Political Betting has a new Angus Reid poll, the topline figures are almost unchanged from their previous poll with CON 40%(nc), LAB 24%(nc), LDEM 19%(-1). The company have shown Labour and the Conservatives unchanged in their last three polls, and continue to show a signficantly lower level of Labour support than other polling companies – a difference at least partially explained by weighting past Labour voters to a lower level.

The poll was conducted on Tuesday and Wednesday, so is the first poll conducted after the announcement of the end of the recession. It’s only one straw in the wind, and we should have several polls coming up in the next week or so, but this first signal suggests it is still business as usual.


40 Responses to “Angus Reid show Tory lead unchanged on 16”

  1. Can anyone please give a brief summary explaining how well Angus Reid did at predicting the last few Canadian elections?

  2. Am I imagining it, or has the last 6 months or so been one of the most stable periods ever for opinion polling? I don’t remember ever seeing so much “no change, no change”.

  3. Yes indeed biz as usual, WMA 40:28:19. Confirms my view that over the last 1 and 2 months there is definitely no trend.

    AR consistently over-estimates the CLead (3 +/- 1.5) but it will be interesting to see how right they may be on the night.

  4. Angus Reid.

    I just don’t buy their figures.

    Either they are consistently way out. Or else all the other pollsters are way out!

    And I know which I suspect.

  5. Given the present political enviroment the Angus Reid polls are the only ones that make sense to me.

  6. for Andy Stidwill and David in France…

    I know what you mean, I find it hard to take the figures given how widely they vary from the rest.

    But given their record in Canada we may have to rethink that
    (from wikipedia http://en.wikipedia.org/wiki/Angus_Reid_Public_Opinion)
    Since 2006, Angus Reid has covered eight provincial elections in Canada—more than any other pollster in the country—and the results have accurately predicted the outcome of each of these democratic processes. The firm also provided the most accurate forecast of the 2008 Canadian federal election.

  7. @DAVID IN FRANCE
    A few days ago when the 7 pollsters gave their opinions of the forthcoming GE result, a Labour supporter jumped on the fact that 2 predicted a hung parliament. I pointed out that 5 predicted a Tory victory. Therefore with the same logic I must agree with you, AR are out numbered. However, it is difficult to seperate personal views from these judgements. If you think that Labour have done a decent job with a few mistakes thrown in, a 16 point Tory lead seems outrageous. If on the other hand, you think Labour have gone a long way towards the ruination of this country, you might ask, why hav’nt the Tories got a 26 point lead.

  8. It is Ground Hog Day again. It seems that we are going round and round and round with repeat after repeat after repeat.

    How I wish the GE was next month.

    By the by I am no relation to the other poster under the banner of – Glenn

  9. I’d be interested to know whether AR’s difference is because of the questions they ask, or because they do not weight-down 2005-non-voters (if I understand correctly).

    If turnout is significantly increased over the lamentable 2005 turnout, other pollsters might regret weighting-down 2005 non-voters

    Either way, you have to admit AR is consistent and reinforce the calm waters message that the WMA shows us.

  10. But whilst they are understating labour they are not particularly overstating tories or libdems. Though compared with their first effort libdems have gone down 3 and labour up 2.

    So given that they are not out of kilter with tories or libdems in a significant way – who is benefiting from the labour share.

    ‘Others’.

    Is not this the problem with polls. ‘Others’ are distorting the issue.
    And to say polls are consistent is not quite right is it? We have had one pollster going from 10% lead, 17 to 9 to 13 and back to 9 again.

    Another went 10% to 13 to 9, to 12 to 10, 12 and then to 9 again.

    If these swings are within sampling errors – well whats the point?

    To be sure, ICM have gone from 11 to 9 to 10 to 11 again. So if ICM are a gold standard what we can say is we have ad 3 months of a steady edging up of the Tory lead.

  11. If I was an Angus Reid pollster, I’d either be very confident or very worried right now. Should their polling be shown to be as out-of-kilter with the final result, it will destroy their reputation.

    That said, their figures stay at a stable level below that of other pollsters. So I like to just lop a few points off AG’s Tory lead to know what is the situation.

  12. Andy Stidwill

    I’m no expert, but Angus Reid were closest for the 2008 Canadian Federal election – and IIRC the Vancouver and BC Provincial ones as well.

  13. Glenn Otto

    It is Ground Hog Day again – but only for you. For the rest of us, it’s July, and it was a very interesting election. :-)

  14. @ NBEALE

    “AR consistently over-estimates the CLead (3 +/- 1.5) ….”

    How do you know this?

  15. Though seemingly biased compared to other polls, the Labour vote always drops by about 3 or 4% on general election day. So it could just turn out to be right.

  16. One group of pollsters use one weighting, AR another. Just because AR’s weighing is different does not mean it is wrong. Just because a group agrees amonst itself does not make it right. Indeed, Groupthink can encourage groups to agree to a common error (as demonstrated by some commentators on here).

    Statisticians will warn you that when models are used in an environment in which the variables are significantly different to the environment in which the model was built, the results should be treated with a very big pinch of salt. The models used by the more established pollsters may be more accurate in more benign economic/political circumstances. Being new to the game, AR may be better aligned to the current UK environment. Given the anti government sentiment at the moment, it may not be inappropriate for AR to use a greater weighting against Labour. It is not unreasonable to beieve that many disillusioned people who previously voted Labour will stay at home during the GE.

    All models and projections need to be treated with a pinch of salt and given a ‘reality check’ from time to time.
    If an alien landed, was told the size of the UK deficit, the cuts and the tax rises that the country was facing as a result, and that the government had previously told us that there would be no bust, would balance the books, and that when we did go bust we would be best place to recover from the recession but were the last, and was then asked if they thought they opposition had a lead of 6% or 16%, how do you think they would answer?

    I think that AR may be closer to the truth than many would dare to believe.

  17. @TONY M
    You may think that Tony, you may very well think that, I with my party affiliations see things exactly as you do. However, in line with my comment above to D IN F, extraordinary though it seems to us, a lot of people do not see the maladies you mention as Labours fault.

  18. DAVID IN FRANCE

    “Angus Reid. I just don’t buy their figures.
    Either they are consistently way out. Or else all the other pollsters are way out! And I know which I suspect.”

    Are they consistently way out?
    Looking at the Tory lead over Labour since early October 2009 only 3 out of the 9 highest leads (17% or higher) are from AR polls.

    I have an open mind but Tony M may well be right that AR is tending to more reliable than the other polls.

    A possible reason for this is that some pollsters often adapt their polling techniques to the previous election, rather than the coming one.e.g. if there was a Tory voter strike from 1997 to 2005, discounting the votes of people who did not vote in the last election will understate the number of returning Tory voters who may have voted consistently prior to 1997. AR have not made such adjustments.

    Also AR tend to take large samples which should slightly reduce their margin of error.

    I personally rely more on AR than other polls to see the trend as they are relatively consistent while some other polls appear to be excessively violatile.

  19. In a way it’s just as well we have AR, as the other existing pollsters all estimated the Conservative lead in 2005. Balances it out a bit.

    The end of the recession will not necessarily make much impact on the political climate. Does anyone else remember the phrase “voteless recovery” between 1993 and 1997?

  20. “underestimated” I meant

  21. I still think Angus Reid are the most likely to be accurate

  22. It gets slightly tedious repeating this every time an AR poll comes out, but here, masochistically, we are again:

    1 – We have recent evidence the other pollsters are accurate – Euro elections less than 1 year ago. There is no evidence AR is accurate as a measure of voting intention.

    2 – If AR are right about Labour share, everyone else is (way!) outside their margin of error, despite point 1 above – how likely is that?

    So in summary, yes AR might be right, and some strange shift in public opinion that makes everyone else’s methodology wrong may have occurred.

    But anyone following probabilistically based reasoning would conclude that it is highly likely that AR is the pollster getting it wrong. Occam’s razor.

  23. AR results:-
    Cons: 40.0%
    LAB: 23.9%
    LD: 19.1%

    England & Wales only:
    Cons: 42.3%
    LAB; 22.6%
    LD: 19.7%

    http://www.visioncritical.com/2010/01/steady-tories-keep-16-point-lead-over-labour-in-britain/

  24. Don’t forget Mike Smithson’s famous law which basically states that most polls overestimate Labour’s share. This being the case, maybe Angus Reid could be the star performers of the coming election campaign.

  25. STATTO
    “1 – We have recent evidence the other pollsters are accurate – Euro elections less than 1 year ago. There is no evidence AR is accurate as a measure of voting intention.”

    But election polls for obvious reasons tend to be more accurate and closer. It proves little about recent polls.

    “2 – If AR are right about Labour share, everyone else is (way!) outside their margin of error, despite point 1 above – how likely is that”

    Yes fair point but bear in mind the fact that some pollsters make adjustments for liklihood to vote and other assumptions which tend to be based on historical factors and some of these assumptions may be wrong based on current events.

  26. @ STATTO

    “1 – We have recent evidence the other pollsters are accurate – Euro elections less than 1 year ago. There is no evidence AR is accurate as a measure of voting intention.”

    And therein lies the incorrect assumption. The voting model has NOT been tested not against a GE in 2010, only a Euro elections in 2009. Many people who could not care about the Euro’s will vote in the GE (and vice versa). Is the Euro election a true validation of GE voting?

    Michael Pidd in his book, “Tools for Thinking: Modelling in Management Science” provides a useful summary on the problems of apparently straightforward validation. To paraphrase, “a problem occurs when the model is used to precict the future. A golden if somewhat cynical, rule among forecasters is ‘when asked for a forecast always give a number or a date, but never both’. The future is uncetain and unknown.”

    “2 – If AR are right about Labour share, everyone else is (way!) outside their margin of error, despite point 1 above – how likely is that?”

    Most of the world used to think that the world was flat. What was the likelihood of them being wrong?

    All of the pollsters will be wrong to some degree. Their models are, afterall, just models – not reality. Who will be the most wrong?

  27. @Old Nat

    I found this on the web regarding accuracy of Canadian pollsters:

    http://www.sfu.ca/~aheard/elections/poll-results.html

    AR were most accurate in the last Canadian election as you mention. Follow the Angus Reid link for more details on this poll. Lots of the sub-links are very interesting.

    On this link, you’ll find recent Canadian polls:
    http://www.electionalmanac.com/canada/polls.php

    Note that AR are much more in line with other Canadian pollsters than they are here.

    The AR Canadian poll uses the same “support” term that appears in the UK version.

    @Tony M

    “And therein lies the incorrect assumption. The voting model has NOT been tested not against a GE in 2010, only a Euro elections in 2009.”

    And no model can be – but they were tested in a GE in 2005 and were again accurate. AR is not tested at all in the UK, so I know which polls I’d trust more.

    “Most of the world used to think that the world was flat. What was the likelihood of them being wrong?” – evidence was presented that contradicted flat earth, model was adjusted – Occam’s razor. Where is the evidence that the established pollsters have suddenly gone astray? There is none.

  28. “Most of the world used to think that the world was flat.”

    There’s very little evidence that most of the world’s population used to think the earth was flat. Where is the polling evidence supporting your assumption? ;-)

  29. US economy grew by 5.7% in the last quarter. As I said to the calls of “rubbish” by someone, I am now more confident that the announcement of the British economy has only grown by 0.1% is a turning point that will ensure that Labour are soon in the mid twenties.

  30. Don’t YouGov usually publish a poll on behalf of the Daily Telegraph on the last Friday of the month??

  31. AR could be wring and all the others right. But it is not long since nearly all the pollsters called a GE wrong and only one called it right. Sure, the pollsters subsequently changed their mothodology, but it highlights how models can drift away from reality and the majority can be wrong.

    At the very least, the AR polls prompt a good debate. Without them there might be little discussin around the differing methodologies and the possibility that, as in previous GE, the methodologies could be deficient or inappropriate.

  32. Mori poll

    CON 40% (-3); LAB 32% (+6); LIB DEM 16% (-4)

    England only

    CON 42% ; LAB 31% ; LD 17% ; Other 10%

  33. Philip JW: The US 5.7% is annualised. We don’t do that over here. Approx. 1.3%. Still huge but somehow the Americans produce larger swings than in Europe. They also revise downwards more often than here.

  34. AR consistently show much higher votes for UKIP & BNP than other polls and that’s why Labour’s share is lower, even though the Tories & LibDems are much the same as other polls.

    I don’t understand why potential UKIP/BNP voters would be far more willing to tell AR about it than other pollsters. Is there some difference in methodology that could do this?

    If Labour is indeed losing votes to UKIP & BNP, it makes it very difficult to predict how this would affect seats, in my opinion.

  35. @ Statto

    “2 – If AR are right about Labour share, everyone else is (way!) outside their margin of error, despite point 1 above – how likely is that?”

    You realise that 29 of the 31 final-day polls since 1983 overestimated the Labour vote? On that basis, the answer to your question is “very probable”. More than 93% probable. Of course, I know that this time is different. But also remember that the “spiral of silence” introduced to deal with “shy Tories” now boosts the Labour percentage, not the Tories. But is this really valid? I can’t see it being uncool to support Labour, even though everyone hates Brown. As the media say, there seems to be no great enthusiasm for Cameron like there was for Blair in 1997.

    Angus Reid have been remarkably consistent; does anyone know where to find their raw data so people can have a look? I can’t seem to find them…

  36. Derek Pierson

    “AR consistently show much higher votes for UKIP & BNP than other polls and that’s why Labour’s share is lower, even though the Tories & LibDems are much the same as other polls.”

    Is that right, or do you mean “Others”? The SNP are making progress at Labour’s expense (though it will bring them very few extra seats). Surely AR can’t be consistently sampling a handful of extra nationalists?

  37. JohnBDick

    If you look at the detailed figures, it’s UKIP & BNP that are higher. SNP can’t make much of a difference for the UK because even a good result for them just moves them from 2% to 3%, because Scotland only has about 9% of UK population (& presumably voters).

  38. Derek Pierson

    That’s the way it should be, certainly, unless there was some oversampling or other error. If UKIP and BNP are up, it may be significant in a few constituencies and the effect in seats is anyone’s guess.

  39. Why does the Angus Reid result never seem to change? IS that dubious?
    Do they ask the same people?