New Angus Reid poll

Political Betting have a new poll by Angus Reid. The topline figures, with changes from their last poll, are CON 40%(nc), LAB 24%(+1), LDEM 20%(+1). Others are at 15%, down 3.

Overall there is no significant change from their previous poll. The Conservative lead is, needless to say, much greater than in most other recent polls, as the level of Labour support is lower. This is not a sign of some sudden Conservative recovery – Angus Reid have been showing the Tory lead up at this level all along – it’s just to do with different methodology.

Specifically Angus Reid assume no false recall when weighting their samples to recalled 2005 vote, which means their samples have fewer people who claim to have voted Labour in 2005 than phone pollsters do. For some reason they also tend to show a higher level of support for others – I’ve no clear explanation for that, it could be their question (they ask who people would “support”, not how they would vote) or it could be something entirely different.

Still to come this year we should have a YouGov poll for the Telegraph, which normally comes out the last Friday of the month (though it’s probably safe to assume it will be a different day this year!), a ComRes poll for the Indy, and Ipsos MORI’s monthly monitor, which was apparently conducted last weekend but has yet to be released into the wild. I expect, as was the case last month, that one of the Sunday papers is going to print it.


124 Responses to “New Angus Reid poll”

1 2 3
  1. Why are you including this poll in your running averages when you feel the need to list a string of qualifications about the result ? I repeat this continues to bring into question the neutrality of this site and as for the neutrality of the comments……. more moderation needed Anthony.

  2. Anthony, any idea why PB chose to commission a poll just a week after their previous one?

  3. Anthony,

    I am puzzled as to why we include these polling figures if they do not ASK the same question as other pollsters….I would certainly answer differently to a how are you going to vote as compared to a who do you support question.

    What surprises me is that the latter question still gives a 40% share to the conservatives….as it seems to me unlikely that all of those would be ‘supporters’ as opposed to ‘voters’….or perhaps the Sun is more in touch with the realities of political affiliation than I’ve credited.

    BTW thanks for this blog most of the comment here is stimulating and interesting.

  4. @Anthony

    “The Conservative lead is, needless to say, much greater than in most other recent polls, as is the level of Labour support.”

    I guess you mean that the level of Labour support is lower?

  5. @Alan W

    “Why are you including this poll in your running averages when you feel the need to list a string of qualifications about the result ?”

    Until we have a definitive result from a GE, there is as much reliability in any of the pollsters Anthony has included. Then, and only then, can the accuracy/methodology of any pollstser be brought into question. In an earlier thread there was a detailed explanation of why AR was being included, until they are/if, proven to be out ona limb, they are as accurate as anyone else.

  6. I think the main thing on this poll is the 40% Tory vote. It is the consistent thing across all the polls. Probably the one thing we can believe. It seems very solid.

  7. I am new to polling prediction and reasoning. However , as previosly stated, I am a big believer in relating projections to the bookmakers odds. It may well be that I am introducing a non acceptable criteria to polling. If so I apologise.

    As at today the odds qouted by William Hill suggest seats at the GE would be as follows:-

    Conservative 350/375
    Labour 150/199
    Liberal Democrat 50/59

    Based on Angus Reid Poll using Electorial Calculas result would be:-

    Conservative 368
    Labour 197
    Lib Dem 54

    The bookmakers seem to favour Angus Reid and not the other polls that show a reduced support for the Conservatives.

    Odds for the most seats are:-

    Conservative 1/14
    Labour 13/2
    Lib Dem 80/1

  8. No update to my model, because I don’t include AR polls in it. The combination of them being new to UK polling, using an odd weighting method that produces much higher ‘others’, and ask who people ‘support’ instead of who they will vote for; means I simply can’t take the poll as reliable.

    Most telling is that it’s *not* a voter intention poll, but presented as one. People can support a party they have no intention to vote for. For instance, people dissatisfied with Cameron might ‘support the conservatives’ but be voting UKIP this time.

    I also have to say I can’t understand the use of this poll in generating a polling average, even a weighted one. This is not a voter intention poll, don’t include it with other voter intention polling.

  9. There is no good reason not to include AR in the polling avergae.

    No two polsters use exactly the same methodology be it timing weighting or questions. AR have a proven track record and their methodology is transparent.

    It’s when you don’t have the details of the methodolgy that you should weigh down the significance of a poll.

    Peter.

  10. @Peter Cairns

    Again, I point out the very good reason, that it’s *not* a voter intention poll. The question asks which party you support, not which party you will vote for. And there is a significant difference in that, as people who don’t intend to vote this year, and would say so, may still say they support a party when asked who they support

  11. Jay, so by your logic more people will vote Labour (as measured by the other pollsters) than actually support them? This would imply quite a high degree of anti-Tory tactical voting, which seems a bit unlikely after 12 years of a Labour government.

  12. Aan W – “bias”

    Bias? I see no bias. Anyone see bias on Anthony’s part?

    AR is as valid as any poll. Most ask the question “How would you vote tomorrow?” which to many is a silly question as the isn’t going to be an election.

    Unless you have powers of clairvoyance, you have no more idea than me or any other poster what is right and what is wrong regarding public opinion.

    The beauty of this site is that all the posters are presented with exactly the same information yet they manage to form incredibly diverse opinions from that information.

  13. @Pete B

    Yes that actually is possible. Because sometimes people vote for parties they do not support.

    And some times people do not go as far as saying they ‘Support’ the party they have voted for in the past. Lots of people who have voted Conservative/LibDem/Labour in the past wouldn’t describe themselves as the supporter of any party.

  14. In fact, I’m going to go as far as saying that all the poll shows is that the Conservatives have a more solid ‘base’ of party supporters. Remember that “don’t know” or “support no party” are removed from these figures. It is otherwise meaningless as a voter intent poll.

  15. JAY BLANC If your analysis were correct then it would seem Angus R is understating the Tory vote given the large amount of recent polling evidence that Tories have a high liklihood to vote. I agree witjh Peter there is no obvious reason why Angus R should be excluded. I suspect many of those who want it excluded don’t like it’s results.

  16. @ Jay Blanc

    Or conversely, most UKIP supporter I know, and there are a lot, will be voting Tory holding their noses as more of Brown is not an option for them. They see the possibility of a hung Parliment where the liberals have a say as the worst possible option.

  17. Alan W – my rules for the polls I report are straightforward, and are pretty much the same as what the British Polling Council consider their criteria for what constitutes a polling company. Anyone publishing anything using sampling and/or weighting methods that are likely to produce representative figures. Beyond that, I seek to explain, but not judge: there are several companies who do things differently to the way I would do things were I running their operation, and a couple I think are doing things that are downright wrong… but I’d rather give people the information and let them make their own judgement.

    John – no one really asks the same question as everyone else. ICM and AR mention “your area”, AR say support not vote, ComRes, AR and MORI include a squeeze question (and different squeeze questions at that!). Those other differences could in theory make *more* of a difference than asking support/vote (I think some people are grossly overestimating the actual difference that wording is making, just because it is the most obvious difference in methodology. It’s just as likely to be something to do with weighting or sampling). What’s clear is that while AR don’t use the word vote, by asking who people would support “at a general election” they are trying to measure the same thing from a different angle.

    You could just as well argue that you shouldn’t combine polls that are reallocating don’t knows with ones that don’t, or polls that factor in turnout with ones that don’t – and it’s an argument I’d have a lot of sympathy with! Strictly speaking, they are all measuring slightly different things. The average, is an average of what the polling companies consider to be their topline voting intention figures.

    If you look at the “…MORE” under the average, you’ll seem I am not really a fan of the whole concept of polling averages anyway. It’s a quick shortcut for someone who wants to know what the polls say without following them in detail. If you have the time to follow the polls properly, I would recommend ignoring it and judging on the broad overview of the polling.

  18. I found J A Brown’s comparison of this poll with bookies’ odds very interesting. Bookies aren’t often wrong!

    In fact we might just as well analyse different bookie’s odds as different poll results in my opinion.

  19. Surely it does not matter that AR use a different methodolgy as long as they don’t change the methodology between polls! The trend should be valid. The Cons have consistently been at 40% with all pollsters for months. The rest are all over the place. But providing it stays at 40% or thereabouts, what the rest does, does not really matter enormously to the eventual Con majority which on my (albeit amateur) Exel model is 22. The AR figures, taken with the detailed ICM poll, change that to 29. But that takes no account of tactical voting in specific marginals. The Cons will pick up a lot more seats from that. Has anybody ever calculated how the swing varied from the national average in the top 100 marginals in 2005 ? Or where that information might be ?

  20. Anthony – I’m trying to understand how AR’s failure to weight for false recall affects the overal numbers here and would appreciate some help.
    If AR don’t weight for false recall they therefore have as you say fewer respondents who claim to have voted Labour in 05. From that it’s easy to appreciate how this could reduce the Labour score, but what I can’t understand is why it doesn’t act to also increase the Con/LD scores also. We know from various polls in marginal seats that there appears to be some level of switching direct from Lab – Con, and given the regions where the Con lead is greatest this would make sense. If AR are systematically underestimating Labour support, as other pollsters presumably would claim, I can’t see statistically how this would mean they remain in agreement on the Con/LD levels of support. For this to happen would surely mean either all the people who voted Labour in 2005 but don’t admit it now say they will either not vote or vote for ‘others’. As we know from various polls in marginals, there must be some level of direct Lab/Con switching, so unless none of these direct switchers deny their previous support of Labour I can’t see how AR can generate these differential figures if the disparity is based solely on the false recall weighting issue.

  21. David (in Oxford):

    Judge for yourself. This is the link to Pippa Norris’ website, scroll down and download the Excel version of the 1992-2005 Parliamentary Database, for all the results and swings for the last election.

    http://www.hks.harvard.edu/fs/pnorris/Data/Data.htm

    The best way to test hypotheses is to get the data out and have a good dig around.

  22. I am in total agreement with A Wells, Peter Cairns, Barry P &
    Shopkeeper Man. Th AR Poll has as much right to a place here as any other – at least its up to date, which is more than can be said for IMORI when we get it.
    It seems very much to me that any poll which does not predict a Lib/Lab pact is classed as rubbish by certain people.

  23. Anthony

    Thanks for the link to the Pippa Norris site. I hadn’t seen that before.

  24. We should remember that telephone pollsters tend to find more Labour supporters than there should be ,often substantially so hence the need to past vote weight downwards . The Internet pollsters previously only Yougov but now also AR tend to sample more Conservative pollsters ( Internet users who answer polls tend to be older and live in the South East and London .
    Others have pointed out the consistency of Conservative support in recent polls at 40% . Contrarily this should not be happening given the differing methodologies , sampling errors of each poll and is an indication that whatever the actual level of Conservative support it is unlikely to be 40% .

  25. To some extent, despite undoubtedly being out of line, I think Angus Reid actually does us a service by making the average projection more accurate. All the polling organisations save NOP overestimated the Labour lead in the run up to the 2005 election campaign by at least 2%.
    I would estimate the Conservative lead to be about 11% at present, which I think is likely to be the result of the next election, giving an overall majority of 1979 proportions (about 40).

  26. These differences will keep us guessing through next year.

    I think the polls will find it difficult to predict the next election. The as ever, result will be determined as much by those who don’t vote as those who do. I think this group might change compared to previous elections.

    I think that John Major was helped because a lot of people who do not normally vote were scared by Kinnock; Blair was re-elected for a third time, in part, because of Howard’s unpopularity in the English marginals.

    Is it not the assumptions around don’t knows, may or may not votes…that make up a lot of the differences in the methodologies.

    For what its worth I think people quite like Cameron there are large sections of the English electorate that can’t abide Brown.

  27. Alec – it probably isn’t solely that. To be honest, think the difference is too big to be just down to the weighting anyway.

    There can be no doubt that the different weighting figures make a difference, but that doesn’t rule out lots of other factors making a difference too. If the only difference between this and an otherwise identical poll was to assume some level of false recall and weight (for the sake of argument) to ICM’s figures, then you would get both lower Labour AND higher Lib Dem and Conservative support. That suggests there is also some other differences, but that’s no great surprise.

    There are some obvious other differences to start with like not adjusting for likelihood to vote like phone pollsters do, and not re-allocating don’t knows like ICM and Populus do – not to mention that it is done online without an interviewer. Even comparing it with YouGov there are major differences in the use of a squeeze question and the approach to weighting (not to mention, they use a different panel, which could produce differences).

    Unless you have companies like ICM and Populus who use methods that are almost identical in many ways, differences between pollsters are very rarely down to single factors, especially when comparing online and phone pollsters.

  28. @ANDREW
    I think there are a number of posters on this site who have thought as you do since the recent hardening of the Labour vote established itself. 11 to 15% is a sensible estimate for the Tory lead.

  29. I’d be very cautious over use of betting odds… The bookies phrase the proposition here as “Most number of seats” not on who will form government. Someone can bet on “Conservatives have the most number of seats” and still win when the election results in a Lab/Lib coalition.

  30. Anthony
    Thanks for the link to Pippa Norris website. That is an amazing volume of data. I will have a ‘dig around’ as you suggest to see if any trends exist that may worry a few more Labour MP’s who thought they were in ‘safe’ seats! Seems to me that the bookies have got this spot on.

  31. @Barry P

    Until we have a definitive result from a GE, there is as much reliability in any of the pollsters Anthony has included. Then, and only then, can the accuracy/methodology of any pollstser be brought into question. In an earlier thread there was a detailed explanation of why AR was being included, until they are/if, proven to be out ona limb, they are as accurate as anyone else.

    ———

    Not so. As far as I recall, Zogby was excluded from Nate Silvers calculations for the US election. His polling was consistently way out and some aspects of his methodolgy were questioned.

    Angus Reid are, arguably, skewing the overall figures. Perhaps. Perhaps not.

    But a definitive decision has to be made on their inclusion. And, if they are kept, a reason (or two) ought to be offered.

  32. Only 5 of the last 23 polls have the Conservatives at greater than +14.

    4 of those 5 are Angus Reid at +17.

    The AR polling sample looks suspect to me.

  33. @DAVID IN FRANCE
    Anthony has given his reasons on this matter at least 4 times to my knowledge. Is there any significance in the fact that the people complaining about this pollster are anti – Tory and much happier with news of Labour recoveries ?

  34. Oh what a stew some people are getting into—we’ll
    all see eventually that only YouGov and ICM polls can be relied on.
    As for comment that bookmakers are (seldom) wrong
    they were in 1970 and very wrong in 1992 to say nothing of the Glasgow byelection earlier this year.

  35. @ David in France.

    “But a definitive decision has to be made on their inclusion. And, if they are kept, a reason (or two) ought to be offered.”

    It has, AW has made it/them. Even in my limited span of this blog, it is blatantly obvious he has far more experience than anyone contributing and, must have by now shown little or no partisan tendencies either way and made well thought out decisions otherwise, we wouldn’t be here now would we?!

    I just wonder what approach the, shall we say, labour leaning members’ opinion would be of AR’s poll if the results were showing a Tory lead of say only 5 points. I dare bet this thread would contain only half the posts it now does.

  36. “I dare bet this thread would contain only half the posts it now does.”

    I expect it would be exactly the same length, but with different people.

  37. @AW

    “I expect it would be exactly the same length, but with different people.”

    You are probably correct. Pictured an obvious reply as soon as I pressed the ‘send’ button. More haste, less speed.

  38. I make a case for excluding AR, which is simply that while all the other pollsters appear to show sample variance around the same (approximate) mean for each Party, AR always understate Labour (in comparison to the others) and overstate Others. So I conclude that whatever underlying quantity AR is measuring it is does not have the same mean as the other pollsers, hence I exclude it from the average.

    I don’t ask Anthony to remove it from his average – I’m happy to do it myself! :-)

    Also, don’t ignore AR – they might be right! :-)

  39. Barry P you of course have a good point—but for me
    it’s really not which party is ahead, at the moment I
    think that is obvious but it must be also be obvious that they can’t all be correct. Fact is that ICM and You Gov have first class records on results going way back . Three GE’s ago I determined to only take ICM seriously. Incidently I wouldn’t be at all surprised to see
    all the polls showing similar results by March

  40. I think that a pollster who is consitent is more likely most people dont change there mind that quick

  41. @ Jay Blanc

    A Lib/Lab coalition? Heaven forbid. If anything i would say that the Lib Dems will side with Tories if there a hung parliament (which is unlikely). Another 5 years of Labour being in government just turns my stomach.

  42. Apologies I meant another 4 years and not 5 years……

  43. For the sake of JRBrown — quote– “new to polling prediction and reasoning” but has faith in wisdom of bookies——
    I looked up details of the last surprise/close election in 1992 with following result—-
    William Hill betting most seats Lab 1-4 Con 5-2
    LibDem 200-1
    We all knew the result later that day!

  44. Why do other pollsters put in false recall surely people know who they voted for last election

  45. @ David Greybeard.

    I agree entirely with your last comment that all polls could well be within 1 or 2 points of the overall majority/lead, whatever it turns out to be.

    However, it has been almost 5 years since any pollster has been able to be compared with a GE result. Trends change, voters response to questionnaires change. Opinions differ in many ways to what they did 5 years ago. By that I don’t refer to one’s allegiances but to how anyone might reply to the same question now compared to the same question 5 years ago. So, until we can, with 100% confidence say that a pollstser’s methods are wrong, even questionable, we should all accept what we are given and then, Mar/June ‘ish, hold a full blown debate on individual accuracies and reliabilities pollsters. But, even the opinions and decisions made then cannot be set in stone, things will change again and fluidity will reign again.

  46. Not a statistician myself, but any chance someone on here could explain the logic of this comment by Mark Senior up thread.

    Others have pointed out the consistency of Conservative support in recent polls at 40% . Contrarily this should not be happening given the differing methodologies , sampling errors of each poll and is an indication that whatever the actual level of Conservative support it is unlikely to be 40% .

    I may be being thick, but to me, the more consistent the result, the more likely it is to be accurate.

  47. It would have been interesting if Angus Reid had been doing polls in May and June this year because they may have come up with the most accurate prediction of the Labour vote in the Euro elections which was of course 15.7%. Nearly all the other pollsters overestimated their share.

  48. @C.L.A.D.

    My guess is it is an attempt to undermine the MORI poll, which has been widely rumoured on PB to show another single digit lead for Cons. Mike Smithson hints at this in the text when he highlights that the AR poll is more up to date than the MORI one.

    Personally, I’m more interested in the next Comres, to see whether it is more or less back in agreement with the others (excluding AR).

  49. @Andy Stidwell

    I think it was only ComRes (22%) which overestimated the Labour share. YouGov (16%) Populus (16%) & ICM (17%) got the share pretty much spot on!

  50. @Andy Stidwell

    There is a difference between overestimating Labour’s share at the Euro elections by 0.3% (both You Gov & Populus) and 1.3%(ICM), and AR estimating the Labour share (compared to other pollsters) at something approching 6% lower!

    I don’t think many people would disagree with the assertion that Labour tend to get a slightly lower vote than the polls suggest, and the Cons slightly higher. But it ain’t anywhere near 6%.

1 2 3