Time for a round up of Sunday’s polls, with new stuff in the Sunday papers from YouGov, Survation, Opinium and ICM.

YouGov’s weekly poll for the Sunday Times has topline figures of CON 33%, LAB 33%, LDEM 7%, UKIP 16%, GRN 6%. The poll also asked a series of questions about how people would vote with different Labour leaders. In a control question asking how people would vote if the leaders remained Cameron, Miliband and Clegg the answers were CON 33%, LAB 31% (so the effect of reminding people of the current party leaders still seems to produce a slight positive Cameron effect or negative Miliband effect). If Yvette Cooper were Labour leader the position would be the same, a two point Conservative lead. If Ed Balls was the leader it would be worse, a three point Conservative lead. In contrast with Alan Johnson as leader Labour would be two points ahead (CON 31%, LAB 33%.

I’ll give my usual caveats about questions like this – people are answering them when on very little information, they don’t know what policies or priorities those alternative leaders would set, how the media would react to them and so on. In the same poll, YouGov found that only 42% of people think they could recognise Yvette Cooper from a photo… if you don’t even know what Yvette Cooper looks like, I’m guessing you don’t have a thorough understanding of what she would prioritise as Labour leader. It’s a response based on a very crude impression of those potential leaders based on what tends to be the very limited public awareness of opposition politicians. Nevertheless, those crude first impressions count, so it’s a good sign for Alan Johnson.

Survation also had a new poll with topline figures of CON 29%(+2), LAB 34%(+3), LDEM 6%(-3), UKIP 23%(-1), and they too asked a series of hypothetical voting intention questions with different Labour leaders. In the Survation poll they displayed a biography and played a video clip of each potential leader and asked people questions about them before the questions. This allowed them to include people with extremely low public awareness like Chuka Umunna, though does of course rely upon the choice of biogs and video clips (given bias is often in the eye of the beholder, choosing clips that even those who don’t like the eventual results think are fair is incredibly tricky). The control question with Ed Miliband had a Labour lead of 4 points. In the Survation poll Yvette Cooper did worse than Miliband (neck and neck with the Tories), Andy Burnham just the same (4 point lead), Alan Johnson and Chuka Umunna did best – both extending Labour’s lead to 8 points. A voting intention question asked after video clips of Labour leaders is obviously skewed towards Labour, but it’s the relative performance between the different leaders that counts, and again it’s good for Alan Johnson, and now also for Chuka Umunna.

Meanwhile the fortnightly Opinium poll for the Observer has topline figures of CON 29%(-4), LAB 32%(-1), LDEM 9%(+3), UKIP 19%(+1).

Finally there was an ICM poll in the Sunday Telegraph. As usual with ICM/Sunday Telegraph polls, this asked the public to predict vote shares rather than ask people how they would vote themselves. The average response now has the Conservatives getting slightly more votes than Labour.


281 Responses to “Sunday polls and alternative Labour leaders”

1 2 3 4 5 6
  1. @ Allan Christie

    That reminds me, the Batmobile is due for its MOT this week. Might struggle on the exhaust emissions this time.

    It’s looking increasingly likely that Scotland will hold the key to the GE result next time.

  2. Roland
    I think you missed my point. You gave a list of why no one in their right mind could possibily vote Labour. I was simply suggesting many would feel that way about voting Tory (more in fact according to polls of ‘will never vote for’).

  3. Martin

    Ah, I see what you mean, I thought you were referring to the standard error of a sample. e.g. people thought the cat party would record 38% +/- 5% of the vote.

    A common goodness of fit test (how far out was the expected from prediction) is a chi squared test (go google the mechanics) but essentially in your example the cat parties observed score would be 1.0 and the others would be 0.

    You’d need to do a similar test (based on UNS or a prediction model of your choice) across the range of possible results according to polls to compare the two. Your swing model and how you average multiple polls will be a integral part of making a prediction.

    You can then compare the two predictions and determine which model made the best prediction.

    You could also do a similar test based on seats if you asked how many seats would party x get in the election and compare with polling predictor of choice.

    It’d still work

  4. BANTAMS

    Ooops Sorry I meant BANTAMS and not BATMAN.

    If the polls stay as they are north and south then I think the SNP will at least hold the balance of power.

    So many different factors though..Is it wipe out Lib/Dems? How many Kippers will be elected? Tactical voting between Tory and Labour voters backing a UKIP candidate best suited to knock out a sitting candidate, The Greens could snatch a few…It’s going to get very bumpy and exciting.

  5. OLDNAT

    The trebling of party memberships is a headache most other parties would love to have.

  6. @Roger Mexico

    Good link on ‘Predicting Elections’. Only so far into it. One thing that occurred to me:

    “In 2010, the most accurate poll had an average error of 1.25%, but the worst polling average error was 3.25%, a figure outside standard statistical tolerances.”

    Given the way the polls are shaping up to deliver dead heats, these margins of error might just make or break a few reputations in 2015, if things are this close.

    Will read the rest tomorrow.

  7. @ ALLAN CHRISTIE

    Thought I’d been found out! There’s a great clip on YouTube of the Maryland police pulling up a Lamborghini & looking in disbelief when someone in an amazing Batman outfit gets out of the car.

    I really can’t call what’s going to happen at the GE. Unless something major happens I really can’t see anything other than another coalition.

  8. “Like ability is one thing, people just don’t see Ed as a PM.”

    —————

    Well, according to the article…

    “It turns out that likeability is closely associated with other desirable traits that a successful leader needs, such as being seen as competent, decisive, in touch with ordinary people and honest. More to the point, it is a powerful predictor of voting intentions and therefore a good guide to what people might do in the general election.”

  9. @Bantams

    Vicar Lane, not Boar Lane!!…

  10. Statgeek

    I think pollsters are going to duck predicting SNP as a %age of Scottish vote. It’s easy enough to avoid by making a GB prediction of say 3.2% the real answer comes out as 4.2% and they smugly claim to be within 1%. On a Scotland only basis the prediction would be a massive 10% out.

    I think UKIP are going to give pollsters enough of a headache.

  11. @ Carfrew

    Being a Bratfud lad I always got them mixed up, Boar Lane joined onto Vicar Lane and vice versa. We used to walk down Boar Lane from the railway station, seem to remember it was near the County Arcade? What’s there now?

  12. Alan

    “I think pollsters are going to duck predicting SNP as a %age of Scottish vote”

    But anyone conducting Scotland only polls can’t duck out.

  13. @Bantams

    I think the County Arcade is what became the Victoria Quarter, with Harvey Nichols and stuff…

  14. @Alan

    So in our example (50%Dog, 30%Cat, 20%Mouse, Cat wins) we’d construct a contingency table that looks like this

    ————————————-
    Row 1 | 0.5 | 0.3 | 0.2 |
    ————————————-
    Row 2 | 0.0 | 1.0 | 0.0 |
    ————————————-

    And calculate the statistic using, say, Fisher’s exact test.

    OK, gottit.

    (PS: Martyn, not Martin)

  15. @Bantams

    Oh, if you mean what replaced the Duchess, it became a Hugo Boss. Great shame…

  16. Oldnat

    You think there will be many of those in the week leading up to the election?

    Seems like an opportunity to be way out due to methodology being unable to cope with the sheer scale of the swing due to completely incorrect assumptions about don’t knows being reverted to a previous election. They might work with most swings we see, but those don’t knows might end up overestimating the Lab vote if they split closer to the 2015 vote than 2010 than assumed.

    It’ll be a good chance to compare how those different methodologies stack up faced with such a big discrepancy. The difference between 52% and 43% for SNP? (I forget what the Yougov figure was) shows that pollsters have significant differences in opinion about how much of a reversion to 2010 should be made and it results in massive differences.

    They can’t both be right.

  17. Alan

    If polls are being paid for, pollsters will oblige! They are commercial firms and profit is the bottom line.

    Reputational damage from refusing to take a commission would be significant..

  18. Anthony, this interests me, although I’m not sure how much you could reveal. Would a pollster (not a specific one but as a rule) ever refuse to take a polling commission and why?

    I could perhaps see them doing so if the organisation attempting to commission it was very notorious and was transparently trying to do something with the result that would damage the pollster’s reputation.

  19. Martyn

    Yes, it’d be a little more complex to do a similar thing with polling data predicting %age of vote as you’d need someway to convert the VI into %age chance of winning an election and your choice how to do that becomes part of the prediction model.

    One way once you have a set of polling data would be to take a random “sample” from a distribution based on the sampling error of the polls (taking into account the VI scores for all the parties are negatively correlated) running it through your swing model and recording the outcome. Then repeat a million times (or so) and count how many Lab OM, hung with lab largest, hung with con largest, Con OM you arrive at and use those fractions as your expected data.

  20. @Alan

    Thank you

  21. @Alec
    “Meanwhile, Cons are on 29% in two polls and 33% in another. Two sub thirty polls in one day! What’s the story here?”

    On those grounds we ought to see a contrived story from “sources” that IDS is planning to challenge Cameron for the Conservative leadership. IDS at least managed to poll consistently in the 30s.

  22. Oldnat

    True. Although being 10% out for the largest party will also do damage. At least they’ll have an opportunity to correct their methodology. I guess a lot will hope the call never comes. (Although the significance of the SNP on the overall result means it’s likely to be of great interest.

    Statgeek

    It just occurred to me that with so many pollsters one prediction falling outside statistical bounds might actually be less significant. I’d almost expect at least one pollster out of 10 to fall outside 95% bounds. (40% chance)

    Testing the hypothesis “Every one of the pollsters failed to measure the election to within reported MOE” might be statistically unproven.

  23. @Alec
    “Something of a collective madness here. Talk of Alan Johnson taking on Ed is completely fanciful, and I’m gobsmacked so many otherwise sane people bother to waste ink on the idea.”

    And so the next thought would be why is it happening then?

    And your answer might be he’s being used as an involuntary stalking horse.

  24. @Rolandgatinois
    “I would say that Labours performance in the last GE does want some getting over and one 5 year term in opposition is unlikely to be sufficient”

    That might be true under normal circumstances but the Cameroon modernizer strategy was so spectacularly wrong* it’s not normal circumstances.

    *At least in the context of the end of the credit bubble.It might have been different if they’d got to try it during the credit bubble years 1998-2008 when people were caught up in the illusion things were going well.

  25. Martyn

    It’s pretty much what electionforecast do, although they have a “better than UNS model” taking local variations and ashcroft polls into account and they try to predict the reasonable ranges of where the parties will end up based on a model that involves some reversion based on historical trends.

    Of course 2015 could end up bucking all those trends with LD, UKIP and SNP all experience extreme shifts since 2010.

  26. @rolandgatinois

    “The fact that Labour has had a 5 point lead, dropping to a 3 point lead, dropping to a 1 point lead over the last months, in my view proves very little, (except their lead has vanished). Nothing to show for Tory austerity…”

    I think you’re right it’s nothing to do with the Cameroon’s attempt at austerity – as shown by Con flat-lining – but that begs the question what is it about then?

    The political class seem desperate to agree it’s purely about Miliband which provides a neat way of side-stepping the real causes: immigration (for some) and not being leftie enough on anti-austerity (for others).

    “Tory austerity…”

    As an aside the comical part of “austerity” is that deliberately driving wages down is making the problem worse not better (as you’d expect) but because of Brown’s tax credit scam it’s not as deflationary in the UK as it is everywhere else so Osborne is getting the credit for Brown’s scam softening the blow of Osborne’s attempted austerity drive while at the same time the cost of the increased benefit bill is trashing the public finances and Osborne’s debt reduction targets.

    It’s enough to make a cat laugh.

  27. Alan

    YG (& others) weren’t too far out with the “easy” bit of the 2011 Holyrood election – the constituencies.

    Back then on UKPR, most people (including me) couldn’t really believe the polling numbers that suggested such a huge overturn of Lab -> SNP seats in the constituencies, but the pollsters were right.

    Today we seem to be seeing the same thing. Few, including the SNP, are saying that current polling will stay at this level till May, but if it did ……

  28. ” Few, including the SNP,” -> Few, not even the SNP,

  29. Oldnat

    I forgot who did well or not in that election tbh. Going back on the site it seemed Ipsos-Mori was in fair agreement withYougov (I am remembering the right pollster who predicted 52% for SNP?, my short term memory is badly affected)

    Now there is a large discrepancy between the two, and the difference is very likely to be methodolocigal (barring the real result is 48% and both got really unlucky on sampling)

    If IM were significantly higher than Yougov in the Holyrood elections, then one might think that in the regime of major swings IM overestimates the size of the swing. I really wonder what the real picture is in Scotland, common sense tells me 43-45% is more likely as it’s a more conservative swing. (Not the party obviously, a conservative swing in Scotland would really confuse everyone.)

    I just have a nagging doubt that maybe 52% is the real picture and until the election there is no way to divine whose methodology is right, both have worked in the past (unless I’m missing a major change in methodology from IM).

  30. MrNameless – “Would a pollster (not a specific one but as a rule) ever refuse to take a polling commission and why?”

    Of course. Companies would probably refuse to work for a client if the reputational damage of being associated with them was too great – for example, I expect most companies would refuse a commission from the BNP or National Front (though I’ve never personally been in the situation). They turn it down when they can’t do it well – I’ve turned down countless requests for constituency polls because YouGov’s panel doesn’t have enough people in individual constituencies. They also turn down commissions when the client refuses to accept unbiased question wording – it’s very rare, as most clients will accept the pollster’s advice (after all, it is what they are paying for), but some do come along who won’t and after a while you reach an impass and part ways.

  31. @Mrjones – “And so the next thought would be why is it happening then?

    And your answer might be he’s being used as an involuntary stalking horse.”

    As far as I can tell, no one inside the Labour is pushing Johnson’s name as a possible leader. That would be the stalking horse option, voluntary or no.

    The AJ talk is entirely from the media, and is entirely and utterly barking. Decent man though he seems to be, AJ wasn’t a shining success at his major front bench role, and the only reason the media tout him is because of his upbringing.

  32. Anthony,

    Interesting on the tangle over clients wanting to use “Bias” questions.

    There was a report earlier this year by, I think, Reform Scotland on the future of Local Government in Scotland that claimed support for raising local taxation on the basis of a poll they conducted.

    As I recall the question asked was along the lines of;

    “Would you support a rise in Council Tax if it lead to better services”.

    I thought that was a bit dodgy as the “better services” bit skewed the whole result”.

    Oh and it wasn’t YouGov who did there poll!

    Peter.

  33. Anthony,

    I’d be curious of your views on the University of Essex’ likeability polls. The polls suggest that there is little difference between the party leaders, and they claim that the likelability scores are a better predictor that who’s the best prime minister…

  34. If you look at the papers, and listen to the news, the silence says that the attack dogs seem at last to have started to move on this morning. Even for those papers totally out to get Miliband, there is a limit. You can only sustain for so long stories based on comments from shadowy unattributable “sources” (that can of course be made up or at least greatly exaggerated), when no-one is prepared to (or indeed able to, if made up) go on the record to give them any credence. I suspect that carrying on now would further diminish those papers’ credibility, as well as risking a backlash leading to a sympathy vote for Ed Miliband. In Labour circles that rallying around in the wake of another reminder of the excesses of the right wing press has already started. From the right wing press perspective, the danger would be that the wider public might start to see through these attacks.

  35. @Christian Schmidt

    I believe this is the “likeability” poll to which you refer: http://theconversation.com/voter-survey-shows-miliband-panic-is-overblown-33975

    There was some discussion on it upthread.

  36. When assessing the import of ‘likeability’ polls one must be careful to differentiate between independent correlation and association.

    It is hardly surprising that positive leadership qualities (empathy/intelligence etc) associate with likeability but that doesn’t mean that assessments of likeability independently correlate significantly with assessments of leadership.

    I am sure even many Conservative supporters think Miliband is more likeable than Cameron but think he would be a worse leader.

    Apparently many Conservative backbenchers detest Cameron.

  37. @PhilHaines

    “From the right wing press perspective, the danger would be that the wider public might start to see through these attacks.”

    I think the danger isn’t so much what the readers of these newspapers think or believe, and there’s some evidence that most papers preach to the already politically converted, it’s their ability to influence the wider news agenda, causing TV and Radio news bulletins and political discussion shows to follow the story. These programmes have much greater capability to influence the political narrative, watched and listened to as they are by vastly more people than read right wing newspapers.

    I should imagine the editors of these newspapers are well aware of that too.

  38. Christian – interesting.

    From memory (I’m at home for a hospital appointment and their book is on my desk at work), Whiteley, Sanders, Clarke & Stewart use the likeability rating in their 2005 and 2010 models based on the BES data, so likeability is the measure of leadership that does have an influence on the vote. IN 2995 Whiteley et al found it was pretty representative of a wider basket of measures of leadership like competence, trust and so on. This doesn’t surprise me – we might be sniffy about such “soft” measures as how likeable a leader is, but I think people ascribe lots of other positive values to people they like, and ascribe a lot of negative values to people they dislike. So it should be important.

    However, thinking about the leaders at the time Whiteley et al did their work likeability and perceptions of competence were quite close (Blair led Howard on both, Cameron led Brown on both). Now we’ve got a slightly different situation – Miliband’s poor ratings seem very much driven by people not thinking he is up to the job of PM and not by people disliking him personally, so we’ve got an interesting situation where there is a big gap between the leaders in terms of perceptions of their suitability for the role, but a much more modest gap in the leaders’ personal likeability.

    That could help explain why Miliband’s poor ratings haven’t stopped Labour being ahead, because likeability actually matters more than competence and he does OK there. Or it could end up mattering more come the election and we’ll have to reassess which leadership ratings really matter.

  39. Alec
    “The AJ talk is entirely from the media, and is entirely and utterly barking. Decent man though he seems to be, AJ wasn’t a shining success at his major front bench role, and the only reason the media tout him is because of his upbringing”

    Well said. It seems that the media have forgotten the AJ remark “pick up a primer in economics for beginners”or just maybe they hoped that they could use it against him in their forlorn hope that he became Leader of the Opposition.

    Nice man he may be but LotO – no chance.

  40. Anthony,

    “In 2995 Whiteley”… My doesn’t time fly when your having fun… Hope the hospital was only something trivial, like;

    Doctor Doctor I keep getting my dates mixed up… Oh it’s quite common if it persists come back yesterday!

    Peter.

  41. Populus:

    Con 34 (+1)
    Lab 36 (+1)
    LD 8 (-1)
    UKIP 13 (-1)
    Greens 4

  42. Interesting Top 10 from Populus:

    Ebola
    EU bill
    Death of Lynda Bellingham
    Islamic State
    Virgin Galactic crash
    American mid-term elections
    Immigration
    Hong Kong murders
    Football
    Tower of London poppies

    https://twitter.com/PopulusPolls/status/530737410708348928/photo/1

    Seems like ordinary people living ordinary lives have little time for politician tittle tattle.

  43. BANTAMS
    @ ALLAN CHRISTIE
    Thought I’d been found out! There’s a great clip on YouTube of the Maryland police pulling up a Lamborghini & looking in disbelief when someone in an amazing Batman outfit gets out of the car.

    I really can’t call what’s going to happen at the GE. Unless something major happens I really can’t see anything other than another coalition
    ________

    LOL well I suppose the equivalent over here would be the MET pulling over a Reliant Robin and out pops Vince Cable.

    It’s looking like another collation or minority gov. If the Libs go into wholesale meltdown then for Labour and the Tories there aint going to be a natural bedfellow to share the bed with.

    If the SNP to rocket up to 40 plus seats I still can’t see them forming a collation with anyone because they would had seen the damage Plaid and the Libs have suffered.

    My bets are still on a minority Tory gov.

  44. I don’t know if people are following events in Northern Ireland where a coalition has pushed through major cuts. Simon Hamilton has emerged as the star – suspect we will hear more of him.

  45. After the last by-election survation showed UKIP was 3-4 points away from coming second in a GE national poll. If they win the next by-E and DO poll 2nd how would that change the game?

  46. “It’s looking like another collation or minority gov”

    Somehow I have a suspicion that collation may indeed be an appropriate word.

    Bramley – thanks for Populus. Conservative looks high.

  47. Howard (a few threads back)

    If anyone has discovered what Table 4 and Table 5 of Survation are, seeing they both have identical descriptions, kindly let me know please?

    Apologies for not replying before but it’s to do with reallocation. If you look at the bottom of page 2 of the tables for Survation’s poll for the Mirror:

    http://survation.com/wp-content/uploads/2014/11/Voting-Intention-Poll-Tables-6-November-2014.pdf

    they say that as “an additional weighting step” they add back in voters who say they were undecided or refused to give a preference “based on a factor of whom they voted for in 2010”. Table 5 should be Table 4 with these extra people added on. ICM and Ashcroft do similar things with their telephone polls to pick up ‘shy’ voters, but don’t publish a separate full table, just the adjusted percentages.

    ICM normally allocate 50% of these voters ‘back’ for Con, Lab and LD only, but pollsters can vary. Populus used to use 50% for Con and Lab but only 30% for LD for example.

    I actually think it’s a little dodgy to do reallocation with an online poll in the same way – the ‘shy’ factor is going to be less because there is no human interaction and respondents have greater leisure to consider their replies – one of the reasons why DKs are usually lower in online polls. But I thought I’d see what ‘factor’ Survation were using for their online poll. And that’s where it got strange.

    I looked at the difference between the two tables in number of voters as against the number of 2010 for each Party who were now undecided according to Table 3[1]:

    Con T5-T4 = 480 – 420 = 60 (2010 now Undecided 53)

    Lab 539 – 478 = 61 (47)

    Lib Dem 156 – 119 = 37 (62)

    UKIP 412 – 377 = 35 (All others 16)

    SNP 82 – 75 = 7

    Green 58 – 49 = 9

    Others 21 – 19 = 2

    The Lib Dem reallocation is possible if high at 60%, but everything else is over 100%. I may have miscalculated or missed something but it’s very odd. It may be that they are reallocating undecideds they don’t have a 2010 vote for or the tables may be showing the wrong things.

    [1] These tables are all said to be “Weighted by normal weighting and likelihood to vote”. Technically ‘refused’ voters should be included too, but nearly all those who refused to give their current VI seem to have refused to say who they voted for in 2010 as well.

  48. @Alan, Martyn.

    That’s exactly what I’ve been doing, but you hit a few pitfalls along the way. For instance, you have to normalise the random deviation so it’s a zero-sum change.

    And it’s hugely tempting to build a “better than UNS” model, by trying to apply regional data. The problem is you’re just putting your thumb on the scales, in ways you figure are appropriate. You really really really have to have a good set of assumptions for doing that, and I don’t think we have reliable consistent and multiple sourced regional data for 2015. The only region I’m special casing is Scotland.

    Follow that on with the decision on if you predict “an election tomorrow”, or predict for election day. And if you do predict for election day, how strong do you calculate momentum, and how much do you weight it against “revert to mean”?

    I’ve already set out that I think electionsetc prediction model fails because it ignores momentum entirely, and just applies a reversion to mean. This is very flawed, because the concept of polling momentum is very well established. And Inertial force builds up as you approach election day, that is moderated not dictated by revert to mean.

    electionforecast is… a weird model. They are doing a lot of things, that have sound reasons for doing them, but do them without good data. The pre-constiuency adjustments are done on tiny numbers per constituency, from raw unweighted polling data. And no, it does not actually run the elections in simulation, it does bayesian inference to make an estimate of the probabilities of an outcome. The problem with that is that you need to have a reliable and relevant data set for it to work. Which means they’re trying to use bayesian inference based on the small data set of past elections, which may not be relevant data to this election anyway. They also suggest they have an entire year long window of averaging for their polling data, which certainly smoothes over the data, but might miss the crucial point that polling does move.

    My model, http://ukelectiontrend.blogspot.co.uk , actually does run the Monte Carlo simulation you talk about.

  49. AC
    For a Tory minority Govt to be viable there would have to be Opposition MPs who would refuse to support a Labour No Confidence motion. Who would they be?

  50. Populus crossbreaks for Scotland – SNP 36 Lab 30 Con 19

1 2 3 4 5 6