Thursday, December 16, 2010

Tory lead drops from five to two in new Harris-Decima poll

Harris-Decima's new poll shows national change within the margin of error, but the shifts that have taken place have narrowed the gap between the Conservatives and Liberals.Since we last heard from Harris-Decima, the Conservatives have dropped two points and now lead with only 31%. The Liberals are up one to 29%.

Certainly not the 11-point lead some other pollsters have reported over some of the same field dates.

The New Democrats are down two to 15%, while the Bloc Québécois is up two points to 11%. The Greens are up one point to 11% in this telephone poll.

The Conservatives, as usual, hold a large lead among men: 36% to the Liberals' 26% and the NDP's 12%. Something that we've also been seeing lately is that the Liberals lead among women, with 32% to the Conservatives' 26% and the NDP's 18%.

In Ontario, the Conservatives (+1) and Liberals (unchanged) are tied at 36%. The NDP is down three to 14% while the Greens are up one point to 12%.

The Bloc is up five points and leads with 44%, a very good number for them. The Liberals are up two points to 23% while the Conservatives are down two to 11%. The NDP is down one to 10%.

In British Columbia, the race is close. The Conservatives lead with 32% (-1), followed by the NDP at 24% (-4) and the Liberals (-5) and Greens (+9) at 21%. That is a very high number for the Greens.

The Liberals lead in Atlantic Canada with 42%, up nine points. The Conservatives have dropped 10 points to 35% here.

In Alberta, the Conservatives have dropped 14 points and lead with 47%. The Liberals are up seven to 24%. A 23-point gap would be a bit of a surprise in Alberta.

The Conservatives are up nine points in the Prairies to 48%, followed by the Liberals at 25%.

With this poll, the Conservatives would win 19 seats in British Columbia, 25 in Alberta, 21 in the Prairies, 47 in Ontario, three in Quebec, and nine in Atlantic Canada for a total of 125. That's one more than my projection for Harris-Decima's last poll.

The Liberals would win eight seats in British Columbia, two in Alberta, five in the Prairies, 48 in Ontario, 15 in Quebec, and 21 in Atlantic Canada for a total of 101. That's an increase of five seats over last time.

The Bloc would win 56 seats in Quebec, one more than October's projection for Harris-Decima.

The NDP would win eight seats in British Columbia, one in Alberta, two in the Prairies, 11 in Ontario, one in Quebec and two in Atlantic Canada for a total of 25, a drop of eight seats.

The Greens would win one seat in British Columbia.

And before anyone asks, you can't have the Conservatives drop 18 points in Alberta, with the Liberals gaining 13 points and the NDP one, without some consequences.

This result is closer to what EKOS has found recently. That gives us two polling narratives - are the Conservatives way in front or has the situation not changed? This poll doesn't show the Conservative gains in British Columbia that even EKOS has reported, but it does show that the Tories are performing well in Atlantic Canada.

So are we still heading to a spring election? It still does not appear that any party but the Bloc is likely to make any major gains, but an election could still be called just to change things up. There appears to be a bit of political exhaustion in Ottawa. A reset may be required, even if it changes very little.


  1. Its interesting that the polls of the last week or so are so erratic in terms of the range of support for all the parties - the Tories range from 31% to 39%, the Liberals from 24% to 29%, the NDP from 12% to 20%, the Greens from 3% to 11% and "other from 0% to 3%. The BQ in Quebec could be as low as 34% or as high as 45%

    I can't remember the last time I saw this much inconsistency.

  2. I'm starting to think that the issue is that almost every pollster, with their myriad of methodological differences and sample biases, reported over the last two weeks. When you look closely, you see that virtually all of the poll-to-poll changes for each firm were within their respective MOEs.

  3. There are two what I see as serious flaws in your seat predictions.

    Last election the Bloc got 10% of the popular vote in Quebec. It was a very efficient vote... and got them elected in 49 seats. They had 13 seats that they won in close races. There were only 6 ridings that they were a close 2nd.

    I am using 14.9% (the margin of Liberal victory in Vaughan in 2008) to indicate a close riding.

    The absolute peak for the Bloc would be to run the table and win all the ridings that they were within 14.9% of winning last time PLUS carry all the ridings that they won last time including the 13 close ones. The absolute peak for the Bloc is 55 seats.

    With their increase overall popularity the BLOC already lost a seat in a by-election in 2009 where the Cons turned a safe Bloc seat (15.4%) into a close CPC seat (5.0 %)

    With 10% of the popular vote the Bloc actually should win 45-46 seats if they go 50-50 on the close races

    the second is that even after giving credence to the poll results in Alberta you totally discount the results in BC. The Green and Liberals are both at 21%. The Liberals are to win 8 seats and the Green 1????? Really??

    What seats do you think the Green have the best chance of making up ground? Why that would be the Liberal seats where the potential Green voters had a quandary last election as May campaigned as hard for the Dion Liberals as she did for the Green. The Liberals under Ignatieff have totally abandoned the environmental movement.

    If the Green gets over 20 the Liberals stand to get wiped out in BC. In 2008 ALL 5 of the Liberal victories were close ones to the CPC. They were a close second in 3 CPC ridings and 1 to the Ndp. They were only competitive in 9 ridings with 19.3 % of the vote. With 21% of the vote and their prime vote splitting party gaining 10 % they will win 8 out of the 9 races they are competitive in??

    If the Liberals and green both get 21% of the vote they will both likely win 1 or 2 seats each. OR if you argue that the CPC is losing so much (down from 44% to 32%) then it would make sense to have the Green and Liberals winning 3-4 seats each.

    The trouble here is that you really do not believe that the Green is anywhere close to 20% and can't see them winning any more than 1 seat. Have you hard coded that Green maximum of 1 into your model?

  4. Hmm 5 national and 1 Quebec poll showed very modest gains for the CPC lately and HD shows them -2 ?

    Sure its all movement within the MOE, which normally is statistically meaningless for a single poll but when many polls are taken together actually becomes somewhat meaningful.

    HD seems out of step, no doubt about it.

  5. Since when the hell did a "close riding" become classified as a 14.9% difference? If Vaughan is the baseline, why not Winnipeg North? Then a 37-point difference can become a close riding! Makes perfect sense.

  6. Whenever I saw the headlines I did not even have to open it to know it had to be harris decima!!!

  7. Funny how when HD was showing a large CPC lead compared to other pollsters, Shadow didn't think they were "out of step". Basically, no matter the result, Shadow thinks it shows the conservatives are awesome, and the liberals are horrible. Some of us loyal readers of 308 are tiring of the uber-partisan rhetoric.

  8. You're my silent majority.

  9. Anon ALL pollsters are wrong 5% of the time.

    A few weeks ago it was EKOS that was out of step with the others.

    Now that we have 6 or 7 polls out at a time its not unusual for one of them to be wrong, statistically speaking.

    Unfortunately we've seen a pattern where by Liberal partisans, like Anon, defend the "wrong" poll because it offers them a glimmer of hope.

    We've moved from a statistically dead heat to a healthy CPC lead.

    Anyone who claims otherwise is an uber-partisan that we're getting tired of.

  10. I don't think Harris-Decima is wrong. I think they might be, for the Tories, on the lower end of their MOE, while Angus-Reid and Nanos were on the higher end of their MOEs.

    The Conservatives appear to be making some gains, but I think it is more along the lines of a 5-pt lead turning into a 6 or 7-pt lead, rather than anything more extreme.

  11. Eric looking at trend lines:

    Nanos - CPC +1
    Abacus - CPC +2
    AR - CPC +1
    Ekos = CPC +0.4
    IR - CPC +4
    Leger - CPC +1 (Quebec)
    HD - CPC -2

    Both HD and IR seem out of place.

    CPC has likely gained around 1-2 points lately.

    "5-pt lead turning into a 6 or 7-pt lead"


    Which would make HD wrong.

    HD - 2 point lead, 2.2 MOE = max CPC lead of 4.2.

    4.2 being far different than the 6 or 7 point lead you're talking about.

  12. No, I think the 2.2 can be used to increase the Conservative number to 33% and the Liberal number to 27%. Boom, six point lead.

  13. Eric do you believe that the Green are at 21% in BC?

    Even at the bottom end of the 7-8% MOE that would give the Green 13-14% That is a major major break through for the Green.

    If they are at the top end of the MOE they are looking at 29% A virtual tie with the CPC in BC OMG!!!

    If it was at all true it should be the story of the year.

    What is the reason behind the break through?

  14. Right, my mistake.

    Although the odds of that happening are far, far, far less than the 5% chance the poll is just wrong.

    So you just have a gut feeling HD is right or something ?

  15. No, I just don't see any reason to call it either way. It's just another piece of data, and the regionals are plausible.

  16. Um, Voice of Reason, the Bloc only got 10% in Quebec but turned that into 40-plus seats? You may wanna check your math on that.

    In BC, Hedy Fry's victory was nowhere near a close one.

    And there can be no presumption that a Green vote means one taken away from the Liberals.

    14.9% is not a close riding. The swing in Vaughan cannot be taken as the standard by which all ridings should be measured for their likelihood to switch.

    Are the Cons up? Yup, no doubt about it, and with their micro-targeting and GOTV efforts, I would bet on them in close races.

    You seem almost offended by Eric's projections. They are based on a model that is applied uniformly to every poll. He doesn't alter the math to make a story. In fact, if Eric has any flaw, he tries to apply the model too literally -- trying to come up with a formula to predict the effect of a star candidate being the prime example.

    You don't have to like the numbers, you don't have to like the poll. You can even suggest his methodology is flawed, but it is not fair to suggest he played with the numbers to get a result he liked.

    If you believe that he would mess with the data, maybe you should ask for your money back.

  17. In every poll, one will always see one or two "wonky" numbers esp. outside of Ontario and Quebec with their relatively larger sample sizes.

    In this instance, the 47% for the CPC in AB and the 21% for the Greens in BC are kinda obvious.

    BTW, speaking of wonky numbers, has Strategic Counsel left the field?

  18. "No, I just don't see any reason to call it either way."

    Because there's at least a 97% chance the CPC have made gains and HD is saying they lost 2 points.

  19. In BC, Hedy Fry's victory was nowhere near a close one.

    She won by 5,300 votes 9.8% collecting 34.51% over the CPC 25%. It was a 4 way race with Adriane Carr getting 18.4% for the Green.

    I would consider it close.

  20. The Bloc won 13 of 19 close races that they were in in 2008. The margin of winning these ridings would not have been reflected in the overall 10% they got nationally.

    The Bloc won 49 seats and were within 14.9% in 6 more. They were only competitive in 55 ridings.

    If the Green double their vote in BC (as the poll suggests) it is almost a certainty that Carr beats Fry and that May beats Gary Lunn.

    You either believe the poll and use it as base for the seat prediction or you don't/

    To have a model that predicts the bloc to win all these PLUS another one based on an increase of 10% in a poll .... well to me that does not pass a smell test.

  21. News of the day was probably Bill Siskay announcing his retirement effective next election.

    Now THAT was a close race, winning by 1 and a bit percent or so.

    Consider it a prime target for CPC pick up.

  22. BCVoR, You don't need a "smell test". It's just math. According to this poll, the Bloc has increased its lead by 6% over the Liberals from 2008, and they have increased their lead over the Conservatives by 16%. That's enough for them to overtake their opponent in 7 seats: Brossard, Charlesbourg, Jonquiere, Papineau, Pontiac, Portneuf and Roberval. Their 2008 total of 49 seats plus 7 new seats = 56.

  23. Let's just have a little brainstorm here ?

    Population = 33,000,000

    10% = 3.300,000

    1% = 330,000

    0.1% = 33,000

    0.01% = 3,300

    Yet the biggest pollster, Ekos, rarely exceeds 3000 polled.

    So we are to believe the results from less than 0.01% of the population represents reality? I've got some suggestions as to what we can do with the info from these pollsters.

  24. Peter, a sample size of 1,000 is statistically considered to be representative, no matter what the population is.

    I suggest we leave it to the statisticians on this one.

  25. From what we've been seeing lately I'd say if an election were held today the Conservatives would get 35%, the Liberals would get 29%, the NDP 15%, the Bloc would get 10% and the Greens would get 8%.

  26. With the variability between pollsters, and the consistency we are seeing, one starts to wonder if they are each polling a certain sub-set of voters each.

    Basically, the people each pollster covers are similar each time, thus an EKOS respondent is unlikely to be asked by AR is unlikely to be asked by HD. But each respondent is likely to be asked a second time by their own pollster. I know I've been hit a couple of times by one of them (AR), but never by another.

    As cell phones take over more and more, as the internet polling groups become more inbred (ie: not a diverse or large enough group of people to stay truly random) we might find the old polls harder and harder to take at face value. Building in bias factors, such as Eric has done, will become vital if you want to know what really is happening.

    Btw, quite the statement on our FPTP system when 21% for one party in one province gets 1 seat, but another party with identical support gets 8.

  27. Peter, a sample size of 1,000 is statistically considered to be representative, no matter what the population is.

    Eric that may be a correct statement but it leaves out one thing. It has to be totally random. As John Northey points out these pollsters aren't totally random.

    Ask the question "How many of you have been a) polled and b) polled by two or more different pollsters?"

    I'll answer for myself, neither. How about the rest on here ?

  28. Hey Goaltender Interference.... Thanks for the detail analysis but you are proving my point

    This polls tell you that the Bloc will take Jonquire-Alma where CPC Blackburn, a cabinet minister, won last time with 52% of the vote.

    That does not pass the smell test.

    In 2008 the Bloc got 38.1 % of the vote in Quebec. This poll has them at 44%. The MOE is 7-8%.

    If this poll was to test the hypothesis that the BLOC vote has changed since the 2008 election the answer would be No.

    Using the poll across ridings would have the Liberals easily maintain their Vaughan riding as the Liberals are to 36% from 2008 from 33.79 and the CPC are down to 36% from 39.18 That means the Liberal have gained 5.28% and have increased their lead in Vaughan from 15 to 20% rather than the 17.5% gain for the CPC. The Poll was off by 37.5% compared to the reality of the Vaughan election where the sample was 40,000 rather than 800.

  29. Both of you stop hurting my head.

  30. With all due respect, Shadow - BS. No matter the poll, no matter the result, your typical response is to spin it to enhance the conservative party's position. Period. Others of us do not deny the liberals are trailing, that they do not have a good chance of winning the next election, or that Ignatieff has proven to be a relatively failed experiment as leader. What we dispute is the idea that Canadians are overwhelmingly turning to Harper's fold and that the LPC is dead - two ideas you directly and subtly insert into just about every post you make here.

  31. Anon if you read above you will see where I said the latest IR poll looks out of place.

    It happens to show the largest CPC gains.

    Your entire comment is, therefore, disproven.

    Now I will ask you to please stop the attacks on other board members while we resume our thoughful, non-partisan commentary.

  32. Shadow - non-partisan community.

    Don't make me laugh.


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.