Friday, December 10, 2010

Liberal growth in new EKOS poll

The new EKOS poll shows that the Conservatives have remained stable over the last two weeks, but also that the Liberals have made some gains at the expense of the New Democrats.Compared to EKOS's last public poll released two weeks ago, the Conservatives are up only 0.4 points to 33.7%. They did have a dip in EKOS's polling last week, when they were at 32.6%, but this is all oscillation within the MOE.

The Liberals have gained 2.1 points from two weeks ago, and now trail with 29.2%. The gain has come primarily in the last week of polling.

The New Democrats appear to have been sinking throughout the two weeks, and are down 2.2 points to 14.4%.

The Greens are up 0.9 points to 10.4%, while the Bloc Québécois is up 0.3 points to 9.8%.

EKOS uses an automated telephone polling system that manages to include both landlines and cell phones. The percentage of respondents who were undecided was 16%, virtually unchanged from two weeks ago.

The race remains close in Ontario, where the Conservatives have dropped one point but still lead with 35%. The Liberals are down two points to 34.4%, while the New Democrats are steady at 15.9%. The Greens are up three to 12.2%. The Tories lead in both Toronto and Ottawa, with 37.9% and 48%, respectively. The Conservative gain in Toronto may stem from, or be a record of, their by-election win in Vaughan. The Liberals trail (but are statistically tied) in those two cities with 36.9% and 42.6%.

The Bloc has gained a point in Quebec and leads with 38.4%, followed by the Liberals at 22.6% (down one). The Conservatives have gained three points and stand at 16.8%, while the NDP is down two to 11%. The Bloc leads in Montreal with 42%. The Liberals are at 21.8% there.

In British Columbia, the Conservatives have finally pulled away from the pack. They've gained six points over the last two weeks and lead with 42%. The Liberals trail with 25.3%, up 10 points, while the NDP is at 21.8%, down seven. The Greens are down five to 8.5%. However, note that in the first week of polling the NDP was leading the Conservatives 35% to 26%, so there is still a lot of polling volatility in the province. In Vancouver, the Conservatives lead with 45%. The NDP is second with 21.4%.

The Liberals score a huge number in Atlantic Canada, at 58.2%. That's a gain of 23 points, and obviously one that is a result of the small sample size and the 95% level of confidence. The Conservatives are second with 20%, down 19 points. The first week of polling had a more typical level of support for the Liberals (36%) and Conservatives (31%), but EKOS did still find the NDP well below 20% in both weeks.

In Alberta, the Conservatives lead with 62.7%, followed by the Liberals at 17.8%. The NDP has dropped five points to 6.1%. The race is surprisingly interesting in Calgary, where the Conservatives have 56.6% support to the Liberals' 29.2%.

Finally, the Conservatives have gained 13 points and lead in the Prairies with 47.5%, ahead of the New Democrats at 18.9% (down seven) and the Liberals at 15.3% (down 10).

This poll would result in 23 Conservative seats in British Columbia, 27 in Alberta, 23 in the Prairies, 46 in Ontario, seven in Quebec, and four in Atlantic Canada for a total of 131. That is only one more than the projection result of EKOS's last poll two weeks ago.

The Liberals would win nine seats in British Columbia, one in Alberta, one in the Prairies, 45 in Ontario, 15 in Quebec, and 27 in Atlantic Canada for a total of 100. That is eight more than two weeks ago, but the gain is primarily in the improbable Atlantic Canada result.

The Bloc would win 52 seats in Quebec, down one.

The NDP would win four seats in British Columbia, none in Alberta, four in the Prairies, 15 in Ontario, one in Quebec, and one in Atlantic Canada for a total of 25, down eight from two weeks ago. The drop took place mostly in British Columbia.

This poll puts a bit of a damper on the 7 to 11 point leads we've seen from other pollsters like Nanos and Angus-Reid. It also differs in that the Conservative strength is out West, whereas some of the other polls have shown surprisingly good Tory numbers out East.

Will this put out the flames of the rumours that we'll be heading into an election in January? Hard to say - each of the parties conduct their own polling and they probably pay more attention to those (unpublished) results.

Nevertheless, we've been blessed with an abundance of federal polling this week and it appears that no two polls tell the same story. That is a bit of a change from the last few months, where every showed almost the exact same thing. While this could just be methodologies and margins-of-error conflicting with one another, it could also indicate that Canadians are starting to be less sure in their voting intentions.


  1. In several of your analyses, you have predicted one Liberal seat in Alberta. Is this based just on their overall percentage in the province? Or do you have an idea of what riding they are competitive in? I live in Edmonton, in one of the two former recent Liberal ridings. I can tell you that I see no chance of a Liberal win.

    I would consider it far more likely that the NDP would hold their riding of Edmonton-Strathcona, than a Liberal win anywhere in Alberta. Yet you show the NDP with no Alberta seats.

    Can you explain?

  2. Yes, it is based on their overall percentage in the province. Based on historical performances, this level of support should be enough to get them one seat.

    The most likely seat would be Edmonton Centre.

    The NDP, at only 6% (or roughly half of their 2008 election result), are not in a position to hold on to Edmonton-Strathcona.

    The model currently does not make projections for individual ridings, but will in the future. At that point I will be in a better position to break down the situation in Alberta. It could very well be that once I can make more precise projections I won't be projecting a Liberal seat win in Alberta.

  3. the poll had the CPC ahead of the Liberals in the GTA. How many seats did your model have the CPC pick up?

  4. Enough to make-up for what would have to be Liberal gains outside of Toronto, considering that such a close race provincially and with the Conservatives leading in Toronto, the Liberals must be doing very well outside of the city.

  5. What do you make of the absurd 17% figure for the Green party in Man/Sask. just days after two byelections in Manitoba where they got 5% in one and less than 1% in the other. I think its about time that one of the polling companies took the lead and stopped prompting for the Green Party and Ekos in particular should stop this nonsensical prompting of "other"

  6. The margin of error in Saskatchewan/Manitoba is more than 8 points, so really nothing odd about it.

  7. ERic,

    You're more familiar with the polling methodology that I am, can you think of any reason why the the top-line numbers in the Ekos polls (i.e., 58% for the Liberals in the Atlantic Provinces) doesn't match their component numbers (i.e., 42% for the Liberals amongst Men and Woman in the Atlantic provinces). It's most obvious in the Atlantic, but you see similar results in a number of provinces (and for all parties). I'm actually surprised that there haven't been any comments to that effect because it's such a weird result. On the otherhand, EKOS is a first rate polling firm, so I can't see them making a typo like that.

    The only possibility I can think of is that the underlying numbers are not weighted but the top-line numbers are (although, if so, I would have thought that EKOS would have disclosure to that effect). Still, if that's the case, what does it say about EKOS' underlying survey if re-weighting shifts voter support by 33% (i.e., from 42 to 58%).

  8. Carl,

    EKOS has corrected that, if you look at the PDF now. It appears they have some sort of recording error. It isn't the first time this has happened in Atlantic Canada.

    They are, in fact, at 58% in Atlantic Canada. The demographic breakdowns were incorrectly recorded.

  9. Or weighted, I should say.

  10. The margin of error in Saskatchewan/Manitoba is more than 8 points, so really nothing odd about it

    using the full margin of error that would put the Green above 9% in Manitoba and Sask.

    And in 2 out of 28 ridings they drew just got an average of less than 3%

    AND you see nothing out of the ordinary with that?

    The concept of polling is to make the best prediction of what would happen in an election. There is no indication that the Greens will increase their vote %.

    Since sept 9,2008 in 128 Ekos has had the Green a 8% once April 13,2009 and below 10% 9 times. That is 119 polls that Ekos has the Green Party at more than 10%. The average is 10.8%. The range of Green support is 8% to 13.4 %. That is an incredible small range for 128 polls with 3.5% MOE. It would suggest that according to EKOS Green support never changes.

    This covers an election with the greens getting 6.8 %.

    At some point people have to call shenanigans on Graves and Ekos as their polling has no basis in reality.

    Does the taxpayer/CBC funded frequency of the EKOS polls give them extra weighting on say Angus Reid in your seat forecast model?

    With this obvious flaw systematics flaw in the EKOS polls are you considering dropping them from your seat prediction model?
    Or is GIGO okay with you?

  11. Eric,

    Thanks for that.

    However, I note that they've only corrected some of the fields. For example, in the Atlantic, the aggregate Green vote is higher than either the Green vote amongst men or women. Similarly, in Ontario, the Liberal vote amongst men and woman is 34.6%(for each), while the aggregate is 34.4).

    And a number of the total numbers still don't look like averages of the subsets. I don't understand, for example, how the Greens can be at 8.5% in B.C, when they're at 8.2% amongst BC men, and 15.8% amongst BC woman. Do woman only make up 5% of BC's population?

    And those are just the easy ones when you look at men and woman, a lot of the age and education demographic numbers look awfully suspicious. I think EKOS may want to re-run their numbers until they're all right. I

  12. Interestingly, if you look back at the November 25 EKOS poll you see some of the same problems. For example, in that poll, the top line number had the NDP at 29% in BC, even though their support amongst men and woman was 17.7 and 24.7% respectively (conversely, the underlying data for the Greens and the Liberals suggest that their top-line number should have been higher). It also seems to have understated Tory (and Green) numbers in Sask/Man (relative to the underlying data) while overstaing the Liberal numbers there.

    You don't see this pattern in the November 11 poll (where all the top-line numbers seem to be mathmatical averages of the male and female numbers - to use the simplest subset to compare with - as you might expect), but you do see it in the earlier polls in October.

    Considering EKOS' tendency to pontificate on the significance of gaps in the underlying data (the "education" gap, the "gender" gap), they may want to make sure they get that number right. And I sure home that it's the bottom numbers that are wrong, rather than the top-line number.

  13. New IR poll out confirms CPC lead:

  14. Star Candidate:

  15. Earl

    That poll is nuts. Tories go up by 4

    Liberals static

    NDP down by 4.

    If you think NDP supporters will shift their vote to the Tories I've got some Florida beachfront property you might be interested in !!!!!!!

  16. OT:

    The problems with Ontario's "Green" program.

  17. Peter you should know better! Just because there is a four point shift in the CPC and NDP vote doesn't mean NDP voters are going to the CPC from the NDP. That's what is known as a statistical anomaly.

    IR is also roughly in line with AR and Abacus in terms of the CPC lead. It's also in line with Nanos and AR as regards the level of CPC support. The only poll out of line this week is EKOS.



    BTW do you really own swampland in FL?

  18. Peter: Hold the deposit on the real estate. The NDP didn't move to the CPC.

    The Liberals moved to the CPC and the NDP moved to The Liberals. The natural migration to the Right. The better the CPC manages the more NDP move to the Martin/Ignatieff Blue Dog Liberals.

    In Vaughan The Liberals basically got ALL the NDP vote , but 20% of the Liberal vote went CPC.

  19. Peter said: "If you think NDP supporters will shift their vote to the Tories I've got some Florida beachfront property you might be interested in !!!!!!!"

    Peter, Peter, Peter. What are we going to do with you? First, did you consider the possibility that maybe the Tories picked up voters from the Liberals, but the Liberals picked up votes from the NDP?

    Second, why do you think it's implausible that the Tories could pick up votes from the NDP. In large swaths of the country (i.e., Manitoba, Saskatchewan and BC) both the Tories and the NDP appeal to the same core of populist voters (that's why those provinces are both the core base of the Tories federally, and have the strongest NDP parties provincially). And you see that trend in other recent polls in Atlantic Canada (which have shown the Tories making gains at the expense of the NDP). In my own neck of the woods, its no coincidence that Ed Broadbent's old riding now goes solidly conservative.

    I realize that in high school civics texts they show the political spectrum as being linear, but reality is a bit more complicated than that.

  20. Peter said: "If you think NDP supporters will shift their vote to the Tories I've got some Florida beachfront property you might be interested in !!!!!!!"

    Not so fast. Outside the City of Vancouver (and its closely neighbouring 'burbs), it's a battle between the CPC and the NDP in BC.

    The Liberals have recovered their traditional urban BC core vote, but the NDP now also seems to be collapsing in suburban/rural BC due to the very public nature of its current internecine warfare.

    The provincial BC NDP is bleeding badly right now as a result.

    That provincial NDP blood-fest also now seems to have spilled over into BC federal voting intentions with the NDP at 18% in the latest Ipsos poll.

    In most of BC, it's a CPC/NDP battle and that battle essentially involves populist politics (read Reform in the 1990's).

  21. Peter you should know that 11% of NDP supporters prefered election outcome is a CPC majority.

    (The highest number of all opposition parties, Liberals were at 4%.)

    In many ways its MORE likely to see NDP support going to the Tories than it coming from the Liberals in many regions.


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.