Friday round up

This morning’s YouGov poll for the Sun has topline figures of CON 31%, LAB 39%, LDEM 11%, UKIP 12%. That’s the third poll in a row showing an eight point lead, so it looks as though the further narrowing at the start of the week was a bit of a blip – full tabs are here.

There is also a Survation poll out for the Daily Mirror which has topline figures of CON 23%, LAB 36%, LDEM 10%, UKIP 22% (Full tabs are here). It’s the lowest the Conservatives have shown in a poll this Parliament, and the closest UKIP have come to them, but that’s largely for methodological reasons. If we go back to my post from earlier this week showing the house effects of different pollsters:

You can see that Survation normally show higher levels of UKIP support than other companies (a good seven points higher than ICM, who tend to show the lowest). As has been often remarked, this is partially because they are the only company to include UKIP in their main prompt, but that probably explains only a small part of the total difference, as there is a more general gulf between the levels of UKIP support that “new online” companies show, and the levels of support that the traditional telephone pollsters and YouGov show.

Part of this could be interviewer effect (though I doubt it – firstly UKIP voters don’t seem particularly bashful or reticent, secondly if that was the reason we’d expect to see YouGov in the same place as other online companies), or it could be a sampling issue of some sort.

229 Responses to “Friday round up”

1 2 3 5
  1. Do we know which company does UKIP’s own polling?

  2. RE – The survation poll. Of some interest, if possibly meaningless:

    “Q10. Thinking back to the General Election in May 2010, can you recall which party you voted for in that election?”

    Con 36.0%
    Lab 28.9%
    Lib 22.9%
    UKIP 3.1%
    Green 2.2%
    SNP 2.0%

    Actual GE results:

    Con 36.1%
    Lab 29.0%
    Lib 23.0%
    UKIP 3.1%
    Green 1.0%
    SNP 1.7%

  3. Anthony
    I notice that you have not included Survation in your list of pollsters (see left). Also, Populus is on your list but not on your chart.

    Perhaps i am mixing up the names (has been known).

    I shall put down the temporary blip in LD down to 8 to Danny Alexander getting his face on TV (in-house LD lefty joke).

  4. Statgeek
    That is ‘really very remarkable’.

  5. Survation last pollster Gipsy Rose Lee has retired, they’ve employed a bloke who dresses like a Cheyenne medicine man who throws coloured sand in the air he’s supposed to be quite good cutting rouge polls to a minimum as we can see.
    Now if they would employ Crofty’s puppies they could rule out any inaccuracy for good.

  6. Howard – haven’t updated all that for *ages*. Populus are not on the chart as they’ve stopped doing regular voting intention polls.

    Not sure why that’s remarkable – like most companies (the exceptions being YouGov (who use party ID instead), MORI and Opinium (who don’t use political weighting)), Survation weight by past vote, so it should always match the actual results of the last election, if necessary with adjustments for past recall.

  7. Mark – No. It is not always public knowledge (not, at least, until general elections when parties have to public invoices!).

    The Conservatives used to use Populus for most of their stuff, but are presumably now using CrosbyTextor
    Labour have in the past used Stan Greenberg’s company, I do not know if that is still the case
    No idea at all who other parties use

  8. Statgeek,

    You have quoted the UK figures for 2010 whilst the pollsters almost always give us GB figures. The GB breakdown for 2010 was :
    Con 37.0%
    Lab 29.7%
    LD 23.6%
    UKIP 3.1%

  9. AW
    No I guessed that but could not resist doing a David Coleman impersonation. You are too young to remember him.

  10. Aren’t there conflicts of interest in polling?

    As in, Conservative newspapers often want to base stories on slipping Labour leads, or public dismay against the EU, or support for Michael Gove etc etc

    What I’m saying is, won’t it be tempting for a new company to just sway polls to their market – as in, create them to sell them?

  11. Howard – while I didn’t spot the reference, sadly I’m not!

  12. @Graham

    Taken in that context, how many of the 52 (reasonable crossbreak estimate of NI population) would vote any of the listed parties?

    Looking at the 2010 election, we can see their vote share:

    Con 0.0%
    Lab 0.0%
    Lib 0.0%
    UKIP 0.0%
    Green 0.014% (0.5% of NI)

  13. @AW

    I was expecting it to be more skewed (per your mention of recall). To be honest I don’t generally pay attention to 2010 data, unless looking at how 2010 voters have changed their vote.

  14. Interesting looking at the weighting by age range

    18-34’s – Survation 389 responses weighted down to 312.
    – Yougov 18-24’s 94 responses weighted up to 217
    25-39 424 responses weighted up to 457

    18-34’s voting intention – 19% Conservative; 15% UKIP; 45% Labour – Survation

    Yougov has the 18-39’s at about 30-33% conservative

    Would it be correct to conclude that Survation’s panel is younger (as they started up doing the big brother type polls), and yougov’s is older?

    So would it be correct to say Survation give a better picture of the younger voter, Yougov of the older voter, due to less error due to better sample size for those demographics?

    Or could it just be that survation better represents the reality television viewer demographic, and that demographic votes differently?

    And looking at local election results, Survation was closer to the mark on vote share, so does that mean reality television teaches people how to vote and gets them interested in voting, so they are over-represented at the polling booth :)

    Well some reality tv stars in parliament should at least add some interest. Not sure they could really do much worse than the lot we have there at the moment :)

  15. Statgeek – in this Parliament pollsters have tended to make very little adjustment for false recall. I’ve been pondering this (for a year or two now, which is very sad) and I keep meaning to post on it.

    ICM and Populus and ComRes all estimate false recall by comparing the actual result of the last general election with people’s recalled vote in their raw samples. Averaging over a number of polls they assume that part of the difference between the figures is sample skew, and part is false recall, and use those figures to come up with “false recall adjusted target weights”.

    In the last couple of Parliaments those false recall adjusted weights have been more Labour than the actual shares of the vote, and the phone pollsters normally had to weight their data to make it less Labour and more Conservative in order to reach their target. The obvious conclusion was that some people were saying they voted Labour when they didn’t, and that raw telephone samples were more often than not biased towards Labour (that’s no criticism – all methods have intrinsic biases – the companies all corrected for this with weighting and produced fine and accurate results).

    Now, roll on to the present situation….

    Currently pollsters are weighting to targets very close to the actual 2010 election results, implying there is little or no false recall. This is because the recalled vote in raw samples is pretty close to the actual shares of the vote, so if they split the difference between 2010 and raw recalled vote, it is naturally very close to the actual 2010 vote.

    In practice this all seems very good, if a sample is almost spot on without any weighting that is a good thing. If you think on it though, it has some worrying implications. False recall has always been there in the past, why isn’t it now? In the past raw phone samples were always too Labour and needed weighting to make them more Tory, why has that stopped happening? Why would they suddenly become better?

    The implication from the current approach taken by the pollsters is that false recall and Labour bias in phone samples have both suddenly vanished at the same time. However, there is an alternative explanation… that raw phone samples are still biased towards Labour and false recall is still there, but in the other direction, so samples should actually be being weighted to a target that is actually slightly *more* Conservative than the real 2010 election result. If the second explanation was true, then the polls would be overestimating Labour support and underestimating Conservative support.

    That would be my conclusion, except, if that was the case, then we would predict that YouGov (who don’t really have to worry about false recall because they use historical panel data from May 2010 to weight, so it doesn’t shift) would show more Conservative results than ICM and ComRes. That is not the case – YouGov’s results actually tend to be more Labour inclined than ICM and ComRes.

    All very odd.

  16. Richard

    I think it’s safe to say no to every question there!

    You can’t safely conclude anything about the make up of the panel from the raw samples (it’s a common enough error, but it doesn’t work like that – the raw sample in a panel poll is not a random microcosm of the wider panel, it depends what emails go out).

    I don’t think Survation mostly use their own panel either, so don’t be mislead by the reality TV stuff (I think they get people from Toluna, a panel providing company, topping up the smaller amount of people on their own panel).

    Sample size does not equate to how representative a sample is – a big sample size can be rubbish if it is strongly skewed in some way or another (I shan’t stretch everyone’s patience by wheeling out the Literary Digest example again!)

    People vote very differently in local elections to how they vote in general elections (and if you ask voting intention for Westminister and local council voting intention, you get different answers), so one should only ever compare local election results to polls conducted of *local election voting intention*. The only local election poll done immediately preceding the local elections this year was by ComRes.

  17. Ok thanks Anthony

    Survation have released a statement attempting to explain the difference between their figures and yougov’s

    “YouGov’s President, Peter Kellner has perfomed an analysis of the effect of prompting, showing that if YouGov used Survation’s methodology, their results would look similar in a detailed piece for those interested here:

    But reading that Yougov statement, it seems to be referring to European elections, not a general election?

    Or is it correct that if yougov prompted for UKIP it would show the same results as Survation once adjusted for weighting?

  18. Odd that the DT doesn’t have an opinion piece about the sudden rise of Ukip, following on from the Ukip is dead article they had a few weeks ago

  19. Richard – I don’t think so, no.

    In Peter’s article he estimates that the effect of prompting for UKIP *in a European Election poll” is 4 points, so certainly wouldn’t bring YouGov’s UKIP scores up into the low twenties… and that’s before you take into the account the point you’ve just made, that polls for European elections do not necessarily reflect polls for Westminster elections.

    Weighting by partyID in 2010 is analogous to weighting by how people voted in 2010, so shouldn’t make any real difference (the ability to weight using data collected at the time, as opposed to weighting by data collected when the survey is conducted, is a different matter and could explain some difference).

    Essentially we know why Opinium and Survation tend to show higher levels of UKIP support than TNS BMRB and ComRes Online (lack of political weighting and prompting respectively), but we don’t know why all the “new online” companies consistently show significantly higher levels of UKIP support than the more “established” pollsters. My guess is still that it is to do with sampling and panel – but that’s something which is largely opaque.

  20. Sorry, survation also had this part explaining the difference, so it is not just prompting, so worth repeating here

    “Why is YouGov The Online Outlier?

    YouGov in particular publishes significantly lower UKIP figures in it’s daily polls (today’s 12% making it an outlier among online pollsters), largely as a result of methodological differences such as its use of Party ID weighting to a 3 year old target, which after all usual weighting has taken place, reduces UKIP’s figure significantly. Other online companies such as Opinium and TNS show UKIP at 19% in their most recent polls, more comparable with the Survation figures, though still with a slight difference perhaps accounted for by prompting (as discussed) or by other minor differences in weighting. ”
    But if you look at the figures before weighting and assume prompting adds only 2%, then that still doesn’t explain the massive difference?

    I think people are going to start losing faith in polls if each company keeps coming up with wildly different figures like this (a look on twitter shows a lot of people asking about this difference), so for the good of your industry you should get together to figure out why the results are so different, and correct if an error in any company, or at least explain better why there are such massive differences?

    The explanations to date don’t seem to add up?

  21. It don’t mean a thing of course, but in UKIPs only by-election outing yesterday,their vote share fell as did the Tories and Labour retained the seat.
    I am tempted to advance the idea that UKIP tend to do disproportionately well in Tory seats\areas.

  22. @ Ewen

    They won these 2 yesterday

    What I see is they do well as the second choice party, they seem to be becoming the Tory opposition in strong Tory seats, and the Labour opposition in strong Labour seats.

  23. Richard – drawing conclusions from a few by-elections is silly anyway (even aggregating large numbers together they are not very useful as a predictor of national support), but even when we do, we tend to ignore parish council elections (they are normally extremely small areas, and often not contested on a party political basis).

    It used to be something of a running joke that if they were doing badly in local by-elections Lib Dems would roll out the result of a parish council election in lower bumpstead where they had won a huge victory over a dachshund.

  24. 23% for the Tories, that’s got to be depressing for them. Wasn’t their lowest ever 21% back in ’95 or something?

  25. Labour have involved the police, because of “serious irregularities” in Falkirk.

    If Len McCluskie were to be replaced for any reason, let’s hope Unite can find someone who can communicate effectively, without falling back on the tiresome and alienating ‘comrades vs the bosses’ rhetoric – someone like Frances O’Grady for example.

  26. Billy Bob

    It sounds like the blairites have been doing more of the cheating than unite, it could end up being completely upside down to what we are expecting

  27. @ Anthony

    I understand why by elections have historically been inaccurate in predicting general elections, but as you keep saying we also need to look at trends?

    And the trend is that UKIP traditionally did very poorly in by elections, but they are now doing rather well comparatively?

    With different poll companies coming out with such widely different results, we need to look at something else to see who is getting it right. As I see it by elections is the only thing other than polls that we can look at, and if you look at those it does look like UKIP is doing better than some poll companies are saying, and I haven’t seen the big drop off in UKIP results in by elections since the May elections that Yougov at least is showing in their polls?

    So that doesn’t make sense to me?

  28. You can squeeze something out of local by-elections if you aggregate up a lot of them, over months, and process the data rather heavily the way rallings and Thrasher do (that is, looking at the swings from when the seat was last fought in cases where the parties contesting the seat where the same and comparing it to provisional national shares at the time it was last fought – then using it to project national shares).

    It ‘s very hard work, we don’t really have the past data to do it for UKIP, and I never see anyone but R&T do this because its hard work. I only ever see people blindly trumpetting results to try and prove what they would like to see be true.

    Even if someone did go to the trouble of crunching the figures, it would be useless to us in telling us the level of UKIP support in a general election because we don’rt yet know how UKIP local support relates to UKIP general election support.

    For example, going back over decades the Lib Dems have always done better at local elections than at national elections (on average by about 7 points), so when the Lib Dems get 17% at the local elections, it doesn’t mean the polls are wrong if they are showing 10. It’s normal.

    With Labour it tends to be the other way round, they tend to do worse at local elections than at national elections, so during the Blair government Hague and IDS used to rack up local election “wins” while Labour were still safely ahead in the polls. It wasn’t because the polls were wrong, just that people voted differently.

    The current levels of UKIP support are new. We have no way of judging the relationship between local and national support. What we do have (we’ve done some polling doing local and national VI in parallel for internal testing) suggests people are indeed more likely to say they’d vote UKIP in a local election than a national one.

  29. @Billy Bob,

    Wow, didnt realise the police were involved now in Falkirk.

  30. Anthony, thanks

    I found an article here from someone who seems to know what he is talking about and has crunched some numbers:

    “So that leaves the question of why UKIP were up 11 points on 2009 in the places they did fight last time. Also in last year’s local elections UKIP averaged just 12% in the 311 BBC keywards they contested, 13 points lower than this year’s average. This is the same average rise as in the (albeit only 41) places were the county council division boundaries this year are the same as the ward boundaries for last year’s elections.

    Polls suggest that UKIP support has risen steadily since the general election from 3% in 2010 to 11.5% now, with a 4% rise since last year. A 4 point rise in the polls for UKIP provides some explanation for a 13 point increase in local election performance, but still leaves a lot to explain. ”

    So that at least would seem to suggest a large difference between poll trends and actual vote trends for the local elections at least for a large set of elections.

    But as you say, it is wait and see if this gets repeated at a general election or not.

    Tories should be very worried, they have no idea if they are at 22% or 32%….

  31. @ Ewen,

    I am tempted to advance the idea that UKIP tend to do disproportionately well in Tory seats\areas.

    I seem to recall them doing pretty well in the great Tory strongholds of South Shields and Rotherham.

    I think Richard not in Norway is right, thought not on the basis of parish council elections. Now that the Lib Dems have been Clegged, Ukip have become the official opposition in the northern and southern Tory/Labour exclusion zones.

    The question yet to be answered is whether anyone would vote for them in a contested seat at a general election. It’s one thing to use them as a vehicle for protest, but quite another to risk getting a Ukip MP or risk letting in a Tory/Labour MP by throwing away a vote on Ukip.

  32. I’d probably vote UKIP in our Parish Council elections (although I can’t because I am out of the catchment area for voting) as I fancy some Purple flowers in the flowerbeds for a change.

    It’d be interesting to know how much work UKIP have put into local elections. I always assumed the Lib Dems did better in local elections because a) they put more work into them and b) made a point of making it less about politics and more about the ‘hardworking local councillor’.

    If UKIP aren’t really working the area much but picking up the votes it makes it a bit of an unknown how that equates with a GE vote other than the assumption that people are using the locals as a protest vote.

  33. AW

    It used to be something of a running joke that if they were doing badly in local by-elections Lib Dems would roll out the result of a parish council election in lower bumpstead where they had won a huge victory over a dachshund.”

    The pups wuffed and larfed at that joke.

    “I thought he just told people off” they said.

  34. SHEVII

    Hardworking local councillors do bugger all, and I should know as I was one.

    Actually I did quite a lot a hard clerical work (is that real work?) on planning committees but that was because I was interested in it. Most of my colleagues turned up to planning meetings with noticeably un-thumbed agenda papers.

  35. Paul Croft
    I’ll give the LD site ALDC its due though. It’s the best local election result site around.

  36. Howard – I’m with you, development control is far and away the most interesting bit of being a local councillor.

  37. MrNameless

    As a Tory party member let me assure you nobody in the Tory party is in the least bit worried by a Survation poll for the Daily Mirror or any resent Survation poll come to that.
    We always look at the broad sweep of polls none of which show Ukip consistantly with such a high percentage of VI.
    If one part of your VI seems out of step with other polls then you can have no confidence in the rest of the data for that polling company is correct.

  38. Resent should be recent

  39. Good Afternoon All.

    Wife and children home from the boarding school as the school year ends.

    Labour not doing as badly as it looked last week.

    Ed M could turn the Unite thing into a positive leadership move.

  40. @ Howard

    Yes- that was why I put it in ‘ ”s!

    I find it an annoying aspect to politics that the more times you get your face in the papers and the more leaflets you put out the more ‘hardworking’ you appear but in fact this doesn’t necessarily equate.

  41. Turk – calm down dear!

    If Labour was scoring 23% you would be crowing. Does anybody here doubt that?

    Would you agree that if the latest Tory onslaught against Labour and the unions does not dent Labour’s poll lead significantly, and for a period of months, then the Tory-led coalition is heading for a huge defeat in 2015?

    Personally I think the bookmakers have got it right. The Tories are a long shot notwithstanding their huge media support.

  42. @Turk,

    Absolutely spot on. No surprise The Mirror chooses that poll. Sigh..

  43. @Turk

    We always look at the broad sweep of polls none of which show Ukip consistantly with such a high percentage of VI.
    Well, we have opinium

    Scroll down to graph, UKIP consistently high since March:

    And TNS: – no UKIP drop:

    And Survation…

    So perhaps you should be a little more worried?

  44. Not sure where the Labour lead is now after what seemed to be a narrowing last week.Don`t think the Union issue is interesting to the public,hence don`t think it will shift public opinion.

  45. @TURK

    “As a Tory party member let me assure you nobody in the Tory party is in the least bit worried by a Survation poll for the Daily Mirror or any resent Survation poll come to that.
    We always look at the broad sweep of polls none of which show Ukip consistantly with such a high percentage of VI.
    If one part of your VI seems out of step with other polls then you can have no confidence in the rest of the data for that polling company is correct.”


    A dangerous view to take if you are keen to avoid any chance of disappointment.

    Sure, there’s a good chance the broad sweep of polls is correct, and Survation is noise. But if one takes a scientific approach, then quite often it is the anomalies that don’t seem to fit the “broad sweep” that are of interest and at times wind up ushering in a big change in thinking.

    Unless you have good reason to categorically rule out their polling as noise, caution might be advised. Survation prompt for Ukip, others don’t. Which could be worth a few percent, and a few percent could make all the difference in a tight election…

  46. Iananthony james

    If Survation showed Labour on 23% then my opinion of Survation polls would be the same, except my opinion would be agreed by most who post on here.

  47. Turk, I have a comment in moderation, but look at Opinium and TNS, they are showing the same as Survation with no UKIP drop since May.

    I remember reading of the Tory mp’s shock at UKIP support at the door when they went to campaign in Eastleigh, and hearing about more shocked Tory councillors in May. Unless you face up that you have a problem, you can’t start to address it.

    But unfortunately we can’t tell if you have a problem or not from the polls :)

    I guess you need to go door to door to get a real idea?

  48. @ Shev II,

    In the last round of local elections didn’t Ukip get a fair number of votes in seats where they didn’t even put up a candidate? I don’t think we can attribute their surge to their activists’ enthusiasm.

    @ Turk,

    I’ll agree with you if the others won’t- that was a rogue poll. Whatever Ukip’s real VI is, YouGov is accurately measuring the trends, and the Tories are going up and Ukip are going down, not the other way round.

    The only danger here for the Tories is that the press are irresponsible and desperate to fill column inches. Any poll that shows Ukip overtaking the Tories will fill the media cycle for a week just because it’s something to talk about, and once that happens people who really should know better will go into hysterics and make the situation ten times worse. (Witness the hysteria in the Labour camp over God forbid, two polls showing a 5-point lead. Even Peter Kellner got sucked into it, and he’s a pollster!) Since the Tory backbenchers are prone to hysteria anyway, it could start a damaging chain reaction.

  49. Gary O’Donoghue
    Chief political correspondent, BBC Radio 4
    There are a million questions still to be clarified about what exactly went on in Falkirk, but there is no doubt that in sending their evidence to the police, Ed Miliband and Labour have chosen to escalate this row with Unite to a stratospheric level.

    -Not exactly the actions of a puppet and it will be difficult to portray it as such?

  50. @Steve,

    Surprisingly, I agree with you.

1 2 3 5