Wednesday, January 4, 2012

Tory and NDP slippage gives Liberals a polling boost in key provinces

Entering the New Year, the Conservatives continue to hold a solid lead over their rivals on the opposition benches. But 2011’s two big winners have taken a step backwards as 2012 begins, permitting the Liberals to move forward with gains in Canada’s two largest provinces. 

An aggregation of every public opinion poll released since the May 2 election, heavily weighted toward the most recent data, indicates the Conservatives have the support of 36 per cent of Canadians, down 3.6 points since the election and 2.2 points since Nov. 2, 2011.

You can read the rest of the article, detailing my latest federal vote and seat projections, at The Globe and Mail website here.

Before getting to this week's update of the NDP endorsement rankings, let's take a deeper look at the projection featured in this article.

Using the widest margin required to have given the proper seat ranges to the parties in the last Ontario election, which featured ridings of similar size to that at the federal level, I calculated what the seat ranges are for this projection as well.

Though the most likely outcome is the 140-99-63-5-1 seat distribution described in the projection, the Conservatives could win as few as 109 seats and as many as 187, based on the degree of error that the projection model has had in the past (and this includes the degree of error the polls have had as well).

Of course, winning as few as 109 seats or as many as 187 is very unlikely. The most likely outcome is 140 seats or close to it.
The New Democrats could win as few as 56 seats or as many as 128, meaning that they could potentially win more seats than the Conservatives.

The Liberals could win as few as 29 seats and as many as 88. At their current levels of support it does not appear that the Liberals could do much worse than their 34-seat performance in May 2011. They could, theoretically at least, surpass the NDP.

The Bloc Québécois seat range stands at between one and 27 seats, while the Greens could win one or two.

In addition to calculating the range, I also recorded what kind of races current levels of support would yield.

Of the 140 seats projected to be won by the Conservatives, 31 are considered close enough races that the projection could make the wrong calls. Interestingly, however, they are spread relatively evenly between the New Democrats (17) and the Liberals (14). But of the 47 ridings in which the Conservatives are trailing by a close margin, 28 of them are Liberal, 16 are NDP, and three are projected to be won by the Bloc Québécois.

This means that in a defensive campaign, the Conservatives would need to fight off both the New Democrats and the Liberals in order to save what they could. But in an offensive campaign, the Conservatives would mostly be targeting Liberal ridings or ridings in which the Liberals stand to win seats back from the Tories.

It is a very different situation for the New Democrats. In the 43 seats in which they lead by close margins, in 19 of them they are being chased by the Bloc. In another 15 they are trailed by the Conservatives and in nine the Liberals are not far behind. This means that a defensive campaign would mostly be fought in Quebec, primarily against the Bloc, but that the NDP would also need to fend off the Tories and the Liberals.

Of the 29 seats in which the NDP is trailing, however, 17 of them are projected to be won by the Conservatives. Another nine are Liberal while three are projected to vote Bloc. That means that if the NDP is moving to gain seats they need to win them mostly from the Conservatives.

For the Liberals, they are being chased in 27 ridings by the Conservatives and in only six by the NDP. In one riding the Bloc is on their tail. The Liberals trail the Conservatives in 15 ridings, the NDP in nine, and the Bloc in one. This indicates that though the Liberals need to prevent their vote from slipping to the NDP, most of their riding-by-riding battles are against the Conservatives.

And for the Bloc Québécois, they trail the NDP in 21 ridings and the Liberals in one, while leading the NDP in two ridings and trailing the Liberals and Conservatives in one apiece. Unsurprisingly, their major foes are the New Democrats.

A rough estimate based on these ranges, then, suggests that there is a 76% chance that the Conservatives would win the most seats based on these numbers, with a 24% chance that the New Democrats could win the most seats. No scenario puts the Liberals on top, and no scenario puts the Bloc back in third.

Now to the updates to the NDP endorsement rankings. This past week being a holiday week, there isn't much to report. But there have been a few minor changes.

Paul Dewar is the only candidate with a new endorsement that can be recorded in the rankings: Mable Elmore, an MLA from British Columbia. That gives Dewar an extra point, bumping him up to 26.9, or 5.7% of the total available points (a gain of 0.2 percentage points). Dewar has said he will have more endorsements to announce in the new year, so we should expect him to get another boost soon.

The endorsement of Thomas Mulcair by Herb Dickieson, former leader of the PEI New Democrats, was not added to the rankings last week but should have been. Dickieson gives Mulcair an extra 0.5 points, meaning he now stands at 119.8 points, or 25.5% (unchanged). The margin between Mulcair and Brian Topp now stands at 67.2 points.

That is partly because Topp has dropped one point as I have removed the endorsement of Joy MacPhail, former BC MLA. As a former MLA and interim leader, she was not supposed to be included in the rankings to begin with. This means Topp is now down to 187 points, or 39.9% (-0.2).

Because of the gains by Mulcair and Dewar, Peggy Nash's share of all endorsement points has dropped by 0.1% to 20.3%.

(Click here to learn more about the endorsement system and here for how the points are awarded. And, as always, you can right-click the list of endorsers and open in a new tab or window to magnify it.)
So, a relatively quiet week. But whatever momentum there is appears to be in the corners of Mulcair and Dewar. Since December 14, Mulcair has made a 2.1-point gain in the points share while Dewar has gained two points. But, as he was starting from a lower share, that is a much more significant increase for Dewar.

The press has also been much more positive for Mulcair, Dewar, and Nash than it has been for Topp. Virtually every report says he is no longer the front-runner, though that seems to be based primarily on anecdotal evidence, gut-feeling, and reports from people within the party (always to be taken with a grain of salt, since a leadership race splits a party into various camps, no matter how amicable they may be).

The next round of debates should shake things up and provide a clearer picture of the race. It might not have the drama of last night's Iowa Republican caucus, but the NDP leadership campaign should still have some surprises in store.


  1. Thanks Eric. Very interesting.

    Just using your Nov projection I see a return to minority Govt unless the Lib/NDP can form a consensus position in which case they can easily remove the Tories as governing party?

  2. "The most likely outcome is 140 seats or close to it."

    No, the most likely outcome is that the polls will continue to underestimate Tory support in this country.

    That's what happens when you're the party that raises more money than every other party combined - you have a HUGE turnout advantage that gives you a 4% boost the polls simply can't capture.

  3. It is quite clear that the Conservatives need to hold on to every last drop of support in order to stay in government in the next election. Since this government's honeymoon support is only 1% better than their election result, it would not be surprising if the Conservatives are already preparing for the campaign after the NDP chooses its leader.

    The most probable outcome would be a Conservative minority government in 2015 with the Liberals doubling their seat count because they will win back the GTA. It will be unlikely that the BQ will hold on to much of its current support due to the decline of sovereignty and Quebec will probably be an NDP-Liberal race.

  4. Four points! Why not make it 10?

  5. "Four points! Why not make it 10? "

    I suspect even his CPC blinkers won't go that far ??

  6. Eric,

    Don't forget to deduct those 10 points from the Greens. I expect the Greens to come in with negative five percent of the vote in the next election, with Elizabeth May being exiled to Antarctica.

  7. These are polls taken when people are disengaged.

    To project from now until 2015 is a fool's game. Had the current government been involved in something game changing I might take these polls seriously. In truth they have not.

    The biggest risk to the Harper government is that given a return to prosperity people will be willing to take a chance on an alternate government because they will tire of the bullying tactics employed by Harper. OTOH I still can't see people risking an NDP led government. So in my mind two things have to happen. The Liberal party has to bounce back and retake first or second party status and the Liberals and NDP need to outnumbered the Conservatives.

    2015 is so far off in political terms that anything said right now is rank speculation. One merely needs to look at the results in Iowa last night and compare them to the polls of two, three and four weeks ago.

  8. Correct me if I'm wrong here btw... but these numbers are heavily waited to most the most recent federal polls from Nanos and Harris Decima, yes?

    So relative to the last poll before the federal election, Harris Decima has CPC down 2, NDP down 2, LPC up 3, Greens up 1 and Bloc even.

    Nanos has the Conservatives down 0.6, NDP down 2.5, Liberals up 5.1, the Bloc down 1.0 and the Greens up 0.6.

    So by totally unscientifically just averaging the two we have the CPC down 1.3, NDP down 2.3, Liberals up 4.1, Bloc down 0.5 and Greens up 0.8. Or CPC 38.3, NDP 28.3, Liberals 23.2, Bloc 5.5, Greens 4.7.

    Obviously this is very much more simplistic than Eric's weighted average of far more polls, but I think it's enough to show that if there is still a systemic error over-representing the Greens and under-representing the Conservatives, it's on the order of a percent or two. I'd be interested to see though how topline numbers change if the polls are represent as only the difference between the firm's last poll before the election (at least where methodology hasn't changed). Thoughts?

  9. "Four points! Why not make it 10?"

    I'd make it six points if I was dealing with EKOS.

    That's on the high end of error though. Some were much closer.

    I wouldn't go so far as ten points. No organization in history has ever counted for that much.

    Even your readers who are most skeptical and hostile to the idea of polls under representing the Conservatives seem to have come around.

    Even if they claim its only "a percent or two".

    CPC on election day 39.9%, EKOS last poll 33.9%.

  10. I know it's a little bit late to ask, but will there be a "Projection VS. Result" for the 2011 Saskatchewan election?

  11. SkelAlex, yes. I just haven't had the time yet.

  12. @Anonymous 21:12 Yah, though it looks like EKOS has corrected that error now. Their problem was they weren't applying a likely voter screen. Once they did that they seemed to do better in the Ontario election. But yah... I def look at Angus Reid and Nanos polls alot differently than I look at EKOS or others.

  13. I *love* the seat ranges diagram, a type I haven't seen before. This is the best representation of the fuzziness of poll results I've seen - and the fuzziness makes the polling results a lot clearer, not more opaque.

  14. Thanks! I was hoping people would enjoy it. I'll do up these calculations more often.

  15. Yah, those seat ranges are fantastic :)

  16. Anonymous @21:16,

    I haven't really come around all that much ;). EKOS hasn't released a federal poll since the election, so I don't see how bringing them up is relevant here. They weren't included in this poll average. Keep in mind too that the election results were well within Nanos' margin of error too. If a pollster is 1-2 points below the actual results that is not necessarily evidence of bias towards/against a specific party's set of voters. It can just be bad luck. This is why I say "if" it exists, because at least for Nanos (and Angus Reid) it's hard to say one way or the other. Just because I bias existed in the past doesn't mean it will continue to exist either - in 2006 the polls underestimated the Liberals' support. In 2008 and 2011 it was the Conservatives who it underestimated. We won't know who it will be next time until we see who turns out in 2015.

    If we want to cherry pick pollsters though, COMPAS overstated Conservative support by 7% in the last election, so if you want to add 4 points for no reason to their polls you'd have been off by 11%. My point, and I think Eric's and Peter's point as well, is that these things should be approached from a methodical and empirical standpoint rather than by a gut check. Not that I don't do the same thing myself here... lol

    What I'm really digging on this new seat range format though is it keeps an "open mind" as to what error the polls may have and builds that in to the uncertainty. I think that's really really valuable.

  17. Ryan it is very easy to approach polling error in an objective manner. I believe Eric's run calculations before showing pollsters slightly over estimated the CPC in '04, were just about right for the CPC in '06, under estimated them in '08, and this time really under estimated them.

    That is a one directional trend over multiple cycles.

    Pretty hard to ignore such a pattern or to argue it'll magically reverse itself when we know what is behind it - MONEY.

    In '04 the CPC had no machine but the Liberals still did. By '06 it was an even match. By '08 they had pulled ahead and by '11 they had a HUGE advantage.

    Let's keep watching the financial returns of the parties every year OK ?

    If the Tories continue to outraise all other parties combined then its clear they are going to continue to have this advantage in elections that the pollsters just can't take into account.

    (Unless as you say they start making guesses about turnout with voter screens.)


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.