Friday, March 4, 2011

Stability in new Harris-Decima poll

The latest poll from Harris-Decima shows very little change over the last two weeks, exactly what their last poll indicated. In fact, at the national level, all of the parties are exactly back to where they were in Harris-Decima's mid-January poll.

Compared to Harris-Decima's last poll taken February 17-27, the Conservatives have dropped one point but hold the lead with 36% support. The Liberals are up one point to 28%, while the New Democrats are also up one point to 15%.

The Bloc Québécois and Greens are each down a point to 9%.

This telephone poll has a margin of error of +/- 2.2%, 19 times out of 20. The margin of error for the Conservative lead of eight points over the Liberals is roughly 3.5%.

In Ontario, the Conservatives and Liberals are statistically tied at 39% and 37%, respectively. That is a drop of four points for the Tories and gain of three for the Liberals. The NDP is up two points to 14%, while the Greens are down two to 8%.

The Bloc is down one point to 39% in Quebec, followed by the Conservatives at 21% (+2) and the Liberals at 20% (+1). The NDP is down two points to 9%.

In British Columbia, the Conservatives are down four points to 31%, while the NDP is up 12 points to 31% as well. The Liberals are down two to 27%, putting all three parties in a statistical tie. The Greens are down six points to 10%.

The Conservatives (-6) and the Liberals (+4) are tied at 33% in Atlantic Canada, while the NDP is down one point to 23%.

In Alberta, the Conservatives are up six points to 62%, ahead of the Liberals at 19% (-2). The NDP is down six points to 6% in the province.

In the Prairies, the Conservatives are unchanged at 48%. The NDP is up three points to 23% and the Liberals are down 11 points to 11%.

It would appear that the only statistically significant shifts in support are the Liberal drop in the Prairies, the NDP changes in British Columbia and Alberta, and the Green slip in BC.

With this poll, the national projection is bumped to 27.9% for the Liberals, 9.8% for the Bloc Québécois, and 7.5% for the Greens. The Conservatives and NDP do not change.

With the results of this poll, ThreeHundredEight's projection model awards 12 seats to the Conservatives in British Columbia, 28 in Alberta, 23 in the Prairies, 51 in Ontario, 11 in Quebec, and 13 in Atlantic Canada for a total of 140.

The Liberals win nine seats in British Columbia, none in Alberta, one in the Prairies, 43 in Ontario, 13 in Quebec, and 14 in Atlantic Canada for a total of 81.

The Bloc Québécois wins 50 seats in Quebec.

The New Democrats win 14 seats in British Columbia, none in Alberta, four in the Prairies, 12 in Ontario, none in Quebec, and five in Atlantic Canada for a total of 35. Their BC result saves them here.

With the Conservative drop in British Columbia, Elizabeth May manages to eke out a seat win in Saanich - Gulf Islands, while the Bloc is not strong enough to topple André Arthur in Portneuf - Jacques-Cartier.

Also out this morning is a poll (February 24-27) by Harris-Decima on the economy. Interestingly, 36% of Canadians now believe the country is out of a recession, up from 19% in February 2010. Canadians are split between supporting the government continuing spending to create jobs (44%) and cutting spending to eliminate the deficit (48%).

Surprisingly, there is still a large portion of Conservative supporters (36%) who want the government to keep spending. And not as many NDP voters as you'd think (55%) agree.

I've heard from another pollster that these results from Harris-Decima are more in line with their numbers than the recent polls from Ipsos-Reid, Angus-Reid, and Abacus Data. I tend to agree that the recent Conservative spike was perhaps a little unrealistic. I expect that in the coming weeks more polls will put the Conservatives in the mid-to-high 30s rather than around the 40% mark. And whether the recent hubbub over election spending and the use by Jason Kenney's office of government materials for partisan fundraising will have any traction remains to be seen. I don't think it will - if the past year has shown us anything it's that the Conservatives are coated with teflon.

33 comments:

  1. The Liberals and Conservatives are tied everywhere outside of Alberta and Saskatoba. CPC are luckyfor their base there.

    ReplyDelete
  2. What it also shows using Eric's seat assessment is that the real power broker is the Bloc not the NDP.

    Interesting ??

    ReplyDelete
  3. "Surprisingly, there is still a large portion of Conservative supporters (36%) who want the government to keep spending. And not as many NDP voters as you'd think (55%) agree."

    Really - I am surprise by your take? Personally, I would think most NDP supporters would want to get the finances back in order. It must be those Bob Rae Dippers that were polled. In provinces like Saskatchewan, BC, Manitoba and now in NS, NDP has a solid record of keeping government expenditures in check. Federal and provincial Tories like Mulroney, Harper and Devine have a history of being big spenders.

    ReplyDelete
  4. I suppose it might be a logical fallacy to look at a more stable poller and assume that they're more correct than a headline grabbers... yet one can't help but think there's more value for pollsters and political columnists in rapidly fluctuating numbers.

    ReplyDelete
  5. I'm not sure how Elizabeth may gets elected with a 10% Green showing in BC - the Liberals and the NDP are running strong active campaigns in SGI so the non-Tory vote promises to be very, very split - and at 10% the Greens are showing virtually no growth from 2008 in BC. (not to mention that whatever the polls say about the Green vote - you have to deduct about 40% from).

    ReplyDelete
  6. Wheatsheaf - Everyone was a big spender in the 1980s. The widely held economic belief was that deficit financing was okay.

    And Harper has had a minority his entire time in office. The one time he tried to bring in an austere budget, the opposition threatened to form a coalition and seize power.

    ReplyDelete
  7. Tories lead the Liberals in Quebec (and have done in several other recent polls). Very interesting indeed.

    ReplyDelete
  8. DL,

    First off, the individual poll seat projections are based on the poll results, not on what they are likely to be on election day. If they had the Greens at 50%, I'd project for that 50%.

    Secondly, the Conservative drop is the main reason the Greens win in Saanich. They increase their support by a very little, true, but Ms. May starts from a high level of support to begin with. She isn't working from a base of 10%, as her performances in London and Central Nova demonstrated.

    I have the Greens at 29.2% in Saanich - Gulf Islands with this poll, followed by the Conservatives at 26.5%, the NDP at 23%, and the Liberals at 20.4%. I agree, the anti-Tory vote is very split.

    ReplyDelete
  9. Its interesting that for the NDP the way the vote is distributed in the HD poll is like a perfect storm. Despite national numbers that are down 3% from 2008 - this poll would yield almost no seat losses. in contrast the nanos poll that had the NDP at 19% wasn't so good seat-wise because of the way it was distributed (though i tend to think that if the NDP vote in Ontario actually went up from 18% to 23% - it would mean a lot more than just one additional seat!)

    ReplyDelete
  10. Eric,
    In the scenarios that you have evaluated, you might want to add one for consideration:
    What would happen in Quebec, hypothetically, if Jack Layton was unable to campaign, and Mr Mulcair became the leader shortly before the campaign began? Would the NDP take some of the Bloc vote?

    ReplyDelete
  11. As no polling data exists showing how Mulcair would do as leader, I don't believe I could write such a thing.

    I'm not sure Mulcair would do better than Layton. People in Quebec already like Layton, and he doesn't have a history (checkered, to be sure) with the provincial Liberals (or even Alliance Quebec), like Mulcair does.

    I think Layton is a good leader for the NDP in Quebec, and might even be better than Mulcair. On the face of it, they have similar backgrounds. Layton grew up in Quebec and went to McGill University, just like Mulcair. Mulcair's French is better, but they aren't really different individuals - at least in terms of backstory.

    Perhaps somewhat unfairly, Layton isn't seen as a Quebecer the way Mulcair would be, but aside from that there is nothing Mulcair has that would make him more attractive to Quebecers than Layton.

    ReplyDelete
  12. Just out of curiosity - how is that the NDP would lose its Quebec seat based on this poll? Its true that this poll has NDP support down 3% to 9%, but Liberal support is down 4% from 2008 to 20% and given that the NDP over Liberal margin in Outremont was 6% - how does your model have that seat changing hands on these numbers?

    ReplyDelete
  13. You have the ndp taking no seats in Quebec, so where does that leave Mulcair.
    You have them with 2 seats less than they now have, with none in AB or Que.

    ReplyDelete
  14. DL,

    An NDP drop to 9% is a drop of 25%. A Liberal drop of four points is a drop of 17%.

    ReplyDelete
  15. yes, but 39% less 25% is about 30% and 33% less 17% is about 28% and that's not even counting the fact that Mulcair ought to get a small "incumbency bonus"

    ReplyDelete
  16. Mulcair doesn't receive a bonus. The NDP has gained in Quebec and he is a veteran MP.

    Martin Cauchon also gets the star candidate bonus.

    ReplyDelete
  17. In your model, doesn't ANY incumbent running for re-election get some sort of bonus? It seems to me that in any riding where you have someone running for re-election - you have to assume that they will be at least slightly more resistant to the uniform trend against their party than if they were retiring and the party was running a generic candidate to succeed them.

    ReplyDelete
  18. As I've explained in Thursday's post, yes, incumbents do get a bonus when their party is dropping in support in a province or region. When the party gains, only first-time incumbents get a bonus. When there is no incumbent, the incumbent party is penalized.

    ReplyDelete
  19. Éric said:
    > In Ontario, the Conservatives and Liberals are
    > statistically tied at 39% and 37%, respectively.

    I know we hear this sort of formulation often, but it is statistically incorrect. When two parties are within the margin of error of each other, they are not statistically tied. There is still a better than 50% chance that the Conservatives are ahead.

    I cannot find what the margin of error in Ontario only is for this survey, but let's say it is 5 points. What this means is that there is a 95% chance that Ontarians would vote for the Conservatives in a proportion between 34 and 44%, and for the Liberals in a proportion between 32 and 42%. So it is possible that the Liberals are ahead, but the chance of this is lower than 50%. My statistics are rusty, but as I remember it would actually be possible to compute the probability that the Liberals are ahead.

    It's not a big deal, but it does sound a bit misleading and that's why it caught my attention.

    ReplyDelete
  20. Surely the CPC takes your analysis into consideration, especially seat projections, They are not interested in another minority. Somebody will cave as 32 years of history has told us. Harper is waiting for 165+ projection and when or if that happens a convenient pill will be added to the nonconfidence vote available at the time

    ReplyDelete
  21. Obelix,

    If I am not mistaken, the margin of error for the gap between the Conservatives and Liberals in Ontario is 6.2 points. The gap in this poll is two points.

    While it is possible that the gap is even larger than reported, it is statistically correct to say that the two parties are tied, as there is no statistical basis for giving one party or the other the lead.

    ReplyDelete
  22. On second thought, I think the wording that the lead is not statistically significant would be more apt.

    ReplyDelete
  23. You really shouldn't even include Ipsos-Reid, or take it seriously as a pollster. It has a history of overestimating Conservative support and giving the Cons larger numbers than any other pollster and giving the NDP the smallest numbers of any other pollster. Just look at the bizarre regional numbers like 6% in Quebec for the NDP when all the other pollsters have them in double digits there. Ipsos-Reid has even done things like predicting a Chrétien minority in 2000, and predicting majorities for Harper in 2006 and 2008.

    ReplyDelete
  24. Can I assume Troy that you also want to throw out the results from Nanos and environics for favoring hte liberals compared to other pollsters, then we can throw out ekos and HD for their green numbers. Crop and leger obviously overpolls ndp voters...


    But then again, that is just as the pollsters relate to each other. We can't expect every methedology to come up with the exact same result.



    But if you want to talk about pollster accuracy, eric has calculated that into his model using numbers from the 08 election, and a few of the provincials elections.

    http://threehundredeight.blogspot.com/2010/04/new-weights-for-polling-firms.html

    Did you know that based on that (admittedly small dataset... well, bigger than the one you used) Ipsos is among the more accurate pollsters? And that it is the HD, Nanos, environics that poll more heavily against the tories that ought to end up in the "less than serious pollster" pile.

    ReplyDelete
  25. Éric:
    > On second thought, I think the wording that
    > the lead is not statistically significant
    > would be more apt.

    Do you mean that with the numbers as they are, the probability that the Conservatives are ahead of the Liberals in Ontario isn't much more than 50%? I'd agree with you there. It is higher than 50%, but with a margin of error of 6.2 points, it's not much higher.

    ReplyDelete
  26. Troy out of the 13 pollsters Eric weights, IR is the 4th most reliable. Its much better than average.

    ---

    Eric good call.

    Statistical tie is frowned upon.

    "It’s certainly not a term that practitioners use,” Nancy A. Mathiowetz, president of the professional group American Association for Public Opinion Research

    http://blogs.wsj.com/numbersguy/whats-a-statistical-tie-anyway-234/


    Even the term statistically significant is often misunderstood.

    When a poll shows one party ahead of the other, regardless of MOE, there's at least a 50% chance that they are indeed ahead.

    In this case the odds would be closer to 80%.

    They just don't reach an arbitrary cut off, at least 90% probability, where the pollster can say with certainty that the lead isn't due to chance.

    ReplyDelete
  27. Troy - that's what the house effects are for.

    ReplyDelete
  28. Shadow,

    Won't use the term anymore, then.

    But my own reading has been contradictory to this statement from the link you provided:

    "That means that a difference between two candidates that is less than twice the margin of error is considered statistically insignificant."

    According to what I have read, that is not the correct method for calculating whether a gap between two parties is statistically significant. In most cases, it is roughly 1.5 times the MOE.

    ReplyDelete
  29. Eric I think its because that's an American publication.

    The 2 times MOE formula works for polls with strictly 2 choices, which is often the case in the US.

    ReplyDelete
  30. I think much of the upcoming election will come down to the candidates themselves.

    While the Liberals poll numbers aren't great I think they are in a stronger position then they were in 2008. I think some of the former MPs who were defeated in 2008 that are running again will have a good chance of getting elected.

    For the Conservatives to reach a majority I think it will be up to strong candidates. I don't think a lot of unknown candidates will be able to beat Liberal, NDP or Bloc incumbents just because they are running under the Conservative banner.

    ReplyDelete
  31. The Liberal vote must be very inefficient to have such a low seat projection despite the 36-28 spread of this poll.

    ReplyDelete
  32. I mean this seat projection is giving us roughly the same numbers as the current parliament, which resulted from the significantly higher, 12-point spread of the 2008 election (if memory serves it was Cons 38 Libs 26).

    ReplyDelete
  33. During the Chrétien government years, I reported extensively on malfeasance by the Liberals. To do the math on the Harper government is to conclude that, while it has no sponsorship scandal on its books, it’s already surpassed its predecessor on a range of other abuse-of-power indices.

    The government’s arc of duplicity is remarkable to behold. And there are more revelations to come. It may not happen in the next election, but there will be a tipping point and the PM and his ministers will pay the price.

    Larry
    Martin

    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.