Two new polls today – the daily YouGov poll for the Sun and the monthly Survation poll for the Daily Mirror.

Survation in the Mirror have topline figures of CON 28%(-3), LAB 34%(+4), LDEM 10%(+3), UKIP 19%(-4), GRN 4%(+1). Lots of sharp changes there since their previous poll, but usual caveats apply – the Tory lead in Survation’s previous poll was rather unusual in itself, today’s large Labour lead also unusual, hence the large changes from one to the other. Note also the drop in UKIP support – Survation consistently show the highest UKIP support, so while 19 is large compared to other pollsters’ figures, its a notable drop from Survation.

Meanwhile the daily YouGov poll for the Sun has topline figures of CON 35%, LAB 33%, LDEM 6%, UKIP 14%, GRN 7%. A two point Conservative lead from YouGov, the first time they’ve shown that for just over a month.

The bigger picture remains the same. The Conservatives probably haven’t moved ahead, or we’d be seeing that across most of the polls, when actually they are averaging out at a tiny Labour lead. Neither is there is big swing to Labour, or we’d be seeing that across most of the polls, when actually they all just seem to be showing normal variation around the margin of error. In terms of the Labour vs Conservative race, 2015 so far has been largely static. The only trend that may be meaningful is the drop in UKIP support.


Now that Survation have published their monthly poll we can compare UKIP’s January and February scores across all the pollsters (I’ve taken an average for those companies who publish more than once a month). There does seem to be a pretty consistent fall in UKIP support, perhaps slightly obscured by the fact that the most frequent pollster, YouGov, shows one of the more modest drops and the second most frequent pollster, Populus, changed their methodology at the start of February in a way that increased UKIP support.

359 Responses to “Latest YouGov and Survation polls”

1 2 3 4 8
  1. @Richard

    Thanks. Am used to seeing Green Party leaders at 50% “Who Is That”

  2. @Unicorn

    When you do a linear regression and come out with a p-value…what’s your null hypothesis? Is it ho: change=0 or change<0?

  3. OK, just thinking out loud, here…

    Unicorn’s point about not throwing away all the intermediate points is a valid one, *but* is this answering the wrong question. We have a poll reading for today, let’s call it x. We know the poll reading on January 1st, let’s call it y. We know the poll standard deviation. Is it as simple as “is today’s reading within onepointwhateverx times sd of January’s reading?” if yes then it’s not statistically significant.

    Let’s strip it further. Let’s denote a drop since Jan 1st as Y, and a rise/no change since Jan 1 as N. So we have a series that looks something like this: Y,N,Y,Y.N,Y,N,N,Y,…….N,N,N,N. What’s the statistical test that deals with runs of numbers, anybody remember?

  4. Youguv, possible start of swing back? Realise its only one poll but if it happens it has to start somewhere.

  5. @Pete
    Youguv, possible start of swing back? Realise its only one poll but if it happens it has to start somewhere.

    Survation possible confirmation that swingback won’t happen.

    Just as likely

    “All I can is Fruitcakes”

    Not only in the UK, however, many who are not getting much sustenance from the bread and butter of the main parties will choose the Marie Antoinette alternative.

  7. All this talk of ‘the oxygen of publicity’ seems to me a shoddy metaphor. Without oxygen you’ll die pretty quickly, whereas without publicity a party will gradually wither. So the calories of publicity is a better image.
    I was thinking maybe of the Bacon Butties of publicity but perhaps Fruitcakes are more appropriate in the circs.

  8. Latest yougov big drop in dont knows and labour below 20percent in scotland..

    Expect reversion to trend tonight.

  9. A timely set of caveats Anthony.

    Conservative leads have been fleeting, ephemeral things. I expect this morning’s will be the same. Onwards to your next weekly review.

  10. Today’s YG Scottish crossbreak

    SNP 48% : Lab 19% : Con 13% : LD 3% : UKIP 8% : Grn 8%

    Mean of last 20 YG Scottish crossbreaks

    SNP 42% : Lab 26% : Con 18% : LD 5% : UKIP 4% : Grn 4%

  11. Colin my paradox is that it would be nice for me if my long-term prediction of UKPR Average cross over during March occurred but that as an LP member I would love to be wrong.

  12. @ Martyn

    Question 1: Ho is Beta = 0

    Question 2: it is the Wald-Woflwitz Runs test I think you are referring to. But various non-parametric trend tests would be more appropriate for looking for progressive trends. (The Runs test look for s pattern of any kind…)

  13. Jim Jam

    “my long-term prediction of UKPR Average cross over during March occurred ”

    The portents must be good. There is a solar eclipse on 20 March, so doom must befall the world.

  14. After last night’s discussion I thought I would look at what weights each pollster is using for social class versus UKIP house effects.

    UKIP voters are predominantly in the C2DE social group, so this makes a clear difference to their score

    Pollsters sorted by ratio of ABC1 to C2DE

    Highest group
    ICM/Yougov 57% ABC1 to 43% C2DE

    Middle group
    Ashcroft/TNS/Comres Populus 55% ABC1 to 45% C2DE
    Mori 54% ABC1 to 46% C2DE

    Lowest group
    Survation 41% ABC1 to 59% C2DE (But it is not clear if Survation actually weight to social class as it moves around a bit – this was the Feb poll reported split)

    Looking at the house effects post

    Lowest UKIP scores – Yougov/ ICM (highest group above) and Mori/Populus (middle group)

    Highest UKIP scores – Survation (lowest group above)
    TNS/ Opinium/ Comres(O) (middle group above)

    So at least for Yougov/ ICM and Survation, I think their house effects for UKIP can be explained to a large extent by their different social class weightings.

    It would be interesting to understand why different pollsters use different weights here – this document seems to say the middle pack is the correct distribution for the whole population at least.

  15. Richard


    As to the last link, it just talks of the 2011 Census. It might be E&W only, or GB, or UK depending on which censuses they used.

    It might not make much difference, but I wonder if they all use a GB dataset, or if there is variation.

  16. @Old Nat

    You give the average of the last 20 YG Scottish cross breaks, which is useful to those of us, such as myself, interested in such things. I am sure you have better things to do with your time, but what about other organisations than YG?

    I am acutely aware that like needs to be compared with like – with the additional caveat that some ‘houses’ change their methodology every now and again, thus adding more uncertainty and confusion to the picture – and some publish less frequently than others, but do we have enough information to produce figures similar to your YG ones for other other ‘houses”, in order to see if there are verifiable trends?

    There are such wild swings on the small ‘regional’ numbers used in GB wide polls – e.g. the Tories ahead of Labour in one Scottish cross-break on Monday!

  17. @Richard

    Turnout – richer folk tend to vote poorer folk not so much. So the calculation should be social class weighted by turnout. So YouGov/ICM probably more accurate. And you may have figured out the UKIP discrepancy convinced me, which means UKIP are probably at the lower end of their scale

  18. JIM JAM

    I’m rooting for you :-)

  19. @JimJam & Colin

    And it isn’t just a matter of UKIP/Tory and how that affects Labour’s chances. There is the additional question of how this English question is seen north of the border. What are Labour’s chances of regaining the vote of many who, at present, still seem to be intent on voting SNP, if they look less likely to win against a stronger Conservative Party?

    And if Labour look less likely to win, does that not also give a boost to the Greens in E&W?

    We’re in for a very long ‘run up’ to May 7, and a very grueling few days after that as well.

  20. JOHN B

    Yes-so many variables. They make forecasting a pretty pointless excercise one imagins.

    re UKIP, I was just looking at their regional crossbreaks in this morning’s poll. They are strongest in ” Mids & Wales”. So much for the popular characterisation of their strength faded east coast seaside towns.

  21. @John B –

    “And if Labour look less likely to win, does that not also give a boost to the Greens in E&W?”

    I’m not sure about that. I would think that many Greens would be more nervous about the prospects of a Conservative victory and therefore more likely to vote tactically for Labour?

  22. @Jim Jam / Colin

    I looked at the YG Con/Lab scores on Monday evening for the month of Feb 2015 (to that date).

    The Tories were just above 32.5%.
    Labour were just above 33.5%.

    The average for each of the previous 2 months was Con: 32% Lab: 33%.

    Given that both parties are close to the rounding point, a couple of Con leads in the last few days of the month could easily lead to the February average ending as 33/33.

    Also, it is worth noting that AW’s “average” prefers more recent polls to older ones. At the start of the month, Lab hit 34 more regularly than they have in the past two weeks or so. The Tories in the same period have shown a number of scores higher than 32.

    On YG at least therefore, there is (to my untrained eye) evidence of tightening.

  23. John B

    Now that Populus have changed their methodology, I’m going to do something similar for their twice-weekly polls.

    I don’t think it worth doing for pollsters that don’t, at least, weight their Scottish sample to take account of SNP party ID.

    The distribution of the other party VIs may still be rather erratic, but we do get an approximation of the SNP VI from YG & Populus, and it is changes in that that would seem most significant.

  24. RAF

    Remember-you’ve always got the comfort of that 35% ceiling :-)

  25. Richard/Roger – differences come from using the National Readership Survey or the Census as a source for social class weighting targets. Essentially the ABC1C2DE social classification is based on the chief income earner’s current occupation or – if they are retired – their previous occupation. In the census, retired people are not asked what job they used to do, hence attempts to come up with approximated social grade in the census tend to run into problems with retired people, and the overall social class figures tend to be different.

    Scratch that, the difference is actually because Survation don’t weight by social grade. Should never overlook the obvious explanation.

  26. @Wes

    Yes, but if Labour, in order to gain the conservative inclined vote turn rightwards, or, putting it another way, fail to come out with positively Green policies, and if, in addition to that, they look less likely to win, then it seems to me that Green inclined voters will stick to their guns and not support a party which, in their view, and when it comes to their priorities, doesn’t seem very different from the existing government……

    The same applies to the SNP vote in Scotland and the extent to which Greens here will ‘lend’ their vote to the SNP – not a foregone conclusion by any means!

    There are an awful lot of ‘ifs’ and the ‘then perhaps’ logic is also speculative. But this GE seems to be so tight that the campaign to win votes must take all things into consideration…..

  27. With all the different variables affecting predictions as well as the increased number of unknowns around smaller parties, tactical voting, weightings and the lack of uniform swing, I really do appreciate the work the more statistics based on this site are putting in in order to try and make any sense at all of what is happening in the polls.

    Imagine if boundary changes had been implemented? I think the more stats based among us might well have found their brains exploding in such circumstances!

    Love coming here. The disparency between the discussions on this site and the narrative re polling in press and online has never been so stark…

  28. Is this why the UKIP vote is declining and could decline still further?

    Mind you, it’s nice to know Douglas listens to me…:)

  29. I see both UKIP and the Greens in Scotland hit 8% on this YG poll. Perhaps a Straw / Rifkind effect?

    If so, will there be a Bennett effect in tomorrow’s poll, and where will the votes go?

  30. John B

    There have been 6 Populus polls since they conducted a large survey of their Party ID data in their last January poll.

    I have arbitrarily selected 6 (3 weeks worth) to calculate the mean figure for. I’ll put up similar data when the next Populus comes out.

    Monday’s Populus Scottish crossbreak

    SNP 37% : Lab 28% : Con 19% : LD 8% : UKIP 6% : Grn 2%

    Mean of last 6 Populus Scottish crossbreaks

    SNP 43% : Lab 26% : Con 15% : LD 9% : UKIP 4% : Grn 3%

  31. YG and Populus Scottish mean scores showing great similarity.

    YG finds some more Tories, Populus more LDs, but both may be finding differences within a similar group of people due to panel composition – or maybe just noise.

  32. I’ve looked through all the seats in GB and for what it’s worth, here’s my prediction of how the election will pan out, as it stands.

    Labour 290
    Conservative 261
    SNP 34
    LD 33
    UKIP 4
    PC 3
    Green 1
    Speaker 1

    I don’t think any party will have a majority, but there is a chance that either Lab or Con could have the most seats.

  33. Election4castUK
    Latest forecast update: Con 284, Lab 280, SNP 38, LD 25, DUP 9, UKIP 1.

  34. 25-poll weighted charts updated (set of ten)

  35. I don’t think any party will have a majority, but there is a chance that either Lab or Con could have the most seats.

    no sh*t!

  36. Good morning all from Mount Florida.

    I wish Scotrail would introduce a fat fare for some of their larger passengers when they decide to take up two seats.

    “I think the exact opposite to your theory is more likely.
    Farage appears at ease and accomplished on TV, as his demolition of Nick Clegg demonstrated, the more airtime he gets the more votes he will stack up. That has been the case so far anyway, although I won’t be upset if you are correct”

    That was a one subject debate which Farage has momentum and form on.
    He could well receive a boost when he hits the air but when people actually go to put a X on the ballot paper I think they will have second thoughts about voting UKIP.

    I reckon they will poll around 9% tops.

    If the TV debates allow the audience to ask question the this is where Farage and UKIP usually screw up.

  37. @ Martyn

    Given that the alleged Ukip decline since Jan 1 is not showing up reliably either in general trends or comparisons or trends for individual polling houses, it is a fairly nifty move of Anthony’s to shift the focus onto the changes for each of the ten individual companies. Had there been one more drop, it could be argued that the decline is reliable using the binomial test. However, even using this rather inventive approach there still isn’t proper evidence of a statistical decline.

    It occurs to me that the one last chance of showing a reliable effect using current data would be to use the Jonckheere-Terpstra trend test. This polls the trends from independent companies into an overall statistic – which may or may not be significant.

  38. @Peter Crawford

    Please can I ask you to put quotation marks around the portion of another comment from which you are quoting ?

    It looks bizarre when you quote somebody (but don’t make it clear it’s a quote) then give a response that contradicts the quoted remark.

    I’m sure (hope) I’m not the only person thinking this.

  39. @NORBOLD
    In your link it says: “It is impossible to square those [Farage] words with Carswell’s comment that immigration has been an ‘overwhelming’ success.”
    Editors of papers and blogs do look for splits and agro, don’t they?
    It seems to me that Farage seeks to limit immigration from current levels to levels which would allow Carswell’s comments on successful integration over 50 years
    to be true again.
    Of course there would be no problem if we had kept the Huguenots out.

  40. Do we know if any pollsters have attempted to account for the changes in voter registration requirements?
    Does anyone even have much idea of the likely effects?

  41. More problems for Survation?

    TSE [email protected]

    Dear @Survation, you’ve deleted the tweet about publishing the Lib Dem constituency polls? Does that mean we’re not getting them?

  42. The latest plethora of polls, as contradictory as they appear to be on the true current status of the Tory and Labour VIs (Ashcroft and Survation pointing Labour’s way, YouGov and Opinium pointing in a Tory direction), they do seem unanimous, apart from Populus and TNS that is, on there being a decline in the UKIP vote. It’s still at a very high level compared to 2010, but the direction of travel is becoming clear. It’s heading south, although how far south it will eventually go before May 7th is a moot point. I never believed the high teens stuff, but I think they won’t go much below 10% come May. Some subsidence still to come then, but maybe not the levels some predict.

    What’s fascinating about the early stage of this decline (i.e the descent from about 15-12%, depending who is your pollster of choice) is that the initial beneficiaries are unclear. I wonder if this is because the “soft top” of UKIP support, the soonest to decamp at the sound of gunfire, is made up of a multiplicity of former voters, ex Labour, ex Lib Dem, ex Tory and ex DKs and, accordingly, the initial diaspora is breaking in pretty evenly balanced proportions and in a variety of directions.

    Ipso facto, then, does this mean that the really “sticky” part of their support is the proportion that would almost uniformly break for the Tories if they decamped? In other words, for the Tories to benefit disproportionally from a UKIP decline, do they need to get amongst those UKIP voters who pledged their troth some time ago (i.e. the voters who’ve made up the consistent 8-10% UKIP support we’ve seen in the polls for nearly three years)? The bedrock, if you like. Bedrocks, by their very nature, are loathe to shift once bedded in!

    Conventional wisdom has been telling us that as soon as UKIP deflate, the Tories will benefit. Maybe if you believe Opinium and last night’s YouGov, but definitely not if you look at the Survation and Ashcroft polls. It’s a delightfully mixed picture and just the sort of confusion that punctures pet theories.

    Fascinating stuff.

  43. CB – forgive my saying again but it is several weeks now.

    IMO – The UKIP may well manage to poll around 10%, although I still think a littel lwoer in the end, but in 90+% of Lab/Con marginals they will be squeezed mercilessly as will the Greens and remaining LDs.

    The UKIIP extrapolated vote share from these marginals imo will be nearer the 5-6% level with probably the extra 3-4% hurting the Tories a bit more but any Green lift would hurt Labour more (even if not direct but from 2010 some lefty LDs choosing Green rather tham Lab).

  44. @DAVE

    I don’t want to name drop (!) but I know Douglas Carswell quite well and have spoken to him many times albeit from an entirely different political viewpoint, but one thing I do respect him for is that he does not appear to share the general UKIP view on immigration. I think many of the people who voted for him in the Clacton by-election would be very surprised (and probably dismayed) by his views. Whether he will make a point about his views during the GE campaign is anothr matter!

  45. @Peter Crawford

    I hardly think that it was inevitable, until recently, that either main party could end up with the most seats. Labour were always ahead, if only slightly at times, until, say, 6 months ago.

  46. @Norbold

    Why did he leave the Tories then? Are you sure he wasn’t being a typical politician and tailoring his message to his audience

  47. “I hardly think that it was inevitable, until recently, that either main party could end up with the most seats. Labour were always ahead, if only slightly at times, until, say, 6 months ago”

    If truth be told, labour, even after snp gains are accounted for in
    scotland, are “still ahead” if you look at the polls as we stand….all the models which show the tories ahead on seats require some form of “swingback” to kick in.

    …sorry if this was caustic

  48. I noticed even in his acceptance speech that he seemed to be distancing himself from the standard UKIP view of immigration.
    My impression, based upon not much, is that Carswell is a rather principled and respect-able right winger whose views and iconoclastic approach probably don’t fit any party.
    Farage, and UKIP in general, appear by contrast to make it up as they go along.
    I’d be surprised to hear Carswell referring to Ting Tongs, even in private. Farage?

  49. I do wonder about Carswell’s place in UKIP.

    Prediction time – by 2020 Carswell will have done at least one of the following:

    Left UKIP acrimoniously
    Assumed the leadership

    If it’s the latter things could get very interesting. He doesn’t have the chummy man-down-the-pub relatability of Farage but I think he’s a genuinely talented political thinker.

  50. If UKIP are fading (a lot/little) it may well be good news for CON:

    How so?

    Support for UKIP – as for Front National and others across the EU like Swedish Democrats – is unquestionably linked to economic despair or depression.

    The same is true of anti-EU sentiment – as YouGov suggest.

    This isnt surprising: when people are struggling AND feel remote from government (whether national or supra-national) they look for alternatives.

    But if people see economic imrpovement (see the YouGov EU In/Out piece) they will start to move away from those ‘protest’ parties and, usually, back towards the incumbent government.

    Hence how, if UKIP are fading, it is likely to be good news for CON.

1 2 3 4 8