Tuesday, April 19, 2011

Liberals and NDP continue to make gains

Five new national polls, from Nanos Research, Abacus Data, Environics, Harris-Decima, and EKOS, were added to the projection this morning. Three riding polls conducted by CROP were also added to the projection. Remarkably, all of this new data changed little in the national vote projection, but it did wreak some havoc at the regional level.
Nationally, both the Conservatives and Liberals are unchanged at 38.7% and 28%, respectively. However, the Conservatives have dropped two seats to 147, while the Liberals have gained two seats and now stand at 80. This is both a low and high watermark for the two parties in this campaign. Overall, and considering that two of the vacancies at the dissolution of the House of Commons were safe Conservative seats, this is still a gain of two seats for the Conservatives over their standing when the government fell. For the Liberals, it is a gain of three.

The New Democrats are up 0.2 points to 17.8% and one seat to 35, still a net loss of one seat but a high watermark for them as well. The Bloc Québécois, meanwhile, is unchanged at 8.6% nationally but has dropped two seats to 45, their lowest point of the campaign. As the third vacancy was a Bloc-held seat, this means the Bloc is actually down three seats compared to where they were before the election.

The Greens are steady at 5.8%, while an independent is now projected to be elected.

Aside from a small uptick in British Columbia, the Conservatives are down in every part of the country. But this should not be overstated, as all losses (with the exception of Alberta, where little is at stake for the party) were of 0.2 or fewer points.

But another drop in Ontario is problematic, as it coincided with a 0.1 point gain for the Liberals and New Democrats. This has caused a couple seats to flip.

This is not an entirely good update for the Liberals, despite the seat gains. They dropped 0.3 points in British Columbia and the Prairies, and they are now third behind the NDP in Saskatchewan and Manitoba. They also dropped 0.4 points in Atlantic Canada. The small gain in Ontario is positive, of course, and with the Bloc and Tories dropping in Quebec a steady number is not bad news for Michael Ignatieff.

For Jack Layton, however, things are looking up. Aside from a loss in Alberta (which does not put Linda Duncan in danger, yet) the NDP is up everywhere, with decent 0.3 to 0.4 point gains in British Columbia, Quebec, and Atlantic Canada. Though the party needs to gain a lot more in the two eastern regions to pick-up more seats, the gain in British Columbia could be reflected in seat gains soon - if it continues.

Things continue to look bad for the Bloc Québécois. They are down another 0.4 points while the Liberals and NDP are gaining or holding firm. The consolation of the Tories dropping might not be enough, as the party isn't poised to take any Conservative seats just yet. Even today's CROP poll for Portneuf - Jacques-Cartier put the Bloc behind the independent incumbent André Arthur.

And in some degree because of that poll, the riding is now projected to be won by Mr. Arthur rather than the Bloc. The other seat change in Quebec is to the benefit of the Liberals at the expense, again, of the Bloc. Nancy Charest, who ran a good campaign in 2008, is now the projected winner in Haute-Gaspésie - La Mitis - Matane - Matapédia. That is a gain, as it was held by the Bloc prior to this election. If this seat is won by the Liberals on May 2, it would be quite a coup considering it is a francophone rural riding well outside of Montreal (it's almost as far away as you can get from Montreal in Quebec!).

Two seats in Ontario have also changed hands. Andrew Telegdi, the former Liberal MP in Kitchener - Waterloo, is now projected to win over the Conservative incumbent Peter Braid. In the north, the NDP incumbent Tony Martin is once again the projected winner in Sault Ste. Marie, meaning the Tories have lost two seats in the province in the projection.

If things continue along this path, the Conservatives and Bloc - the two parties everyone expected would come out of the election stronger than they entered it - will be the net losers of the 2011 election.

We are now less than two weeks from election day. While there is still plenty of time for both Stephen Harper and Gilles Duceppe to turn their ships around, there may also be enough time for Michael Ignatieff and Jack Layton to make some real inroads and threaten the Conservatives' minority.


  1. The problem with all of this very close analysis is that it fails to absorb the critically important mood of this election. The Liberals are despondent and the NDP are euphoric while the Tories are serenely confident. These qualities will matter far more than fractional movements of in the polling data come Election Day. They have everything to do with whether people will get out of bed and actually go to the polling booth.

  2. That is a different issue - I report on the polls and what they say. And, in the end, they are a reflection of the mood.

  3. When you make up your projections are they based on the numbers of decided voters or are you also accounting for undecided voters? I think that the pollsters should include the undecided voters in their calculations because it gives a true picture of what is happening. When they base their numbers on decided voters that is obscuring the message.

  4. ...And also, Dio is putting anecdote against measurable data. I have sufficient faith that with high frequency of polls, any 'mood' of the electorate, if it really exists beyond the most recent news article read by the commentator, will be reflected by voting intentions over the following day or two.
    I think, (another anecdote, lol) that Ignatieff was charicaturised by attack ads for a long time, but those perceptions are shifting slowly. This weekend, a whole lot of people wil be arguing politics around the Easter dinner table, and Ignatieff now means a little more to the electorate than he did 3 weeks ago. I am not a Liberal, but I think the Liberals have run a good disciplined campaign, and between tweedle Dee, (Lib) and Tweedle Dum, (CPC) it is the Tories who have everything to lose this weekend. Their message does not bear close discussion, as it consists of infantile assertions, while the Liberals have all policy content on their side. But my anecdoteis worthless against the data, so we will see by next Tuesday who will win the dinner plate war this weekend.

  5. Eric has done a great job on the polling data, but this week the polling data seems to be disconnected from what is happening on the ground. I'll guess that the Conservatives are actually ahead a few seats more that what the public polls are giving them.

  6. Hi, I am a bit confused at all this "hype" around the NDP. I get the polls have been favourable to them lately but all this talk about the NDP replacing the Liberals as opposition, etc. seems to be EXTREMELY far-fetched. I would think the chances of the Liberals winning a minority government are greater than the NDP being the opposition. I think a lot of these voters saying NDP now may not necessarily vote NDP on May 2nd (I think there was a poll discussed on this blog to that effect).

    I know multiple friends that when asked who they want to vote for, they say Jack Layton because they like him but then follow up that with a "but I am going to be Liberal because there is no way I want an NDP government or to split the vote". I think the NDP support is VERY soft and this talk of an NDP opposition is a pipe dream. There are about 50 ridings that I can see are pretty much guaranteed to go Liberal.

    Also, on Andrew Telgadi, I think he pulls it out because 17 votes in nothing and I think you will have at least some of those thousands of people that voted Green or NDP show support to Andrew. Same could be true for the KW Centre riding with Karen Redman, she lost by 400 votes I think last time. For some reason I think all of the Cons talk of a majority may actually prevent them from getting a majority as people may take that possibility into consideration in voting with strategy. I think the Cons were better off not having that sound bite by Harper asking for a Majority government. The Canadian public is not "right" and the best hope I thought for the Cons was enough splitting of the vote to win ridings (like look at Nunavut - they could win with 35% of the vote). When Harper says he wants a majority government you may get people in ridings like KW that say "you know what I'm voting for Andrew even though I like Layton".

  7. Great work Eric. I make sure to read the site every day. Just wanted to let you know that you have a couple issues with the links at the start of the post. The one for Abacus goes to the Nanos poll, and the one for EKOS goes to a story from cyberpresse.ca about the CROP riding polls.

  8. Eric

    I think I understand your methodology but can not figure out the factors that would take the Yukon seat of Larry Bagnall that he won in 2008 by 45 -33 to your prediction of 53 - 25.

    His margin of victory in 2006 of 48-24 was drastically reduced in 2008. Ie the long term trend seems to be the CPC chasing the Liberals down in this riding.

    Mr. Bagnall almost cried in parliament when he voted to keep the LGR. I think that he felt his time as a MP was up.

    Do the polls in BC outside of the lower mainland show a huge Liberal surge? CPC decline? What other factors in your model would drive Mr. Bagnall to a landslide win?

  9. I agree the polls are a reflection of the mood. So far, Harper and the CPC are hanging in there with a majority within reach but not guaranteed. That is the overwhelming message of the polls. Other than that it is small noise, as to what it means to Canada, with a few seats shifting around between the 3 opposition parties.

    I can't share the NDP's alleged euphoria while Harper is still sitting at the highest seat count by far. I am biased and would like to see one or more opposition parties really stripping seats away from the CPC. That's not happening, but I'm finding the analysis of this site very interesting (if not encouraging) anyway.

  10. @Dio

    "Mood" is notoriously hard to predict - both in predicting what the mood is, and predicting the effect that mood will have when it comes to getting out the vote. There really isn't any valid way to work it into a statistical model.

    Sure, there are certain trends that tend to hold from one election to the next - but I've yet to see any accurate predictions of election day turnout modifiers. There have been plenty of those made across the last few elections, and none of them really panned out. They either were flat out wrong, or didn't happen to near the extent that was predicted.

  11. Anonymous 10:02,

    Most pollsters combine decided and leaning voters, so my model takes both into account.


    Thanks, that's clearly a mistake. I'll correct it.


    There is a brief explanation of methodology for the North here: http://threehundredeight.blogspot.com/2011/03/methodology-of-projection-model-and.html

  12. This projection is hard to believe Eric. The Conservatives are still about 5% higher than polls predicted before the last vote.

    I'm sure the Conservatives are quite happy to be between 39-40% in the polls, since they haven't changed their strategy one bit since the onset of the campaign, we can assume they are happy with what the polls are telling them.

  13. Eric based on this latest data and the fact the early polls are just three days away can we say with confidence that a CPC majority is not going to happen ?

  14. I suppose the slow drip of the projection should elate me, but I'm inclined to be more frightened of the couple of polls showing the NDp surging at the Liberal's expense, putting a majority in play. (That and the fact that a major shift in the projection like +-3% is certainly possible depending on turnout on election day itself)

    I don't think if polls continue to show a 26-22/25-20 LPC/NDP split that we'll be seeing that CPC seat number stay where it is.

    Although on that note; there is significant seperation between some polls. NANOS and HD show little of what EKOS/ANGUS/FORUM and ENVIRONICS are showing.

  15. DIO has hit upon the important, but unknowable, question of voter turnout.

    Tight now it is very on the edge number wise, and a shift of even a very few points can make a huge difference.

    The latest poll from the CBC and EKOS, shows a 12.5% lead for the Tories

    ¤ 37.4% CPC
    ¤ 24.9% LPC
    ¤ 20.0% NDP
    ¤ 8.4% Green
    ¤ 7.8% BQ
    ¤ 1.5% other

    Using s simple regional proportional swing model (less sophisticated than Eric's but "in the ballpark") this gives a seat break out of

    Projected Seats
    Tories Grits Dippers Bloq
    11 16 5 0
    9 8 9 49
    55 38 13 0
    22 2 4 0
    27 0 1 0
    19 5 12 0
    1 1 1 0
    144 70 45 49

    The Dippers climb a bit, but fundamentally similar to the status quo.

    By contrast if we make (not unreasonable) assumption that the Tories and Dippers are motivated (add 2% to their totals) and the Grits are depressed watching Iggy crash and burn (deduct 2%) we end up with the following seat distribution:

    Tories Grits Dippers Bloq
    12 14 6 0
    10 6 14 45
    63 28 15 0
    23 1 4 0
    27 0 1 0
    21 2 13 0
    1 1 1 0
    157 52 54 45

    A Conservative majority and Jack Layton as leader of the official opposition... a very different kettle of fish.....

    We will see what (if anything) the pollsters start to do with the enthusiasm gap where the Tories and Dippers feel good and the Grits feel hopeless....

    Glad I don't have to sort this one out! - Good luck Eric!

    The Vorlon


  16. "..threaten the Conservatives' minority"?

    Granted anything is possible but how does this become anything other than a long shot?

    You may well be right about Telegdi winning in Kitchener but this would have to be a slam dunk for the Libs to have any chance of winning the election.

    There would have to be a red wave through Southern Ontario to make this a horse race and I do not see it in any of the numbers.

    Don't the current numbers suggest a status quo parliament +or- 5 seats coast to coast?

    I understand that "mood" is irrelevent but I would suggest that where the leaders are campaigning (not on any given day but in a general sense) gives us some clue as to how the parties believe their fortunes are moving.

    For example, if Ignatieff were pushing in Calgary for three days and Harper was in SK for two days it would imply that the Libs are on a roll.

  17. BTW, I think the Conservatives are keeping their powder dry. I think they are very happy to sit just below a majority so no scare tactics can gain traction, then in the final week turn it on.

    I'm only online, so I don't know about tv, but I've only seen Liberal ads. I think the Conservative ad assault in the final week will be epic, especially in the regions they need.

  18. Anonymous 11:34,

    I'm referring to the possibility that if the Liberals and NDP win enough seats, the Conservatives won't be able to govern as a minority.

    I don't mean to say that the Liberals will win more seats than the Conservatives.

  19. Eric

    from your methodology explanation:

    "For Northern Canada, I have applied the proportional swing of the national results and individual candidate factors. I have also looked at the historical record and applied some swing-corrections only when they appear to be have been a historical tendency."

    So 2008 national results CPC 37.6 Lib 26.2

    Eric's weighted national polls April 19 CPC 38.7 liberal 28

    CPC up 1.1 Liberal up 1.8

    based on this the and the trend of CPC reducing the margin from 24 % point to 12% points from 2006 to 2008 generates a 28 % point margin for Bagnall in 2011?

    What candidate factors would you have applied to get this prediction? Supporting the Long Gun registry actually really popular in the North? Did you attribute the CPC gains in the last election on the Dion Carbon tax and think that the Ignatieff cap and trade would be more acceptable?

    This does not pass a reasonability (smell) test. The raw polling numbers do not appear to drive this prediction.

  20. Michael Ignatieff is campaigning in the GTA today. I think that tells us all we need to know about seat projections.

    This is akin to Stephen Harper campaigning in Calgary for a day, which we don't see happening.

    I think internal party polls are telling us where the election stands right now. The Liberals are not just sending Ignatieff to Toronto but Justin Trudeau. If they are sending the big guns to Toronto, they are in trouble.

  21. Vorlon, the enthusiasm gap you describe is only one part of your equation. There is also the delusion that most NDP voters can be enthusiastic with a Harper majority. A minority of NDP voters, sure. But not the majority.

  22. the polls on April 26 will be very interesting.

    After the Easter weekend when the family sits down and Mom and Dad explain what is going on there will be a how taxation and social programs work and how a vote for the coalition actually impacts the family's finances.

    After the 2008 thanksgiving weekend Nanos rolling poll jumped from 32-28 to 34 - 27 and still under estimated the CPC by 3.5%

    Family time helps the CPC.

    the Friday before the vote the Nanos poll was over 5% low on what the CPC got on the Tuesday.

    If Nanos is a polling using the same methodology as in 2008 the actual CPC vote will be in the 44-45 pt range.

  23. Here's a "mood" that I've perceived quite regularly from those who typically vote Lib or NDP: pissed off (at the CPC).

    This, I think, might compel those in tight ridings in the recent past to vote for the candidate who has the best chance of unseating the Conservative incumbent.

    Great site though, Éric. I've followed it regularly since the beginning of the election. I certainly don't pretend to be an expert on such things, and so can't speak with too much authority, but the disparity in the polls suggests that the only poll that really matters is the one on May 2nd; how do we know which poll released today or tomorrow is closest?

  24. I think the true NDP would be ecstatic with a CPC majority and NDP tied with The Liberals and Bloc as official opposition.

    As great a campaign as Jack has had he is 61 years old the NDP need the 5 years to organize a leadership succession plan. If they run a better more reasonable leadership search and campaign than the Liberals... not hard to imagine... The NDP become the natural opposition in 2016.

    The NDP have a chance of having a leadership team while the Liberals will continue to destroy each other in a leadership Campaign.

  25. "Michael Ignatieff is campaigning in the GTA today. I think that tells us all we need to know about seat projections.

    This is akin to Stephen Harper campaigning in Calgary for a day, which we don't see happening."

    If he was going to ridings like York West & Toronto Centre your comparison would be valid. But he's not, so it isn't

  26. @Anon (12:00)

    Liberals in the GTA have nowhere near the position of the Conservatives in Calgary. While the Liberals typically do very well in the GTA, it is nowhere near the dominance the Conservatives see in Calgary (with a full sweep of the city, and over 50% of the vote in each riding). There are some definite gains that could be made in the GTA. Oak Ridges-Markham was won by the Conservatives by 0.7% of the vote last time around. Same with Mississauga—Erindale.

    While they would not be gains, both Brampton ridings were also very close last time and the Liberals want to make sure they hold on to them.

    There's also another handful of ridings that were about 5% in popular vote between 1st and 2nd in 2008 that the Liberals could consider up for grab (or in need of defense) this time.

  27. Thanks Eric. Have followed your site for over a year. This is my first post. In line with Ignatieff in Toronto, I think the race in Parkdale-High Park is a lot closer than your model suggests. NDP'er Nash (MP 2006-08), has maintained a high profile in the riding since the last election, whereas Liberal Kennedy has been a lackluster representative; plus having the 6th worst voting record in Ottawa. On the ground, Nash signs outnumber Kennedy's. She is also actively supported by the two municipal councilors that cover the riding, and by the local popular MPP Cheri DiNovo, who took the provincial seat from the Liberals when Kennedy resigned. Kennedy may have been a star years ago, but the shine appears to be fading. How much more can he milk establishing the food bank in the '80's, his only real credential as a progressive.

    South Parkdale Jack.

  28. Eric, how do you assign voter preference for the Yukon, NWT and Nunavut? Thanks

  29. "still under estimated the CPC by 3.5%"

    BECAUSE OF TURNOUT-nothing else. Can we end these stupid myths of "ballot box bonus", "shy Tories" once & for all? What is called "voter enthusiasm" should be called "voter habit". Those of higher income & older people vote in higher numbers than those of lower income & younger people. That's a better indicator of turnout than if you're gaga of lukewarm for a particular leader

  30. Eric,

    To what extent is the "gain" for the Liberals driven by the reduced weight (due to "aging") given to older polls (say, from the begginning of the campaign) which showed the Tories well ahead? I ask because the changes you describe in the post don't appear to be consistent with the changes reported in yesterday's batch of polls.

    I suppose we can figure that out if you do poll specific projections today.

  31. Dio some polls are reflecting the CPC strength, the LPC panic, and the NDP gains.

    But a lot of the other polls don't show these trends. They don't reflect that "mood" you were talking about.

    Eric assumes the truth is in the middle.

    But if one set is right and one set is wrong then yes, the CPC could well be on its way to a majority.

    Peter shouldn't be trying to rule out a CPC majority at this point. Still way too early. Anything could happen.

  32. Most people think that the Liberals lost ground in the final weekend in 2008 because of Stephane Dion had that disastrous interview with CTV (and Mike Duffy got rewarded with a seat in the Senate as thanks for his good work on that). The fact that it was Thanksgiving was irrelevant.

  33. Is it wise to be making these sorts of changes to your seat projections and talking about "big gains" (yesterday's post) when the changes in support in the polls were all less than a percentage point (and in some cases as low as 0.1 or 0.2 points)? That's well within the margins of error for these polls, especially considering the higher margins for regional results.

    Statistically speaking, you can only say one thing about a change of 0.5 percentage points: support remained unchanged.

  34. My projections are an aggregation of all polls, so even small changes are important. The aggregation does not have a statistical margin of error because it isn't a poll, and the inclusion of all public polls smooths out the MOE of each of the individual polls.

    And in the end, I'm reporting on how the projection has changed. With the latest numbers, a few seats change hands.

  35. Anonymous said@ 12.07

    Vorlon, the enthusiasm gap you describe is only one part of your equation. There is also the delusion that most NDP voters can be enthusiastic with a Harper majority. A minority of NDP voters, sure. But not the majority.


    And that, for many Dippers is INDEED the question..

    For the NDP to "possibly" get to be the official opposition and supplant the Liberals as Canada's party of the left, the price "might" have to be a Liberal collapse that leads to a Conservative Majority...

    Becoming #2 is a long held NDP dream, the Holy Grail..... The Conservative Majority is the NDP nightmare..

    What if the two go hand in hand? - Some Dippers have a lot of soul searching on that one I bet...

  36. Well, I'd disagree (somewhat) with Eric. As your model is at least in part built upon measurements with errors (recent polls), it simply cannot be described as "it does not have a statistical margin of error". The errors may be too complicated to compute, but trust me they are there. It is true that as an aggregate of multiple polls the model's MOE should be smaller than for any individual poll; it is probably further reduced by other numbers your model is using which are error-free (like the results of past elections).

  37. Ekos has a new seat projection (April 19):

    Con: 146
    Lib: 69
    NDP: 44


    "It is remarkable how similar these seat projections are to the current composition of Parliament. While the Liberals are poised to lose several seats to the NDP (mostly in Quebec), the Conservative and Bloc numbers are relatively stable.
    It is interesting to note the voter inefficiency of the NDP vote in Quebec. While Bloc support has plummeted ten percent in the last few weeks, the party has benefited immensely from seat splitting among the other parties. The current projections put them only one seat behind their performance in the last election."

  38. First,

    Granted, there might be some way to calculate it. But because it uses older polls as well as new polls, and weighs polls according to three different factors, etc. it would be virtually impossible to calculate.

    In the end, the vote on May 2 will determine the error. Perhaps in the future I will use that error as the "MOE" for future elections.

  39. How similar is the Dion melt down to the Ignatieff Howard Dean scream : Rise UP?

    and Ignatieff repeats it.

    The CPC are saving money on not having to put it into an attack ad.

    At least Dion immediately realized he goofed.

    As much fun as it might be at a travelling salvation show Canada does not want an evangelistic snake oil salesman to be PM.

  40. Éric, to follow up on First's comment, you can't actually know whether small changes are important or not. Aggregating multiple polls eliminates _some_ random background noise (=fluctuations within the margin of error of polls due solely to change in sample), but certainly not all of it. (You're not surveying 100% of the population, after all.) For example, today's projection shows the Libs at 38.1% in Kitchener-Waterloo and the Cons at 38.0%. This surely means nothing more than, "the Liberals now have a very slightly better chance of winning KW than the Conservatives". If tomorrow has the percentages the other way round, that could mean an actual trend, or it could just mean one bad poll got added to the mix. (One poll in 20 being outside the "margin of error" as commonly reported, and all that.)

    What is really helpful about your site is the reminder that national polling numbers can hide very different regional trends, and that most poll fluctuations have less meaning than pollsters typically assign them in press releases.

    Focusing on which marginal seat has flipped on a particular day rather detracts from that.

    And surely you could attempt to calculate at least the lower bounds of your margin of error, for example by assuming that all surveys in your model were conducted simultaneously.

    Perhaps that's something you can think about more after the election is over -- the model is an excellent resource, I'm just suggesting a bit more caution in the interpretation.

  41. South Parkdale Jack.

    Funny I see more Kennedy signs that Nash ones. So I guess it depends on where you are looking.
    A more convincing argument than this would be to say that the NDPish candidate won the municipal election so the area may be feeling a stronger NDP vibe recently.
    Using the number of signs YOU see (based on your perception as an NDP fan..which is naturally going to notice Nash signs more than anyone) as a basis for why you think the projections are wrong seems rather silly.
    But I guess polls are just as silly at times so maybe next year Eric can have people provide counts of signs to add an extra weight to the projection - seems like these are decided voters and easier to count than those that won't answer the phone or take on-line polls. = )

  42. I notice a lot of people commenting here are talking about voter turn out, so I did some looking about. The only thing I could find with a quick Google search was this:


    I don't know how this holds up for other polling firms, but I notice that according to this site, historically, the polls have tended to overstate Liberal support and understate Conservative support. Can any one provide more comprehensive data about polling predictions versus actual results for the past several elections? Given how important small shifts in support seem to be in determining seat count in this election, I would be very curious to know if the trends hinted at on the one site hold up across the board.

  43. Rick,

    Some of this is taken into account in my projection.

  44. Bugging you on Esquimalt - Juan de Fuca again19 April, 2011 15:23

    Esquimalt - Juan de Fuca so off in the projections. It's a tight three way race to some, but at most, it's a tight NDP-Conservative race to most observers. This is the problem of applying regional wide polling data to a very complex race losing a popular and 'independent' incumbent.

    You might find that adding factors like the amount of spending by a party's EDA's could get a better projection? That would certainly skew things towards Troy's favour. Also, what about how long a party's candidate has been nominated for? (aka, this is Troy Desouza's 3rd campaign vs. Lillian Spasyk was just nominated) You could also factor the number of party leadership visits per riding as a measure of how serious a party is for a pickup to one race or another. Finally, taking a look at census canada data could help your numbers in the future, as Esquimalt is increasingly adding suburban Conservative-leaning households in Langford.

    UVic polisci professor believes Esquimalt - Juan de Fuca will be an NDP versus Tory race. http://www.youtube.com/watch?v=zjVfOMmLE1A#t=1m07s

    To bug you again, your projections for Esquimalt - Juan de Fuca are way too good for the Liberals to take seriously. I don't know anyone in Victoria who thinks their credible. There's a reason many media outlets look at Esquimalt - Juan de Fuca as one of the Tories easist pickup opportunity this cycle--Keith Martin isn't running, and for this riding, that makes all the difference.

  45. @Éric

    Perhaps start with MoE on single-poll projections, and try to expand out from there over time? Likely something for after the election though... That'd be brutal to start trying to deal with mid-campaign.


    That link is comparing the differences between last pre-CAMPAIGN poll and the actual results, not between the last poll before the vote and the actual results. It is not necessarily about polling accuracy, and could just as easily be a "pre-election bump" in polling numbers for the party leading in the polls at the time (something that was discussed in many comments in the first few days of the campaign).

    Based on their findings, we should expect a drop on May 2nd for the Conservatives (as the party leading in the polls at the time) in the area of 6% (give or take) from their numbers on March 25th.

  46. To add to what I and franzca wrote about errors in your model. If indeed your model is too complicated to explore the error propagation analytically, then you can always resort to numerical (Monte Carlo) error estimation techniques. I do it often in my research. You basically run your model not once, but hundreds or even thousands of times, every time replacing the polls (mean) numbers with randomly generated numbers with given means and MOE. Of course, you have to assume something about the error probability distribution function; the simplest approach is to assume it to be normal (Gaussian). So at the end you'll have not one prediction per riding, but say a thousand. Then you can easily compute any statistics you want - mean, std., two-sigma confidence interval (which I think what reported MOE is) etc.


  47. Two things to keep in mind:

    (a) This poll IS a poll of the current mood. It's not a projection of election day, but what would happen if the election were held right now. As the election gets closer, things like get-out-the-vote, "do I like them enough to give them a majority", "do I vote strategically to keep someone out", and "do I even feel like voting at all" come into play. I don't doubt that this analysis accurately reflects the overall mood right now. The last week is going to show critical movement though.

    (b) People talk a lot about strategic voting, and how the war between the Liberals and NDP may let the Conservatives slip by. But that's only if the voters are very stupid.

    In 135 races, it's CPC vs LPC. In 70 races it's CPC vs NDP. Meaning that in at least 205 races, voting for the second place guy will stop the conservative party, and voting for the third place guy won't. The number of races between Liberals and New Democrats? 30. Voters are stupid if they don't re-elect the incumbent.

    Let me put it differently. If you tell people "vote Liberal to stop the Conservatives", there are anywhere between 70 and 100 seats where you might actually help the conservatives. Likewise if you tell people "vote NDP to stop the conservatives", there are anywhere between 135 and 165 seats where you might actually help the conservatives.

    Plus... none of this accounts for Quebec, where right now it's impossible to tell who is in second place.

    This election is going to be decided on local trends, not the national mood. The Liberals, Conservatives, and New Democrats are all considered toxic in different parts of the country. The question is where unsatisfied voters are going to go (or who will benefit from them staying home).

  48. Question about Renfre-Nippising-Pembroke. I don't live in the riding and I haven't been through there, but apparently former Liberal and now independent candidate Hec Clouthier is doing very well. It would take a lot to unseat Cheryl Gallant with the tories doing so well in the polls, but your forecast of 0.7 per cent for an independent seems wildly off what local media are reporting.

  49. Anonymous 17:51,

    Yes, that could be. I don't have any way to model Clouthier's potential support, however. I strive for uniformity in how I handle each riding, so I can't just guess that he'll get 15%.

  50. Brendan, the best way to determine if someone is going to vote is whether they voted last time.

    Is there some huge shake up or increased voter enthusiasm this time around aka Obama 08? Nope

    In fact, the students won't even be on campus and April 30/May 1 is the busiest moving day in Canada outside Quebec.

    Voter turnout is polling lower than last time at 57% and this might even be an overestimate. We'll see on May 02 but I don't see any groundswell of voter enthusiasm neither do pollsters.

  51. Got a remark re Renfrew-Nipissing-Pembroke

    Again during the campaign the CPC candidate Gallant has done a fubar !!

    So indeed Clouthier, who represented this riding through at least three elections and won does resonate !!

  52. Following up on First's comment about simulations, it's worth reading Nate Silver's methodology explanation for his US election estimates. (See http://fivethirtyeight.blogs.nytimes.com/methodology/ .)

    Not directly transferable, for all sorts of reasons (five parties rather than two, stronger influence of national factors etc.), but demonstrates that it should be possible to calculate the robustness of seat predictions. Assuming enough time and data, of course.


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.