Friday, April 1, 2011

Wild swings in projection as Liberals roar ahead

My grandfather once gave me three words to live by: patience, tolérance, détente. The first two mean the same thing in French and English, while the last means to relax, to take it easy. In other words, not to sweat the little things. And with Nanos Research the only polling firm reporting six days after the campaign has started, I think I need to take his advice to heart. But from what I can tell, at this time last year Harris-Decima, Ipsos Reid, Angus-Reid, and EKOS had all started reporting regularly, in addition to Nanos. What's the hold up?

At least yesterday's Nanos poll had a bit of a swing that had some effect on the projection. But today's Nanos is a bit of a snorer, and is virtually the same as yesterday's. The result is that there is little change in the projection.
The Conservatives are unchanged at 38.6%, but have picked up one seat in Quebec and are now projected to win 151 seats. The Liberals, meanwhile, are down 0.1 points to 27.2% but are still projected to win 73.

The New Democrats are steady at 16.6% and 33 seats, while the Bloc Québécois is down 0.1 point to 38.9% in Quebec and are down one seat to 51. The Greens are up 0.1 points to 6.7%.
Regionally, there have been very few changes. Both the Tories and Liberals are down 0.2 points in British Columbia to the benefit of the Greens.

In Ontario, the Liberals are up 0.1 points to 33.4%.

Quebec has had a tiny wobble, with the Liberals and Conservatives both picking up 0.1 points.

And in Atlantic Canada, the Conservatives are down 0.5 points to 37.2%. The New Democrats take advantage, gaining 0.4 points.

The only seat change was Beauport - Limoilou, which has switched back over to Conservative MP Sylvie Boucher. But as you can see in the riding projection, I have the Tories tied with the Bloc in the riding. Boucher gets the nod because she is the incumbent.

I'll have a full report on the Nanos poll later today. Word is that EKOS will be reporting today, so we will have another set of numbers soon.
And if you're still puzzling over today's post title, it was written by a fish.

26 comments:

  1. Yes, I was puzzled by your' fishes headline, lol. Just a word of advice though, outside Quebec, most anglophones wouldn't know that a fish has anything to do with April fools day. That is a French tradition that never really crossed over to anglo pop culture

    ReplyDelete
  2. Nanos daily polls Thursday mathematically extracted (Assumption the first 3 days were all the same):

    CPC 39 Liberal 27.7 (down 12.3 from Wednesday) NDP 20.2 (up 11.7)

    The Dion Liberals switching back to NDP as Mr. Layton shuts down "tar sands"?

    Mr. Ignatieff disappoints in verbal scuffle with Winnipeg senior?

    ReplyDelete
  3. Bluegreenblogger,

    Perhaps not, but it should give people a clue that I wasn't serious!

    BCVoR,

    I don't think your calculations are valid. Let's take the polls as they are and as they come. Tomorrow we'll have three completely new days of polling.

    ReplyDelete
  4. Looks basically as if we're stuck in the MOE. I can't see any great changes in that number set.

    ReplyDelete
  5. Good April fools joke Eric on my calculations. They are too valid.

    Very simple math and as long as the sample size is a constant very accurate.

    If the Liberals surged yesterday and today they remain the same in a rolling poll it is intuitive that they have fallen back.

    The concept of the rolling poll is to spot trends ... the Liberals went up 12% and then down 12% to remain the same. However right now they are trending down. It won't show until the Wednesday 40 is aged out on Saturday and reported on Sunday. (It will falsely reinforce my CPC stronger polling on the weekends theory)

    Sorry for ruining the surprise.

    m 28 t (28 w 40 t 28) --- current rolling
    (m 28 t 28 w 40) t 28 --- previous rolling

    surely you can see both rolling come out to 32

    ReplyDelete
  6. Don Cherry Canada's representative at royal wedding
    NNW Insider: NNW has learned that Don Cherry has agreed to represent Canada at this month's royal wedding in Endland. Yesterday, PM Harper sent his regrets to Prince William and Kate Middleton, saying that he will be unable to attend their wedding, due to election timing.

    ReplyDelete
  7. Yeah, that's probably a joke.

    ReplyDelete
  8. Last election, Stephen Harper announced that he would be going to the Governor General to seek an election well in advanced.

    My guess is that they took that as a sign that an election was 100% possible and started to spend money on polling.

    I think the polling firms this time around were more cautious, as while an election was quite likely for quite some time, they wanted to make sure it was certain before spending the big bucks on polling.

    ReplyDelete
  9. I was wondering about your post title. Of course April fools. Also it seems the swing in the Nanos poll yesterday was a blip or a bad sample since the swing back averaged over 4 days is 1% from Lib to NDP today that means it changed back by 4% again yesterday.

    Daily rolling polls can fool people. They really need to be averaged over a week to mean anything. Sample blips are just that sample blips. A hockey game on TV during polling could cause that swing.

    ReplyDelete
  10. Eric you don't think I would post a joke do you?

    ReplyDelete
  11. Send Don Cherry to the royal nuptuals. He'd add a bit of colour at least. Thanks for the poll results too. :)

    ReplyDelete
  12. Is this fish thing something i saw in a french textbook once ?

    Some kids had taped a fish to a woman's back and were laughing hysterically. She was unaware.

    ReplyDelete
  13. BC Voice of Reason, how do you know those earlier numbers?

    First data point is that using D1, D2, D3 and D4, the results were (cons, libs) = (39, 28).

    Second data point is that using D2, D3, D4 and D5, the results were (cons, libs) = (39, 32).

    Third data point is today's, D3, D4, D5 and D6, and the results are (cons, libs) = (39, 32)

    From this, we know that D5 is a better day for the Libs than D1, but Im not sure what else we can deduce without knowing the specific values of the first few days of polling.

    ReplyDelete
  14. Jerome this is a rolling three day tracker.

    D1,D2,D3 = first data point
    D3,D4,D5 = current data point

    Also your numbers are slightly off.

    1st = (38.4, 28.7)
    2nd = (39.1, 32.7)
    3rd = (39.4, 31.7)

    We don't actually know the relationship between any of the days because any change can be attributed to either the value of the first day being dropped from the series or the new day being added to the series.


    BCVOR is wrong because he could be spotting trends that happened two days ago. Here's a hypothetical Liberal sequence to demonstrate:

    M = 20
    T = 40
    W = 30
    R = 30
    F = 40

    So on Wednesday the tracker would show 30%, on thursday it would show 33.3%, today it would show 33.3%.

    A gain followed by stagnant numbers does not nessecarily mean the Liberals are trending down.

    I've demonstrated the opposite.

    ReplyDelete
  15. Final point is directed at Earl, HtH, and the media.

    Nanos tracking has showed CPC numbers going UP for two days in a row.

    Tell me again how Harper is losing this campaign or blowing his shot at a majority ?

    The public likes his campaign, his platform, and his style.


    Maybe we should wait until the numbers actually show a CPC drop before having a collective freak out ?

    ReplyDelete
  16. Shadow

    Good point... on fun with numbers However I would put my assumption that the Liberals polled the same average of 28 in the first 3 days of Nanos poll is much closer to reality than saying the Sunday March 27 poll was below the Liberals all time historical low of 20%.

    Will you be shocked when the rolling average for the Liberals drops 4-5% because Wednesday's 40% drops off when the Sunday's rolling poll numbers are released?

    Not me.

    I won

    ReplyDelete
  17. BCVor i'm just sticking with the math.

    Remember the MOE of a 400 sample is kinda high so assumptions about what are realistic ranges for Liberal support on any given day are out the window.

    That's the reason why Nanos waited until day 3 to release a blended sample.


    He didn't want people trying to spot "trends" based on 1 or even 2 days worth of wobble.

    The MOE is just to high to draw any conclusions.


    Meanwhile EKOS has a new massive sample poll out showing CPC and NDP trending up, LPC trending down.

    ReplyDelete
  18. EKOS is out:

    http://ipolitics.ca/2011/04/01/conservatives-open-up-daylight-on-liberals-as-ndp-make-gains-ekos-poll/

    ReplyDelete
  19. BC Voice of Reason:

    you're likely correct, or at least its likely that your numbers are close to the real thing. and i agree that wednesday was likely a very good day for liberals in that polling sample. but you dont have the info to make exact calculations, unless im missing something.

    ReplyDelete
  20. Interesting that the Ekos poll just out has the exact OPPOSITE trend from Nanos - they have the NDP gaining ground from the Liberals since the start of the campaign!

    I remember in past election, a number of people tormenting themselves trying to figure out what the results from individual days were in nanos's three day roll. Of course you can't do it unless you know what the results were for three individual days at some point and then calculate backwards. When the NDP dropped 4% in one day from Wednesday to Thursday - my immediate assumption was that EITHER the NDP had an incredibly stellar night on the first night of tracking which was Sunday and then that was dropped OR for some reason the NDP had a really bad night on Wednesday night that was bad enough to drive down the three day average by that much - or some combo of both. The fact that NDP support stayed at 16% today makes me suspect that there was probably a very bad outlier night on Wednesday that will not get dropped until we see numbers on Sunday morning.

    ReplyDelete
  21. Eric,

    Does your riding-level prediction model tries to factor in a "following the leader" mentality, which apparently is present in some ridings? Take my riding (Hamilton Mountain) as an example. It's always been very competitive (was many times Lib, PC, and NDP). Here is the plot I made for Liberals only, which compares the national popular vote for Libs (red) vs. Hamilton Mountain vote for Libs (blue), since 1968:

    http://www.physics.mcmaster.ca/~syam/A.png

    The national data can be a bit hard to interpret (as initially there was no Bloc, which reduced popular vote for all major parties nationally), but at least for the last 10 elections the "following the leader" trend seems to be very pronounced, at least when it comes to Liberals: when they do well nationally, the riding vote becomes disproportionally higher, and the opposite is true when they do poorly nationally (they are punished disproportionally in the riding).

    Sergey

    ReplyDelete
  22. In EKOS

    Liberal Fortress Toronto CPC 40.6 Liberal 34.8

    Nationally Female CPC 31.5 Liberal 26.5


    Especially in an EKOS poll this is terrible news for the Liberals....

    This might be the start of a rout. Watch to see the Quebec federalist vote move to the CPC if there is a sense of an impending CPC majority without Quebec.


    Just a small observation once again Quebec is way over polled relative to its population.

    ReplyDelete
  23. Some of these projections are so far out of touch with reality that its hard to believe any of it.

    Winnipeg North for example.

    ReplyDelete
  24. Ah yes, the "one tiny weird thing makes everything wrong" school of thought.

    By-election results are fickle, and the model has to be standardized. Winnipeg North could prove to be an exception, but I have to use uniform methods in treating by-election results.

    ReplyDelete
  25. I just read a previous post I did in discussion with Shadow it end with

    "Will you be shocked when the rolling average for the Liberals drops 4-5% because Wednesday's 40% drops off when the Sunday's rolling poll numbers are released?

    Not me.

    I won"


    I would have sworn I ended this with

    "Not me.

    I wont"

    A dropped T makes me look like even more of a jerk than I already am. I hope I didn't tick anyone off as it just made me look bad.

    ReplyDelete
  26. "Will you be shocked when the rolling average for the Liberals drops 4-5% because Wednesday's 40% drops off when the Sunday's rolling poll numbers are released?

    Not me."

    Well, how was your prediction?

    ReplyDelete

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.