Friday, April 22, 2011

Drip, drip, drip - Conservatives and NDP make gains

Five new polls have been added to the projection and the slow progression of the New Democrats in the model continues. Polls from Angus-Reid, EKOS, Forum, Ipsos-Reid, and Nanos were added this morning.

There might be quite a few of you impatient with how slowly this projection moves, but I have to point out that the stellar NDP growth in Quebec and in Canada as a whole is quite new - just days old, even - and the projection model will take some time to buy that this is really what is going on and that it is something that will stick. But I assure you that if the NDP and Liberals are still tied at the national level on May 1, they will also be in the model. In fact, if this current rate of increase and decrease continues at the same pace, the New Democrats will pass the Liberals in national support late next week.

But if things reverse themselves - and to paraphrase Paul Wells, in Canadian politics the least interesting thing to happen is the most likely - the model's caution will be well-rewarded. And if they don't, there is still enough time for it to be reflected in the final projection.
Nevertheless, there still have been some major changes in the last 24 hours. The Conservatives remain stuck at 38.6%, but they have gained three seats and are now projected to win 150. The Liberals are down 0.5 points to 27.4% and four seats to 76, back under their standing at the fall of the government.

The New Democrats are up 0.8 points to 19% and one seat to 36, while the Bloc Québécois is down 0.2 points nationally to 8.2%. They remain at 45 seats, with the Greens unchanged at 5.7%.

The remarkable growth of the NDP can be seen in this regional breakdown. They are up about a full point in British Columbia, the Prairies, Quebec, and Atlantic Canada, an amazing amount of change in only one day. They are also up 0.5 points in Ontario and 0.3 points in Alberta.

Much of this has come at the expense of the Liberals, who are down 0.1 points in the Prairies, 0.3 points in Quebec, 0.4 points in Alberta, 0.5 points in Ontario, 0.8 points in Atlantic Canada, and 1.2 points in British Columbia. This has allowed the Conservatives to take a few seats from the Liberals, despite losing support in Ontario and remaining flat in Atlantic Canada.

In Quebec, the Bloc has dropped 0.6 points to 34.4%, while the NDP is now second in the province with 20.2%. Nothing like this has ever happened in the 30 months that I have been doing projections, so its significance cannot be understated.

The Conservatives pick-up two seats in Ontario and one in Atlantic Canada from the Liberals. The two Ontario seats are Brampton - Springdale and Kitchener - Waterloo, where Parm Gill and Peter Braid are now the respective favourites. In Atlantic Canada, incumbent Conservative Gail Shea is once again projected to win Egmont on Prince Edward Island.

The New Democrats have picked up their seat in British Columbia. Don Davies, the incumbent, is the projected winner in Vancouver Kingsway.

In this case, the Liberals suffer most from the NDP surge. But who will be the next to fall?

There are 15 seats in the projection where the NDP is within 10 points of the leader. Two of those seats are held by the Bloc Québécois, five by the Liberals, and eight by the Conservatives. So the NDP could be a bit of an equal opportunity spoiler. But in a lot of close Conservative-Liberal seats, the NDP's increase in support could turn more than a few ridings over to Stephen Harper.

UPDATE: Projection with last week of polls only

As I've explained on numerous occasions, my projection model works slowly. It is skeptical to new trends until they can be proven over many polls and many days. But what if we just take the last week's worth of polls?

The result isn't incredibly different. Nationally, with only the last week of polls, I would project 152 seats for the Conservatives, 71 for the Liberals, 44 for the Bloc Québécois, and 40 for the NDP. In Quebec, which is where everything is happening, the Bloc would win 44, the Liberals 14, the Conservatives 11, and the NDP five. Remember that several of the polls over the last week pegged the NDP at less than 20% in Quebec, and Nanos still has them behind the Bloc, with Ipsos putting the gap between the two parties at one point. So a consensus remains elusive.


  1. Thanks for doing this! It's one of my favorite election resources. Just wondering how the Ipsos poll is weighted v. say, the Nanos poll.

  2. Not much of a difference. Nanos sample was 1,200, Ipsos 1,000. Nanos a little newer (April 19-21) than Ipsos (April 18-20), and then the small differences in track record rating.

  3. This is just where the modeling is flawed. In a short, 6 week campaign, we need to be able to see how things look when a surge has become very clear. Fine that outlying polls don't affect things too much, but every major polling firm over the last week has now confirmed the surge. Gotta show us what that looks like!

    Doug Johnson Hatlem

  4. Yeah, to heck with rigour - entertain us!


  5. Thanks for another update Eric! I was feverishly clicking "refresh" waiting for it.

    0.8% growth in one day for any party is astounding. For the NDP it is breaking the mould. It feels like a major political realignment is going on in the country. A lot of the news coverage has noted the pattern in the last several elections of an NDP mid-election surge and then fall back. I think this one is somewhat different for three reasons.

    First of all it is happening among voters who have never been NDP-universe before in Quebec. It takes a lot to tip a voter over to a party they have never supported before and they would seem, therefore, to be more likely to stick once they are over.

    Secondly, the NDP high water mark in the last election was, as I recall, a Nanos poll that put them on 21%. Their high water mark now is the 25% they are showing in Forum and EKOS, a good 4% above last election's high point. They may well have crossed a psychological tipping point.

    Third, this surge is coming later, generally, than in the prior two elections, especially considering that we have the long weekend underway and it will be followed by the royal wedding. Over the long weekend people (except my politically obsessed family) won't be talking about the election, and the royal wedding will suck a lot of the available print space and air time that would otherwise go to politics. I'm not entirely sure what effect this will have, but it may mean less of a chance for the NDP's surge to be derailed in some fashion. After all, when everyone is back on topic after the wedding, there will be at most four days of campaigning left.

  6. "This has allowed the Conservatives to take a few seats from the Liberals" - shudder! I hope this meme isn't picked up in MSM. As I've just said on Twitter, it will be a real problem if this slow moving modeling of the prediction actually ends up influencing the vote by making people scared that support for NDP will wind up helping the Conservatives.

    Poll results can influence outcomes! (which is why elections Canada is smart to insist that results not be published on West Coast until they've finished voting).

    If you are in communication with Nate Silver, you should inquire with him about this problem (I think he would find it rather fascinating - going to go over to 538 and point things out in comments). Looks to me like your model is very, very similar to his. But this may just show the weakness.

  7. Thanks for doing this. Very useful. It would be interesting to see how surge of the NDP is causing some seats to fall to the Conservatives because of the vote splitting. I wonder how many seats are currently going to someone with less than 50% of the vote.

  8. Doug,

    An update has been added looking at just the most recent polls.

  9. Wow. NDP within two points of the Bloc in Jeanne—Le Ber. It's just been a huge shift, hasn't it?

  10. Eric,

    Excellent work to date and I think the statistical conservatism of your model is warranted, considering that the NDP "surge" is relatively new and should be given some time to "firm up" before automatically being considered as a definite reality. On another note - are there any other polls expected to be released today?

  11. Thanks, Eric! Interesting times for a number whiz, aggregate pollster such as yourself. What if things come in somewhere between the 5 seats for NDP in Quebec from averaging the last weeks polls and the most outlandish prediction (31 as you said on twitter!). Say, something like 14-16 seat win for the NDP in Quebec. Where would that likely have them nationally percentage and seat wise? Who would the seats come from most likely? Okay, I'll stop being so pesky now and get on with my good Friday. Much appreciated!

    Doug Johnson Hatlem

  12. @Doug:
    Pollsters releasing public polls are under incredible pressure to find something newsworthy in their numbers. This leads not only to overhyping changes that are well within the statistical margin of error, but also to the "Green jelly beans cause acne" effect (see ): if you look through enough data (change in party X support in province Y etc.) for any newsworthy change, you'll almost certainly find one, even if absolutely nothing happened in real life.

    Éric should be commended for sticking to his model. Nanos had the NDP at 13.2% on April 9th, which was clearly a freak result, but that didn't stop the media from asking Layton how he felt about the NDP heading for crushing defeat. Now Nanos is on CTV talking up the big jump to 23.7%.

    Maybe there's a big NDP surge. Maybe there's a small NDP surge. Maybe there's very little change at all, and just a coincidence of outlier polls.

  13. Doug,

    If the NDP gets up to, say, 24% nationally and in the low 30s in Quebec, they could very well win 15 or seats in the province. But they won't take them from the Conservatives, and a lot of the Liberals seats are safe in Quebec, so most of them would come from the Bloc.

    I'd expect the Tories to retain their current crop of MPs, the Liberals to be reduced by 1/3 to 1/2 in Quebec, and the Bloc to drop about a dozen seats, if not more.

    But I think an NDP breakthrough with 5-10 seats is most likely if things continue this way.

  14. Hey, Eric - Don't give in to the pressure (I know you did, a bit, by publishing results with only the last week's polls). The unique value of your site is its multi-ingredient recipe. The surge is still, at least temporaly, fairly narrow. Political (and sports, and romance movie) junkies would love to see a big come-from-behind surge that absolutely *floods* seats into the protagonists' totals.

    Too bad that's not reality. If your projection was more responsive, it would be less relevant.

    Keep up the good work!

  15. Eric, thanks for taking time working on this... I know it's a lot of hard work, but it's great to be informed!

    Like others, I'm skeptical that your model is as flexible as it needs to be. In a US-style election, it would work since election campaigns can last 1-2 years. In Canada, though, with such a fast-moving campaign, it doesn't work as well to trends.

    The NDP trend in Quebec is well established now, and Nanos- which was initially slow to respond to it as well- has started to finally pick up to pace. It will be interesting to see where Nanos is early next week... they could have the NDP over the Libs by then as well.

    The kind of slow response your model has to trends can easily result in it being way off at election time, as it could still be showing results from two weeks prior.

  16. I disagree with one of your remarks. The NDP trend of growth in Quebec is certainly established, and reflected in the model (they are up four points in the last week, eight since the start of the campaign). But the trend of the NDP challenging for first in Quebec is not established. In fact, it is only two days old.

  17. I hope Liberals will realize they made a big mistake by dumping the idea of a coalition with NDP, and pretending they could take on Steven Harper all by themselves. They should have made the coalition official since the beginning of the campaign, and had only one (Liberal or NDP) coalition candidate per riding, and then spent the 5 weeks of the campaign making the coalition case for the electorate. As most Canadians didn't mind the Lib/NDP coalition idea even without much of explaining, I am pretty sure the outcome of the election would be dramatically different than what we'll likely get now (Conservative majority).

  18. Eric did the right thing here. He's worked on this model for months, (years?). He's tested a number of factors, including incumbency and star power. It's designed to predict the outcome as accurately as possible.

    However, it is fun to see snapshot seat projections based on one poll. Just as it's fun to see the ceilings. I know they aren't realistic outcomes, but I still want to see the possibilities.

    Your audience is hungry for more, Eric. You may soon be a victim of your own success.

  19. I appreciate your analysis, Eric. You provide information with a clinical approach. I continue to feel that your models underweigh the CPC vote, particularly where committed voters are concerned. As we get closer to the election, I think your seat totals will move the Conservatives into majority territory.

    The focus of this campaign has become the race for second place. Expect the Liberals and NDP to use up the available oxygen to attack each other, leaving PMSH to cruise to victory on May 2nd. As I said before, note that there has been absolutely no change whatsoever to the Conservative's campaign theme since Day 1.

    Another website's anaylsis shows the Conservatives winning a huge majority based on the latest Ipsos Reid poll. I think their approach is biased --- because they favour voting reform --- but I still contend the CPC is well past 160 seats at this point. It could be more if the Lib - NDP votes split evenly in Ontario and BC.

    BTW, I freely admit that I'm a CPC voter.


  20. It has been barely a week (April 13) since the release of the Compas poll which was met with what could best be described as "derision". The commentary that accompanied the Compas poll indicated that there would be "surprises" for the NDP in Quebec but also noted the relative weakness of the NDP in areas of traditional strength. Half of the NDP seats are in Ontario and your model (and hence a number of polls including the latest from Nanos) still show the NDP below their 2008 vote level. The same applies for the western provinces including BC.

  21. Ipsos-Reid and Ekos both polled over the same days - April 18th to 20th. So how is it that Ipsos gives the Tories 43% while Ekos gives the Tories 34.4%? Any idea what's going on here?

  22. You have to think that there will be less political pollsters getting commissioned after this election.

    Either Ipsos Reid or EKOS will lose all credibility if the other one is accurate. If the actual result is in the middle then they both will be outside their statistical margin of error.

    The Angus Reid panels were a week ahead of the field showing the NDP being tied with the Liberals.

    Why do AR and IR do so few polls?

  23. Once again Ipsos-Reid shows a massive result for the Conservatives compared to other polling companies, and extremely weak numbers for the Liberals (21% nationally)! They also show the Conservatives far stronger in Quebec than any other pollster. I posted a comment previously about this on the 308 blog, but I still find it odd that Ipsos consistently has a very high Conservative number relative to other polls. What's up with that?

  24. We can clearly see here and will continue to see that a strengthening NDP benefits the Tories the most. When I went through all the ridings in Eric's projection yesterday, I believe that there we only two riding where the NDP could take a seat from the Conservatives, Pontiac and Nunavit, both cabinet ministers and both very unlikely. Growth for the NDP picks up seats for the Conservatives from the Liberals.

  25. Dan EKOS asks other questions on policy, direction of the country etc before the vote intention question. It also prompts. Ipsos and most others don't.

  26. Dan,
    Ipsos in every poll I've ever seen has the Tories 5-10 points higher then any other pollster. Not sure if the difference is methodology, or who is correct, but important to note that the difference is consistent.

  27. Question for Eric,
    Do you see the NDP taking enough seats from the Conservatives out west so as to make a majority even more difficult to achieve? In all the seats with NDP vs CPC races, what happens if the LPC vote collapses?

  28. Anonymous 12:02,

    Every pollster prompts except Nanos, and Ipsos only doesn't prompt for the Green Party.

    Anonymous 12:07,

    In the West, the NDP can win seats from the CPC. But in Ontario, they help the LPC lose seats to the CPC.

  29. So there's talk today of a Green party steal in Vancouver Centre. The Conservative vote may have shifted to the Green Party after their candidate dropped out, with the collapse of the Liberal vote in BC, this could be one to watch.

    I think it's probably spin but who knows.

    Anyone with a link to the real poll is most welcome to post it.

  30. Eric I didn't see one seat that was a possible Conservative NDP switch in the west except Burnaby Douglas which is already held by the NDP. Name the ridings please.

  31. Most other polsters except EKOS and CROP do not ask other questions first. It is clearly the resaon why EKOS numbers have the Conservatives low and the Greens ridiculously high, as well as the prompting.

    If they say ask a question about the environment as an election issue, it leads people to say they would vote for the green party. As far as I can tell CROPS weighting system is garbage and it is a poll within a poll skewed by the initial questions.

    The others, especially those that only ask two questions, aka "If an election were held today who would you vote for? If you are undecided who are you leaning towards?"

    That's it. There is absolutely no reason to ask other questions first, other than to skew the result. Even if a pollster says, we have ten questions we'd like you to answer, some people will hang up who wouldn't if there are only two questions, especially during hockey playoffs.

  32. Surrey North, is the only other possibility in the entire west, but since there is no indication that the Conservative vote is dissipating in BC, unlikely. Also Surrey North will already be highly organized by the Sikh community and the temples won't be changing their allegiances at the drop of a hat. They already know who they are voting for.

    In Vancouver Island North the NDP rise would have to come from the Conservatives, and there's no indication that is happening. In Esquimault Juan de Fuca, an NDP rise is more likely to give it to the Conservatives and it's held by the Liberals.

    The numbers just aren't there.

  33. "As most Canadians didn't mind the Lib/NDP coalition idea even without much of explaining, I am pretty sure the outcome of the election would be dramatically different than what we'll likely get now (Conservative majority)."

    I wouldn't be so sure about that. At least according to yesterday's Ipsos poll, Canadians prefer a Tory majority to an Liberal/NDP/Bloc ooalition by a 48%-41% margin (which would translate into a massive Tory landslide, given the 3-way split between the 41% in favour of a coalition). Moreover, Canadians were evenly split between an NDP/Liberal coalition and a Tory majority (which, again, given the likely vote split,would likely translate into a hefty Tory majority).

  34. It would seem from my perspective as if this is a really critical juncture for the Liberals to position themselves as a 'Real Alternative' lest they continue to suffer a death from a thousand cuts caught in the centre between the right-wing Conservatives and the centre-left NDP.

    I don't see it as particularly likely on a national spectrum that there are many voters who would flip from Tory to NDP or vice-versa from one election to the next- it's a bit too drastic of a ideological shift to feel organic, but significantly more plausible that the centre could collapse. With the NDP appearing to make a case that they're capable of gaining enough support in at least vote percentage across much of the country and Ignatieff not seen as a particularly charismatic leader, in spite of the earlier statements against coalition with the NDP, as a Liberal strategist, I'd begin looking much more closely at the plausibility of such a thing.

  35. Might I argue a little with your banner head prediction - "Minority Conservative Govt"?

    We need to sever the House of Commons and the Govt. The two things are only marginally connected. There are, as soon as the writ is dropped, no MPs. Buy there is still a government, and Harper is still PM (tho' he's not, like the rest of them, an MP).

    So far, so good. Now comes the election results, and we have elected MPs. There is no change in the government until the House meets. The ONLY question is: who has the confidence of the House?

    Unless one other party has the majority of the House, Harper will act as if HE has the confidence. He presents the Speech from the Throne. The opposition parties will oppose it. There will be a vote. If the government lose it, then Harper goes to the GG and says, "I've lost the confidence of the House, and I therefore resign." The GG says, "Is there anyone in the House who DOES have its confidence?". Harper is BOUND to give his best advice. If he senses that there is a working relationship between the opposition parties, and if the Liberals are the largest party after the Cons, he will say, "You should ask Mr. Ignatieff if he can form a govt." Which the GG does. Ignatieff meets the House, presents a Throne Speech, and, again, there is a vote. If Iggy wins it, then a Liberal govt (with other party support) is in operation. If Iggy LOSES the vote, HE goes to see the GG, and he may recommend Layton.

    The process will be somewhat speeded up if Libs, NDP and Bloc say, as soon as the results are known, "we will not again support a minority Con govt, given how they behaved last time."

    It may be that Layton is recommended by Harper, because Layton's figures on trust, competence, etc. are better than his and a lot better than Iggy's.

    This is not an unknown situation: it happened with the first minority Labour administration in the UK in the '20s. Labour were the third party, but McDonald became PM with Liberal support.

    So your headline might more accurately read "Conservative plurality: government uncertain."

    Keep up the good work, Eric, and kudos to your readers for their politeness!

  36. @Vanya: There are lots of voters who switch between the Tories and the NDP. Most voters have only a weak sense of ideology, if any. Many quite literally have no idea what left and right mean, or what the differences between the parties are. They vote on the state of the economy, personalities, "feelings", and scattered issues unconnected to others.

    Also, even some better informed voters switch between Tory and NDP, especially economically left, socially conservative voters, particularly working class westerners who find the Liberals alien and corrupt.

    The incredibly low (by elite standards) levels of political knowledge of average voters is not well understood by many of the political junkies at sites like this, although it is well established by political scientists who study voting behaviour. My statements about the Tory/NDP switchers are based on analysis of Canadian Election Study academic polls, that I'd be willing to share, but don't have access to over the holiday weekend.

  37. Vanya,

    I think the problem is that, given this week's narrative, the Liberals can't credibly position themselves as a real alternative to the Tories. That gambit only works when you're clearly well ahead of the other alternative party and the Liberals aren't.

    But you're right that the collapse of the Center is the big risk for the Liberals and it's a real one. The traditional Liberal tactic of telling center-left voters that they have to vote Liberal, because only the Liberals can stop the Tories, has been turned against them. And, on the right, the Tories now have the threat of the NDP (or an NDP/Liberal coalition) to peel off "blue" Liberals.

    As for forming a coalition with the NDP - I think that would be a disastrous move for the Liberals. Apart from the problem of having already ruled it out (which, if nothing else, would crippled the legitimacy of the exercise) and the problem that such a coalition would be less popular than the sum of its parts, it would be a long-term error, for two reasons. First, if you think that the NDP uprising (Iggy did ask Canadian's to rise up - this probably wasn't what he had in mind) is temporary aberration, perhaps caused by a "plague on all your houses mentality", and that Canadians don't really want to buy what the NDP is selling (for example, I doubt many English Canadians would be so keen on the parts of the NDP platform that are so popular in Quebec - the NDP is playing the old game of running different campaigns in French than in English), than the Liberals would be damned foolish to tie themselves to the NDP. On the other hand, if this the first step of a possible long-term break-through for the NDP (perhaps akin to the Tory revival in 2004), then entering into a coalition with the NDP would constitute a surrender of the center-left to them and would spell the end of the Liberal Party (although, truthfully, that's a real risk right now - these days the Liberals remind me of nothing so much as the old PC party of the late 1990s).

    At this point, the Liberals have to decide what they want to do. Do they want to try to hold the Tories to a minority? Then they keep targeting the Tories, at the risk of being wiped out by both the Tories and the NDP as center-left and center-right voters split off. Or do they try to preserve their status as the official opposition? Then they train their guns on the NDP, even if it gives the Tories a free ride during the last week of the election. Or do they try to do both, and risk succeeding at neither. Not a lot of good options.

  38. After going over numbers carefully last night, I realized that there is a substantial typographical error in this paragraph ...

    "There are 15 seats in the projection where the NDP is within 10 points of the leader. Two of those seats are held by the Bloc Québécois, five by the Liberals, and eight by the Conservatives. So the NDP could be a bit of an equal opportunity spoiler. But in a lot of close Conservative-Liberal seats, the NDP's increase in support could turn more than a few ridings over to Stephen Harper."

    Isn't it two seats held by CPC, 5 by LPC and 8 by BQ?

    Doug Johnson Hatlem

  39. This site is getting annoying to me. You have the recent polls in Quebec showing the BQ and NDP neck and neck, with a seat projection of 2 for the NDP? It seems to me your projections are falling too far behind the fast-changing reality.

    Note the prairies, where Ipsos has the PC vote at 60, and Ekos has them at 37, the most recent polls, the same day. Too small sample yes, but... WTF?

    And here in Vancouver Centre, there is still a Conservative candidate, who ran as her party's nominee for mayor a few years ago. She could sneak up the middle, so we still need to think about strategic voting here. I don't believe an internal poll showing the Greens in second here; I think the Greens will place fourth, just like last time, but I could be wrong, as could EVERYONE in this strange election.


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.

Note: Only a member of this blog may post a comment.