Thursday, November 4, 2010

Good Conservative lead, better NDP result in Abacus poll

We last heard from Abacus Data during the New Brunswick provincial election. They're a new pollster at the national level, and it's good to have another polling firm reporting.

Their newest poll shows an eight point lead for the Conservatives over the Liberals, but also has the New Democrats at a very high level of support.Abacus has found that the Conservatives are at 33%, which is certainly within the norm. They have the Liberals at 25% and the New Democrats at 21%, numbers which might raise a few eye-brows but look a lot more plausible when you consider the 3.1 point margin of error.

Abacus uses an online panel in its polls, much like Angus-Reid. The undecideds in this survey numbered 16%.

In Ontario, the Conservatives lead with 36%, followed by the Liberals at 31%. The NDP, at 21%, is riding high. The Greens are at 12%. This would result in 50 Conservative MPs, 37 Liberals, and 19 New Democrats.

The Bloc leads in Quebec with 40%, while the Liberals, Conservatives, and New Democrats seem to be in a logjam. They're at 19%, 17%, and 16%, respectively. The Bloc takes advantage of the weakness of the three federalist parties, and would win 54 seats. The Liberals would win 13, the Conservative seven, and the NDP one.

The Liberals lead in Atlantic Canada with 42% and would win 23 seats. They are followed by the NDP at 23% (four seats) and the Conservatives at 21% (five seats).

Wary of small sample sizes, Abacus clumps British Columbia, Alberta, Saskatchewan, and Manitoba together. The Conservatives lead in these four provinces with 45%, followed by the NDP at 26% and the Liberals at 19%. The Greens are at 10%. The Conservatives would win 63 seats here, while the NDP would win 21 and the Liberals eight.

In total, the Conservatives would win 126 seats. While they hold a nice-sized lead, they are too weak to make any gains. The Liberals would win 83 seats, the Bloc 54, and the NDP 45. That would be an all-time best for Jack Layton's party.

Obviously, this poll is bad news for the Liberals. It isn't exactly stellar for the Conservatives, either, as they are doing badly east of Ontario and could even be doing better in that province.

It's terrific news for the NDP, but primarily because of the great result in Ontario.

It's difficult to look at a poll in isolation, but as this is Abacus's first, we have no choice. I look forward to seeing future polling results from Abacus Data, so that we can delve a little deeper into how the parties are faring.


  1. I looked a little into their polling methods, and they seem similar to Zogby Interactive,an American pollster, more so than Angus Reid. Zogby doesn't exactly get a high reputation down south of the border, so I'd take this in stride.

    However, can't really tell how well a poll actually fares when they've only got one out. Do you know of any plans they have on how much polling they plan on doing, Eric?

  2. Undecided seems far too low at 16%. Usually we see figures around 30%.

    Don't know what that's about but I wonder if this internet method is catching young people who won't vote and is artificially inflating Green and NDP by about 2 points each.

    35-27-19-10-8 would seem far more believable.

    We'll just have to wait until this polling outfit can be added to the house effects chart.

  3. Volkov, they seem intent on being around, so I suspect we will hear from them more often.

    They did not do very well in the New Brunswick election, however. As that is my only point of reference, they have a low weight in the projection.

  4. Shadow,

    You may be right about the internet voting point. It could net more younger voters, or even "progressive" voters, or even simpler, non-traditional voters. However, I don't know how Eric could ever compensate for that in his polling averages, you know? It's not as simple as picking the number "2" and switching around numbers.

  5. Volkov if week after week there is a sustained bias towards NDP/Greens that other pollsters don't show then when Eric configures his house effects table we'll have a pretty good idea what that figure should be.

    2 was just an eyeball from what we've seen lately, its by no means scientific or accurate!

    If this firm does have a sustained (and inaccurate) relative bias then Eric can account for it through his weighting system.

  6. It looks like Abacus (like Angus Reid) weights its sample to be representative of the broader Canadian population (I've never quite figured out what the methodology is for doing that - if anyone knows, I'd be curious). So, their sample shouldn't be too skewed with respect to observable characteristics (age, gender, etc). Of course, that can't control for unobservable characteristics. So if, as Volkov suggests, internet use is positively correlated with say being "progressive" and negatively correlated with being a social conservative (both being plausible hypothesis, although I don't know how true they - AR tends to get a lower (and more accurate) measure of green support and higher measure of Tory support) you'd end up with skewed results. AR seems to have been able to deal with this problem in their internet polling (at least at the federal level), so maybe Abacus can too.

  7. Shadow: Undecided seems far too low at 16%. Usually we see figures around 30%.

    Would that "we" be the people who state that 87.3% of all statistics are made up on the spot?

    EKOS gives an "undecided/ineligible to vote" number in each of their surveys. In the October 28th report (fine print, page 3) it was 13.2%, a typical number. Allowing for variations in polling methodology, the Abacus 16% is certainly plausible.

    Don't you hate it when those pesky facts keep creeping in?

  8. "Don't you hate it when those pesky facts keep creeping in?"

    John my facts aren't wrong, the situation is just more complex than you believe it to be.

    I followed the link to the actual survey and saw that they included leaners in their figures.

    (As does EKOS btw.)

    Nanos, who polls committed voters, and Environics who have a very, very soft second nudge question show undecided voters in their 20's.

    A segma poll taken before the '08 election showed 30% of the population was truly "undecided".

    Before the '06 election the number was around 50%.

    "Allowing for variations in polling methodology"

    appears to be the operative phrase indeed.

    Now that we know that figure includes leaners there is nothing unusual.

    If they were trying to say that only 16% of the population was truly undecided, (as you seemed to think) well now that would be something different!

  9. ...and on most Ipsos polls the undecided rate is about 6% - so go figure. One thing for sure is that "DK/NA" is vastly higher on polls by Nanos or Environics that do not prompt party names at all - than on all the other polls where they prompt.

  10. DL brings up again the issue of "prompting".

    I think he has a good point.

  11. Angus-Reid definitely does a pretty questionable job of weighing its online samples. They consistently seem to overestimate the "hip" urban party compared to other pollsters, and to results - the NDP in Canada, Projet Montréal in last year's municipal election, Smitherman in Toronto this year, the LibDems in this year's UK election.

    If this polling firm has a similar methodology, and it seems like it does, that, combined with the unusually high NDP number, makes me take this firm with a grain of salt.

  12. Joffre said: "[AR] consistently seem to overestimate the "hip" urban party compared to other pollsters, and to results - the NDP in Canada, Projet Montréal in last year's municipal election, Smitherman in Toronto this year, the LibDems in this year's UK election.

    In fairness, though, AR wasn't the only one to overstate Smitherman (Nanos was off on that one too - only Ekos got it right, and only because they were polling during the last week of the campaign when the big shift happened) or the Lib Dems in the UK (probably because Lib Dem support actually fell in the dying days of the campaign).

    For what it's worth, AR was the only polling firm to nail the 2008 election result (i.e., be within the margin of error of the actual results for all 5 parties (though, since I'm hesitant to give too much weight to that since it could be a fluke).

  13. DL Ipsos Reid both prompts and includes the leaders' names doesn't it ?

    That would account for the 10% difference from other pollsters who include leaners and a strong second nudge question.


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.