delete
  3. You forgot to put North numbers in it?
    Abacus pdf said North (3 territories) NDP 67%, LIB 33% and other 0%, and 2 NDP seats and 1 Liberal seat, not 2 Libs and 1 NDP?

    ReplyDelete
    Replies
    1. A poll of three respondents is not worth taking into account - though I put the numbers in for fun anyway, and still get 2 LPC, 1 NDP, as the LPC start from a better base in Nunavut and the Yukon.

      Delete
    2. Scratch that, I input the numbers incorrectly. Nevertheless, the usefulness of a poll of three people remains (it has a margin of error of +/- 57%!).

      Delete
  4. inflated Bloc numbers again, pfff

    ReplyDelete

COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.