Scottish YouGov poll

There is a new YouGov poll of Scottish Parliamentary voting intentions tonight, commissioned by the Scottish Greens and YouGov’s first Scottish polling since last October. Topline figures are:

Constituency: CON 15%, LAB 41%, LDEM 8%, SNP 32%
Regional: CON 15%, LAB 40%, LDEM 7%, SNP 26%, GRN 6%

The last Scottish polling we saw was a rather surprising MORI poll earlier this month. Until then Scottish polls had been pretty consistent in showing Labour ahead, often by quite hefty margins. The MORI poll either reflected a remarkable turnaround for the SNP – showing Labour’s lead dropping by 11 points in the constituency section since the previous MORI poll – or a rogue poll. The hope was that the next Scottish poll would give us a clue.

Obviously we can’t be confident what way the trend is moving with YouGov, since this is the first poll for over three months, but in the constituency section it certainly shows a Labour lead consistent with the other polling prior to MORI’s last one. The regional vote shows an even larger lead for Labour – it’s unusual to have such a difference between the regional and constituency votes, so that might be a bit of a blip.

84 Responses to “Scottish YouGov poll”

1 2
  1. Reposting here! Not wishing to sound partisan, but the idea of Iain Gray as First Minister feels like an embarrassing low for Scottish devolution. The whole Labour front-bench are either unknowns or second-rate retreads who’ve hung on while more charismatic, able figures have imploded.

    Salmond remains a massive asset for the SNP for now, so that should narrow the polls during the campaign. But Labour’s remarkable 2010 performance probably has given them a real boost. If the SNP goes into opposition, it’ll be a hard slog.

    On the other hand, it will be interesting to see an unimaginative Labour group trying to implement cuts. Even if they restart council tax increases, how will they square their cuts over the next 4 years with their current rhetoric? Blame the Tories?

    The LibDems might poll badly, but a lot of their seats are fairly safe.

    Fascinating electoral contest- although the actual politicains are uninspiring.

  2. this poll is in line with the others. the mori poll in Scotland does not use past vote weighting like UK mori polls do – a very surprising decision from then. Mori was also statistically very wrong in its Scottish polls last year as well because of this different methodoogy

  3. Labour, to an extent, foresee making Holyrood Labour’s Benghazi… a base from which they will try to topple the Tories at W/Minster….

    The question is whether that is what Scottish voters want? Or will they prefer a more inward looking government, intent of focusing on local issues important to Scotland specifically, and must less to the UK generally.

    I don’t know the answer.

  4. “According to, the figures would give Labour 59 out of 129 seats at Holyrood, up 13 on 2007. The SNP would get 35 (down 12), the Tories 19 (up two), the LibDems nine (down seven), the Greens six (up four), with one Independent MSP.”

    If this was the result next May, which of the following scenarios would be more likely to emerge?

    A Lab/LD coalition
    A Lab/Green coalition
    A minority Labour government

    Would be interested to hear the views of Scottish posters on this question.

  5. Doug – no MORI polls use past weighting, either in Scotland or Great Britain as a whole.

    I think you are thinking of TNS-BMRB, who use post vote weighting for GB polls but not for Scottish polls.

  6. @Steve

    Can kind of see where you’re coming unfortunately from although hopefully Boyack and Chisholm would be in cabinet.

    This poll looks 3/4% too high for Labour on the regional and the same amount too low for the SNP.

  7. Those Regional List figures do look odd.

    All other polling has Salmond miles ahead as the choice for FM – regardless of VI. That should normally reflect itself in a higher SNP VI on the List – even if constituency voting remained heavily Labour biased.

  8. @D Abrahams

    I assume A Lab Minority coalition supported by the greens but this poll is probably exaggerating labour’s regional vote by a few %.

  9. ‘On the other hand, it will be interesting to see an unimaginative Labour group trying to implement cuts. Even if they restart council tax increases, how will they square their cuts over the next 4 years with their current rhetoric? Blame the Tories?’

    Labour, LDs and the SNP all seem to be keeping quiet on the university funding question. Only the tories have announced plans for a contribution.

  10. Steve “Not wishing to sound partisan…” And yet going on to do so:-)

    It’s true, I think, that there is a general lack of ‘quality’ in Holyrood and particularly telling that the star quality that does exist comes from ex-MPs. This is true of all parties.

    I suspect that the poll is gnerally how things will go. Alex Salmond has ‘brand recognition’ but there’s no general wave of enthusiasm for the SNP. The truth is that, on most things, there’s no great difference between Labour and the SNP.

  11. Anthony,
    Is there a link to the poll details on the you gov website? I cant seem to find one.

    The only thing I’ve found is a list of their Scottish polls, including one last October (although it seems to have also asked Westminster VI, whcih could have skewed it) and it is not drastically different from this one – except in that regional list vote, with Labour up a few points now and the SNP down a few points now.

  12. Landocakes,
    yes, AS does have the brand recognition, but its surprising that he consistently polls higher than his party – its difficult to predict whether that will help the SNP during the campaign, as the SNP have pretty much been campaigning constantly (at a national level) since 2007.

    Also, this time they wont be able to call the SNP “Alex Salmond for First Minister” on the regional list.

  13. I think you are totally missing the story here which is whether YouGov is now capable of conducting a Scottish poll.

    The sample here on Scottish Parliament first vote has the SNP AHEAD BY 11 PER CENT – 42 TO 31.

    The ADJUSTED figures then trun that on its head to produce Labour ahead by 9 per cent – 41 to 32!

    The reason for this is that YouGov are using political identifiers based on survey data at the time of – or just after – the Westminster elections which have Labour totally dominant.

    In other words it doesn’t matter what their own panel shows up they will always show Labour ahead.

    A 20 per cent switch from sample to adjused figures is in my experience unprecedented and totally questionable.


  14. Anthony – thanks for that – why then is the past vote question in the UK Mori tables but not the Scottish tables? seems weird

  15. John Ruddy,

    I hadn’t realised that this change had taken place, however I am delighted that the SNP will no longer be able to use ‘AS for First Minister’ on the regional ballot (or any other party variations thereof).

    Personally, I blame that marketing wheeze for much of the ballot confusion last time.

  16. Doug – they just ask it out of the goodness of their heart, they don’t use it for anything!

    (More practically, I think MORI’s monthly political polls are dedicated political polls so they have the room for lots of background stuff and demographic questions. Their Scottish voting intention questions are just added to the start of their Scottish Omnibus so they probably try to minimise the number of questions)

  17. Ex-Pat,
    I’m sure that Anthony will correct me if I’m wrong, but the point of weighting by westminster vote is to ensure that the sample is representative of the scottish population.

    Since Labour support is traditionally higher amongst women, would you say it was unfair of the poll to weight for the balance of women/men in the sample?

  18. John – I don’t think it was actually an attempt at sensible conversation, but YouGov don’t weight by westminster vote anyway.

    They weight by party identification, which isn’t specific to an election (if people vote differently at the two elections, as many do, they may define themselves as one rather than the other, or may see themselves as a voter who isn’t firmly aligned with either, or many not know how to answer the question and say don’t know. The bottom line is as long as you are weighting the individual data to the correct totals at the time you gathered it, it will do its job)

  19. While counting in Eire still goes on, here are the (hypothetical) results in seats according to various electoral systems on the basis of parties’ scores in the first preference voting (psephology is quite good for keeping my mind out of what is actually happening in North Africa):

    Full PR, no threshold (e.g: Netherlands, Israel)
    FG 60, LAB 32, FF 29, IND (on the assumption that they go together as a single party) 21, SF 17, GP 3, SP 2, PBP 2

    Full PR with 3-5% threshold (e.g. Sweden, Czech Republic, Denmark etc.)
    FG 62, LAB 34, FF 30, IND 22, SF 18
    Not very different from the previous, but allowing for a FG+IND majority

    Full PR with threshold for 5/6 of seats, 1/6 of seats as bonus to 1st party (Greece)
    FG 80, LAB 28, FF 25, IND 18, SF 15
    Still no OM for FG, but pretty close.

    Italian system, adapted to single parties (instead of coalitions): First party gets 54% of seats, other parties get the rest by PR with 4% threshold:
    FG 90, LAB 25, FF 22, IND 16, SF 13
    (Most stable system in theory since it guarantees OM no matter what, fails in Italy proper because of relentless splits, defections and recompositions in Italian parties).

  20. Anthony,

    “YouGov’s first Scottish polling since before the general election.”

    Err… I’m sorry Anthony, but that is just complete and utter pants. There have been *** THREE *** published YouGov surveys of Holyrood VI since the UK GE in May 2010. Please see:

    YouGov / Scottish Mail on Sunday Survey Results
    Sample Size: 1010 Scottish Adults Fieldwork: 31st August – 3rd September 2010

    ht tp://

    YouGov / SNP Survey Results
    Sample Size: 1212 Scottish Adults Fieldwork: 17th – 18th August 2010

    ht tp://

    … and there was a YouGov Survey for the Scotsman (fieldwork end date 21 Oct 2010) which has mysteriously disappeared from the YouGov online archive. Glad if you could tell us the new url. Thanks.

  21. Since Anthony seems to have an allergy against showing changes from the previous published poll by the same pollster when it comes to Scottish polls, here is the +/- change from the YouGov/Scotsman poll in October 2010:

    Constituency (FPTP)
    Lab 41% (+1)
    SNP 32% (-2)
    Con 15% (+1)
    LD 8% (n/c)
    oth 4% (-1)

    Regional (AMS)
    Lab 40% (+4)
    SNP 26% (-5)
    Con 15% (n/c)
    LD 7% (-1)
    Grn 6% (n/c)
    oth 5% (-1)

    [Stuart – how embarassing, I’d forgotten we’d done that – hence the text at the top of the report about the first since the election. I’ll have to change it – AW]

  22. ” showing Labour’s lead dropping by 11 points in the constituency section since the previous MORI poll ”

    Err… Anthony, what on earth are you on about?? There was only a 5 POINT drop in Labour support between MORI’s November 2010 and February 2011 polls.

    For the love of God, please store this url in your browser:

    ht tp://,_2011#Opinion_polls

    [Read my sentence again Stuart. It doesn’t refer to Labour’s share of the vote – AW]

  23. R HUCKLE – previous thread but just as relevant on this one
    I really can’t see that happening, as they [“Scottish” Labour] will want to keep quiet and be seen to make sensible decisions. I suspect Miliband and other powers within Labour realise that the Tories would use any issues in Scotland against them at Westminster. SNP would be wise to attack Labour on the basis of them being constrained by Westminster politics and therefore not able to act in the interests of Scotland.

    You make a valid point, which ultimately may serve to undermine the Lab/L-D “wheeze” in using the AMS system for Holyrood and Cardiff Bay that they hoped would give them a share of power. It also serves to demonstrate the ersatz nature of the “Scottish” in the trading names the three large London headquartered parties use in standing candidates for election in Scotland. Scots tories have declined since the Scottish Unionist Party allowed itself to be absorbed by its London cousin, and the disbelief shown even by partial observers like the BBC over Iain Gray’s assertions that he would have defied his London bosses over Megrahi. The L-Ds have a similar problem, particularly since they abandoned their federalism, joined Calman and now are pushing for Calman minus.

    AW makes a perfectly respectable and valid point re YouGov’s party association weightings, but as Scotland has become more politically sophisticated in showing different voting petterns in the four different electoral systems imposed upon it, there must be some doubt whether a single party association is adequate it their weightings. At least we haven’t too long to wait before we have the real polls themselves.

  24. @ A Brown
    “Can kind of see where you’re coming unfortunately from although hopefully Boyack and Chisholm would be in cabinet.”

    I believe Chisholm is retiring a the election, which is a blow to the Parliament as a whole, since he’s one of very few independent thinkers in the Labour party – or at least one very few who is senior enough to get away with it.

  25. I think Anthony wells is skimming too lightly over the dubious methodology of the You Gov Scotland polls.

    Although Robertson is hardly unbiased, I do think he and others are on to someting as he says:.

    SNP campaign manager Angus Robertson said “the way YouGov’s data was weighted, using Westminster rather than Holyrood party loyalty, cut SNP support, which was 13 points ahead of Labour unadjusted.”

    I just find it unconvincing that somehow they wildly overcontacted SNP supporters.

  26. Tom – “I just find it unconvincing that somehow they wildly overcontacted SNP supporters”

    There is no possible alternative explanation. If there were not too many SNP identifiers in the sample they would not have been any need to weight them down. Weighting targets are fixed. They are not decided on a poll to poll basis, they are fixed in advance.

    YouGov use a panel for sampling and aleady know the proportion of people of each party ID they need for a sample (it’s the proportion they weight too in the end). The email invites sent out to people to respond are normally in the same proportion as desired in the eventual sample (sometimes it’s adjusted for past response rate patterns – I suspect this needs to be looked at in this poll)

    Any difference between the two is down to differential respose rates. Basically, if 100% of the emails YouGov sent out to their panellists were responded to, there would be no need for any weighting (this simplifies things a bit, but is basically correct), if there is a lower response rate amongst some demographic groups, then you need weighting to sort it out.

  27. As a general comment –

    Degree of weighting is less important than what you weight to. The best way to explain this is an example. Imagine a you were surveying a population that had 10% left handed people, 90% right handed people and that this was the key meaasure in getting a representative sample.

    In the first sample, you ended up with 500 left handed people and 900 right handed people. To get to the correct proportions you would need to weight down the left handed people by 0.28 and weight up the right handed people by 1.4. For the left handed people, that’s comparatively hefty weighting.

    Now imagine a second sample. In this one there are 100 left handed people, and 900 right handed people. This sample has the correct proportions to start with, so doesn’t need any weighting at all. Perfect!

    Now, after weighting, which sample is more representative and why? Is it the one with weighting, or the one that didn’t need any? Both samples got exactly the same number of right-handed people, so they should be equally good at representing that part of society. Right? The first sample got five times more left-handed people than the second sample, so which one would better represent the left-handed people? Hmm.

    The result of having to do lots of weighting is simple. It decreases effective sample size – to understand why, go back to our example above. The actual problem is that the first sample had too many left-handed people. Pollsters solve this problem using weighting, but it could equally be solved just by interviewing 400 less left-handed people to begin with. The result would be the same, except for the sample size, which would be 400 lower.


    Now, go back to that first sample with 500 left-handed people, and 900 right handed people. If instead of weighting to be 10% left-handed and 90% right-handed, you instead weighted it to be 25% left-handed and 75% right-handed. Now you would only need to weight the left-handed people by 0.7 and the right-handed people by 1.17. There would be much less weighting (and the ignorant might look at the poll and be much happier with it)… but the sample would be wholly unrepresentative as it would have the wrong proportions of left and right handed people.

  28. Thanks for your response, Anthony which I believe I understand.

    Perhaps it means that SNP supporters are much more enthusiastic than Labour and will therefore be more likely to turn out on the day althogh I assume the certainty to vote aspect of the poll is meant to adjust for that.

    However, I am also left with some doubt as the the You Gov accuracy about who is an SNP supporter.

    I do anticipate the result being close with Labour slight favourites to be largest party.

  29. the unweighted figures for the Scottish poll were

    Constituency Vote


    Regional List


    quite a large turnaround between SNP/LAB

  30. @ An Duine Gruamach

    If you mean Malcolm Chisholm, he is not retiring. Fortunately, he is as keen as he ever was to represent his constintuency.

    We are looking forward to Malcolm being elected again ( Malcolm himself is never complacent about it & is working hard).

  31. Tom – the Yougov weighting is just by party ID – it asks which party people identify with.

    If we instead asked “Thinking about Westminster elections, which party do you most identify with” or “Thinking about Holyrood elections, which party do you most identify with” I think we’d get different answers, and some people would give different answers to each question.

    However, I think they’d also be much less suitable for weighting! If we mentioned elections, then people would start thinking more about how they voted, rather than which party they identified with, and we’d get a much less stable figure.

    Party ID is actually significantly different from voting intention – for example, in 2005 and 2010 there were many people who were Labour identifiers, but who abstainted or voted Liberal Democrat. Party ID is also much more stable than voting intention – for example, the British Election Study found very little change in party ID between 2005 when Labour won by 3% and 2010 when the Consevatives won by 7% – at both elections Labour had a large lead in the proportion of people who identified with the party

    This is why it’s useful for weighting – whatever the sort term ups and downs of party support, the proportions of people identifying themselves as being “a Labour person”, “a Conservative”, “a Nationalist” or whatever are comparatively steady.

    It’s probably not stable enough to be suitable for weighting for a non-panel survey – there’s always the chance of sudden change and therefore I would not recommend it for phone polling, but for a panel based company like YouGov, movement isn’t a problem anyway, since they collected panellists party IDs back in May, and can weight fixed historial data to a fixed historical point.

    That said, we do need to track the movement so new recruits to the panel can be intregrated over time. Not an issue at the moment, since most panellists were on the panel in May, but it all needs keeping under constant review.

  32. Tom – missed your other point. Differential response rate isn’t necessarily anything to do with the enthusiasm of the party supporters – I looked at it in some detail when we were trying to work out why the polls overestimated the LDs at the election, hoping to find some great disproportionate surge in LD response rates. There wasn’t one.

    Rather it can just be down to demographics (people in some age groups or social groups are less good respondents ANYWAY, and sometimes that correlates with party).

    Anyway, at the moment YouGov isn’t factoring in likelihood to vote for the Scottish election – we only do it during election campaigns. That said, we’ll probably switch over very soon. Historically this tends to work against Labour, though not by much (during the UK general election, when we started turnout weighting it increased the Conservative lead over Labour by one point on average. It won’t necessarily have the same effect in Scotland and without Labour as the party of government, but I suppose it gives an idea of the possible scale of the difference – relatively small).

  33. Thanks Anthony. That was all extremely helpful.

  34. Stuart Dickson

    The YouGov/Scotsman October poll is now at this reference:

    I’ve a feeling that the Scotsman published the results in two batches several days apart and YouGov later consolidated the tables together here

  35. @ Amber Star

    That’s good to know – much as I hope he is still in opposition after the election. One man of his integrity is worth of ten Richard Bakers, Karen Whitefields or Iain Grays.

  36. Anthony

    I assume that the Herald gave Angus Robertson of the SNP the tables so he could discuss them and make the criticisms described above. (YouGov will presumably put them up tomorrow).

    When the October poll was published, it had a similar problem and I wondered if it was something to do with YouGov’s algorithm of polling disproportionate numbers of panelists with Others IDs so as to get better granularity among them.

    Could this be related to over-selection of SNPers?

    Incidentally why are YouGov not doing voting likelihood for Scotland, when they are doing it for Wales?

  37. Roger – the oversampling of Others in YG polls isn’t part of some cunning plan (it’s just a positive unintended consequence), it’s something that will be sorted out in the fullness of time as it is a waste of money to sample all those unneeded others only to weight them down.

    It isn’t connected to this issue – the Scottish sampling is done differently.

    I was going to mention Wales but thought we were intended to switch over this month rather than last month and didn’t want to jump the gun! Yes, ITV Wales wanted to start factoring it into their polls from now, so we made the switch. Since we’ve already done it for Wales I expect we’ll do it very soon for Scotland too.

  38. Its looking pretty good for Labour on these numbers.

    Its looking OK for the Tories (relatively speaking).

    Its looking dissapointing for the Nats and looking utterly desperate for the Scottish Lib Dems.

  39. Anthony

    One thing (well one among many!) I don’t understand is why (when weighting is done by Party ID regardless of election) that the weighting of SNP respondents was done differently in your October poll between the questions for Westminster and Holyrood.

  40. Oldnat – I’m not sure where you get that idea from. Unless there were different samples (or one of them had turnout weighting) they shouldn’t have been (and if they were different samples, they’d still have been weighted to the same targets!). Let me check the tables.

  41. Oldnat – looking at the link Roger has put up earlier for the Oct poll, both questions are on the same weighted sample. They’ve been weighted together to the same figures.

  42. OK Is it my Maths, or my lack of comprehension :-(

    Lab SNP
    Weighted 494, 295
    Unweighted 443, 396
    Weighting +12%, -16%

    Lab SNP
    Weighted 452, 393
    Unweighted 403, 506
    Weighting +12%, -12%

  43. Ah – that’s the effect of the weighting, not the weighting itself. The weighting was the same, but since Westminster and Holyrood voting intention are different things (related, obviously, but different) it can different effects upon them.

    Those cross breaks are the voting intention at Westminster and at Holyrood. There isn’t a cross break by party ID, which is what we weight by.

    The effect of weighting isn’t necessarily universal across the board, that applies as much to demographic weighting as it does to political weighting.

    Say (these things are always better with an example) – that 30% of men voted SNP at Westminster elections, but 40% of men voted SNP at Holyrood elections.

    Now imagine you had a sample with too few men, and you needed to weight men up by 10%. The result on the SNP Westminster support would be to boost it by 3%, but the effect on SNP Holyrood support would be to boost it by 4%.

  44. Anthony


  45. yougov overestimated the SNP and underestimated Labour in all their polls from Jan-May 2010, from Jan-May 2007 and from Jan-May 2005. So if anything the methodology is pro-SNP not anti. Please do check the tables for yourself.

  46. Doug Stanhope

    Maybes aye, maybes naw.

    By definition, no poll can be known to be an over or under estimate of VI at that moment in time.

    These damn voters just won’t stand still once they have been counted! :-)

    If you had restricted your comment to the final poll before each election, you may be right (or not).

    Why did you want to include earlier polls?

  47. Oldnat – i agree with your sentiment ie they are just snapshots of opinion. I included the Jan-May polls because they are ‘election year’ polls, nothing else. but I do think my basic point is still very important, the yougov polls overestimated the SNP in 2010, 2007 and 2005 so the methodology should be looked at to stop this happening.

    I do think my basic point is still very important, the yougov polls overestimated the SNP in 2010, 2007 and 2005 so the methodology should be looked at to stop this happening.

    I’m sure Anthony will correct me if I’m wrong, but I think any queries here are related to weighting – esp. on party ID & paper readership – rather than to YouGov’s methodology, and in the context of the 2011 General Election only the 2007 figures have any real relevance.

    The last poll YouGov conducted before the 2007 General Election was for the Daily Telegraph, here, with fieldwork: 1-2 May. That poll was

    Constituency Vote
    Actual,  YouGov,  Party
     32.9%,  37%,  SNP
     32.2%,  31%,  Lab
     16.6%,  13%,  Con
     16.2%,  14%,  L-D
      2.1%,   5%,  Others

    Regional Vote
    Actual,  YouGov,  Party
     31.0%,  32%,  SNP
     29.2%,  27%,  Lab
     13.9%,  13%,  Con
     11.3%,  10%,  L-D
     14.6%,  18%,  Others

    From where I’m looking, I’d say that the YouGov figures were pretty close to the actuals, and YouGov’s small overstatement of the SNP and Others could have been due to last-minute factors such as media coverage.

    But, even more relevant to current polling is that YouGov’s party ID & paper readership weightings have changed at least twice since then, and the $64,000 question will be [after the event, of course]: Have the changes improved the accuracy or made it worse?

  49. Hi barbazenzero – sorry if I used the wrong term. Not an expert just interested in this stuff!. I agree that the figures are not wildly out all the time, but I do think overestimating the SNP and underestimating Labour – however small – at the last 3 elections is relevant and not just the overestimating of the SNP and underestimating of Labour in 2007. Ipsos-Mori did it in 2010 as well.
    all fun and games!

  50. Doug Stanhope

    Also TNS (or at least the old System 3) seemed to exaggerate support for the biggest parties – SNP and/or Lab – maybe because it was face to face?

    I suspect (and it’s no more than that) that systems designed for specific populations may not necessarily translate well to other environments.

    Angus Reid had a lot of problems adjusting to the GB market – despite having an excellent methodology for Canada.

    YouGov’s methodology may not translate well to Scotland – no matter how well or badly it performs in England.

    At a guess we are going to continue to see wide variations in polling from now till May. As NewsNet Scotland said in their headline today

    “SNP and Labour will both win election, new poll evidence shows”

1 2