Wednesday, November 3, 2010

NDP biggest gainer in newest Harris-Decima poll

The latest poll from Harris-Decima shows Conservative gains and Liberal losses within the margin of error at the national level, but also a New Democratic surge of three points, outside of the poll's 2.2 MOE.Compared to Harris-Decima's last poll in mid-October, the Conservatives have gained one point and lead with 33%. The Liberals are down two to 28%, while the New Democrats are now at 17%.

The Greens are unchanged at 10% and the Bloc Québécois is down one to 9%.

The data from this telephone poll does not include the number of undecided Canadians.

The three-point gain for the NDP puts them back into a good position. Their gains came primarily in Ontario, but also in British Columbia and Quebec.

In Ontario, the NDP gained five points and now stand at 17%. The Liberals lead with 36%, down two, while the Conservatives have dropped four to 35%. The Greens are up one to 11%. This would result in 48 Liberal seats, 43 Conservative seats, and 15 seats for the NDP. That's a loss of seven seats for the Tories compared to Harris-Decima's last poll, and a gain of eight for the NDP.

In Quebec, the Bloc is down four points but still leads with 39%. The Liberals are down two to 21% while the Conservatives are steady at a woeful 13%. The NDP is up one to 11%. The Bloc would win 55 seats, unchanged from the last poll, while the Liberals would win 15, the Conservative four, and the NDP one. These projections are also unchanged.

The Conservatives have gained six points in British Columbia and lead with 33%, followed closely by the NDP at 28% (up two) and the Liberals at 26% (up one). The Greens lost five points, and now stand at 12%. The Conservatives have picked up three seats, and are now projected to win 17. The Liberals are down two seats to nine while the NDP is down one to 10.

In Atlantic Canada, the Conservatives are up 11 big points to 45%, pushing the Liberals down nine points to 31%. I think it is safe to say that this is due to the small sample size. The Liberals would win 17 seats (down four), the Conservatives 12 (up three), and the NDP three (up one).

The Conservatives are up eight points in Alberta to 61%, and are projected to win all 28 seats. That removes the one NDP and one Liberal seat here from the last poll.

And in the Prairies, the Conservatives lead with 39%, followed by the NDP at 26%. This would result in 19 Conservative seats (down two), five Liberal seats (up one), and four NDP seats (also up one).

In total, the Conservatives would win 124 seats, down one from Harris-Decima's last poll. The Liberals would elect 96 MPs, down seven, while the NDP would elect 33, up eight. The Bloc is unchanged at 55 seats.

Troublesome for the Tories is that the Liberals and NDP could combine for 129 seats, five more than the Conservatives are projected to win in this poll.

Though at first glance it would seem that the good news is for the Conservatives, who have widened the gap between themselves and the Liberals to five points, the reality is that this would change very little for Stephen Harper in terms of seats. He actually loses one from the last poll and 18 from his current caucus, with the New Democrats taking full advantage. At 33 seats, this is still a net loss for the NDP compared to the 36 they have now, but in terms of trends this poll is good for Jack Layton.


  1. I appreciate that you're trying to make polling data exciting, but don't you think calling changes from poll-to-poll 'surges' and 'trends' slightly disingenuous?

    When I look at the Canadian polling numbers from the past year, the only trend I see is a flat line for all the parties.

  2. Ne trouvez-vous pas étrange que la taille de l'échantillon soit de 1000 et que la marge d'erreur ne soit que de 2,2%? C'est ce que le communiqué d'Harris-Decima indique. Il y a certainement une erreur.

  3. Échantillon de 1000 et marge d'erreur de 2.2% selon le communiqué d'HD... Certainement une erreur, non?

  4. Bertrand, c'est 1000 par semaine, donc 2000 pour les deux semaines.

  5. BC Premier Gordon Campbell just resigned!!!!

  6. @ Éric

    Merci! Je n'avais pas vu, effectivement le «each week».

  7. Éric: The latest poll from Harris-Decima shows... a New Democratic surge of three points, outside of the poll's 2.2 MOE.

    Yes, but... HD had the Dippers low two weeks ago compared to their general trend. This week they're above the trend line. Add the two margins of error and it's not clear there's been any change.

    They ain't that good and they ain't that bad.

  8. " Add the two margins of error "

    Yeah statistics doesn't work like that ...

  9. " Add the two margins of error "

    Yeah statistics doesn't work like that .."

    Well, yes and no. You're right that when the margin of error (19 times out of 20) is 2.2 a 3 percent gain is statistically significant.

    But that wasn't John's point. His theory is that the "real" level of support in the population has not changed significantly over the past few weeks (or at least has changed by less than 3%, which is probably right) and that all we're seeing is statistical fluctuations from different samples. And we can't rule out that possibility, because the 95 confidence intervals of both polls overlap (by a fair margin). Both this poll and the last poll would, for example be consistent with a true level of NDP support at 15.5 percent. I'll grant you that it isn't a terribly likely possibility that there has been NO change to the "true" population level of support for the NDP, but I doubt there has been a "surge".

  10. "Despite widening the gap to 5 points the tories win less seats" ITS ALL ABOUT ONTARIO. The libs gained back the lead over the tories and the NDP made big gains there so the spread means nothing.

  11. Shadow: " Add the two margins of error "
    Yeah statistics doesn't work like that ...

    Ummmm... I haven't taken a Stats course since my undergrad days, but I fail to see the problem. Both polls were within the MOE of the trend line. Neither casts doubt on the hypothesis that the NDP is on that trend line.

    Would you care to expand?

  12. John if you actually go back and look at previous HD surveys from the end of August to their first survey in september the NDP dropped 2% to 14 points.

    They stayed there for another 2 weeks.

    And now have increased outside the margin of error.

    So the suggestion that a movement that held for an entire month, and coincided directly with the gun registry debate, was all an illusion is highly problematic.

    John, support for a party CAN and DOES change on a weekly basis.

    Artificially sketching a line between all polling and assuming that any deviation was just MOE is incredibly unlikely.

    The number with the highest probability of being right in a confidence interval is always the EXACT number given.

    Real movement in public opinion is a far, far more likely scenario than a perfectly flat trend line from which all deviation is just MOE.

  13. One thing the recent seem to be telling me is that the cpc core voter is 30%. There is 30% of Canadians that would vote for Harper if he ran on an NDP platform to nationalize the banks and drive big business out of Canada.
    The Liberal core is steadfast as well. For a decade there were 35-40% of Canadians who would vote Liberal no matter what and no matter who was the Leader or what they ran on or what they did when they got elected. Not even almost managing the separation of Quebec from Canada could break the Liberal ‘s hold on their core.

    Then Adscam came along and knocked loose 10% leaving the Liberal Core at 25-30%.

    1) the failure to address the corruption of adscam, and adequately apologize and fix the party so it won't happen the minute they are back in power

    2) not having a real leadership convention after the Dion convention was hijacked by special interest minorities in the Liberal membership.

    3) and playing gotcha politics rather than developing grassroots policy

    has knocked loose another 5-10%.

    This is shown that only 20% who think Ignatieff would be the best PM. The 28% of people who say they will vote Liberal might not if Ignatieff is leader. They are breaking away from the core.

    Now there are 30% of Canadian that will vote CPC no matter what and 20% who will vote Liberal no matter what ... and 10% voting for the BLOC. Give the NDP 15 core support and the Green 5%.

    That leaves 30% of Canadians open to vote on issues and policies. If Harper and the CPC election machine can win 1/3 of the non- committed, undecided 30% he will get his majority government.

  14. Shadow: Real movement in public opinion is a far, far more likely scenario than a perfectly flat trend line from which all deviation is just MOE.

    Marvellous claims can be made if an inconvenient original statement is replaced by a fairy-tale concoction of choice.

    What was that original statement? Why, it's Add the two margins of error and it's not clear there's been any change.

    "Not clear" has somehow been transmogrified to a "perfectly flat trend line". Which any reasonable person would discount, meaning that the only other offered alternative, however silly, must be the correct explanation.

    Applying the law of the excluded middle to a non-boolean situation is a tired old device, but it may well convince the hard-of-thinking. Less credible readers will understand that NDP support has obviously moved over time, and most likely in the last two weeks in particular. However, there's no evidence that any significant movement has happened.

    Don't expect the NDP to skyrocket based on the last two polls. It is nice, however, to see Shadow arguing that the Dippers are on the roll.

  15. John public opinion takes the shape of a wave, not a line.

    And this statement is simply indefensible:

    "However, there's no evidence that any significant movement has happened."

    Its called a Harris-Decma poll. You're on the thread discussing it. Statistically significant (outside MOE.) movement occurred.

    BTW you would be easier to understand if you'd get your terminology right.

    Adding two MOE's together ?

    Everyone knows that compounds error and if using subsamples taken weeks apart introduces longitudinal bias into your results.

    (Not what you actually meant to say but non-regular readers will be forgiven if they can't follow you.)

  16. Shadow: Everyone knows that compounds error and if using subsamples taken weeks apart introduces longitudinal bias into your results.

    "Longitudinal bias"... what a lovely term. It's a pity that it's not one known to any statistician. As Wolfgang Pauli said, this isn't right; it isn't even wrong.

    Feel free to submit the last comment on a posting. However, it would appear more authoritative if it had at least a grain of truth.


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.