Tuesday, June 17, 2014

According to Better Together's own logic, the No lead has slumped by 3% in tonight's YouGov poll

The results of tonight's YouGov referendum poll have just been released, and they show the firm consolidating its status as one of this campaign's extreme No-friendly outliers -

Should Scotland be an independent country?

Yes 36% (-1)
No 53% (+2)

Now, the percentage changes I've listed above are accurate, and are from the most recent YouGov poll in April.  It's worth making the point, though, that the No campaign have for several weeks been insisting that there was a more recent YouGov poll than that - the one conducted by Progressive Partnership.  That's a load of garbage (Progressive are an entirely different pollster with their own weighting procedures), but they've been doing it so they could claim that YouGov were showing a six-point increase in the No lead (in fact both YouGov and Progressive had shown a decrease in the No lead).  So according to the No campaign's own logic, the changes in tonight's poll must actually be reported as follows -

Yes 36% (+4)
No 53% (+1)

In the interests of consistency, we can doubtless look forward to a "Momentum for Yes" graphic appearing on Blair McDougall's Twitter feed any second now.  What's that?  Don't be cynical - of course they're not going to do another "Oceania has always been at war with Eastasia" pivot, and suddenly decide that the Progressive poll was not a YouGov poll after all.  These guys have integrity.

This is also an appropriate moment to repost an observation I made about Kevin Schofield of the Sun (who commissioned the poll) before the results were published -

I was also amused to spot Kevin Schofield almost biting someone's head off for accurately pointing out that the last YouGov poll for the Sun showed a 60-40 lead for No. It's actually normal practice for a newspaper to make the comparison with the previous poll commissioned by themselves, even if the same polling company has conducted a poll for another client more recently.  So clearly Schofield is determined to make the comparison with the poll that gives the highest baseline figure for Yes, in order to push the "momentum for No" spin. He's nothing if not transparent.

So how does tonight's result rank for Yes when compared to previous polls from this most No-friendly of firms?  Pretty well, actually - it's the third lowest No lead reported by YouGov during the campaign so far, only being beaten by the previous two polls which showed leads of 15 and 14 points respectively.  Those numbers are well within the standard margin of error when compared to tonight's poll, so it's perfectly conceivable that nothing has changed, or even that Yes has made further progress.  In normal circumstances you'd say the latter possibility is far less likely, but the situation is considerably less clear-cut when you've just had record-breaking highs for Yes reported by both Panelbase and Survation.  As recently as the 1st of March, a YouGov poll was published showing a No lead of 18 points - 1 point higher than tonight, and yet at the time that was as good as it had ever been for Yes.  It must also never be forgotten that YouGov were showing a No lead of 30 points as recently as last August, meaning that it's still true to say that the gap has essentially halved since then.  So we have a pattern that is at least consistent with the possibility of Yes being at an all-time high across all the pollsters (with the trend being masked in this particular YouGov poll by random sampling variation), but obviously we'll have to wait for further evidence to see if that is actually the case.

We've now had four polls from different firms over recent days (I'm excluding TNS-BMRB because the fieldwork for that was much, much earlier), and three of them have shown decreases in the No lead, with only one showing an increase.  And it's not only the case that YouGov are contradicting a general trend - the extent of the increase they're suggesting in the No lead is less than the decreases reported by ICM, Panelbase and Survation.  The chances that there have been any real world gains for No of late therefore seem vanishingly small - unless of course they happened only over the last few days as a direct result of the absurd reporting of the Lally/Rowling non-stories.  I think that's unlikely, but even if it is the case my guess is that it will prove to be a pyrrhic victory for the Abominable No-Men.  Our lazy journalists may have fallen hook, line and sinker for the 'Cybernat' nonsense, but they'll get bored with it quickly enough, and any temporary benefit the No camp have cynically squeezed out of the whole thing will melt away.

In another sense, You Gov isn't the outright odd-one-out of the four recent polls.  YouGov and ICM may be contradicting each other by showing very different directions of travel, but the common factor with both polls is that they can be seen as reversions to the mean after an unusual finding in the previous poll.  By contrast Panelbase and Survation are both showing tiny No leads that are well below those firm's normal ranges.  So if we set aside the possibility of a very recent swing to No caused by the Rowling/Lally garbage, that essentially leaves us with two roughly equal possibilities - a) the position has been relatively steady recently, or b) Yes have surged to their highest ever level of support.  At least one or two more polls will be required to resolve the uncertainty.

We heard a couple of days ago from someone who took part in the YouGov poll (or at least I assume it was this poll).  He mentioned that he was asked for his country of birth.  I'll be watching like a hawk to see if those figures are published, because on the one previous occasion that YouGov released that sort of data, it turned out they had far too many English-born people in their sample, and far too few Scottish-born people.  If that's a consistent problem for the firm, it could cast doubt on the size of the No leads that they typically report.  We saw with the weekend's ICM poll that reweighting to the correct country of birth target figures from the census was sufficient to reduce the No lead by 2.4%.

As soon as I saw Kevin Schofield mention that a YouGov poll was on its way, I had an inkling that it wasn't going to show further movement towards Yes, because I had only just read an astonishingly conceited article on the YouGov website that effectively said : "We're completely ignoring the fact that other major pollsters are showing a virtually tied race. Only our own polls matter, and because we're showing a wider gap, Alex Salmond will be desperate for a game-changer."  Back in the real world, there is far more than just one possible scenario for a Yes victory.  The first scenario is that Panelbase are the most accurate firm, in which case it doesn't matter if every other pollster continues to show a No lead between now and polling day, and it doesn't matter if the Yes vote in the Poll of Polls never gets any higher than 45% or 46% - all we need is a tiny further swing, and we'll be there.  In some ways that would be the most satisfying way of all to win, because it would absolutely stun the London establishment - but they'd have no-one else but themselves to blame for failing to take the Panelbase results seriously enough.  A second scenario is that ICM and Survation are closest to the truth, in which case a more middling swing is required - although even then it's still possible that Yes could win after never quite taking the lead in the Poll of Polls.  And the third scenario is that the No-friendly pollsters are right, in which case Yes will need a heftier swing and an outright lead in the Poll of Polls - but that won't necessarily require a "game-changer".  The SNP pulled off exactly that kind of swing over a short period of time in 2011 without any specific game-changing moment (unless you count Iain Gray taking cover in a sandwich shop).

*  *  *

I've been having a look at the datasets from the ComRes regional poll for the Borders and Dumfries & Galloway.  The exact swing to Yes since the previous poll in January is 1.2% (although over on Planet McDougall that somehow constitutes "momentum for No").

English-born respondents make up a whopping 24% of the ComRes sample - I've no idea if that means they've been over-represented, but it certainly goes a long way towards explaining the size of the No lead in the poll.   (For comparison, the 2011 census showed that just 9.6% of the Scottish population are English-born.)  Among Scottish-born respondents in the Borders and D&G, the Yes vote is a full five points higher than among the full sample.

I must say that I think ITV Border are being a bit thick in the way they're approaching these polls.  If they can afford telephone polls of 1000 people in the Borders and D&G, then by definition they can afford Scotland-wide telephone polls of 1000 people.  That would tell us far more, and it would also attract much more interest in ITV Border programmes that release details of the polls (ie. Lookaround and Representing Border).  But as things stand, it's only STV that are bothering to commission national telephone polls - or at least ones that are intended for public consumption.

*  *  *


The mild disappointment of the YouGov poll hasn't had a huge impact on the trend shown by the Poll of Polls - the last update showed the lowest No lead ever, while tonight's update merely shows the second-lowest No lead ever.

MEAN AVERAGE (excluding Don't Knows) :

Yes 43.8% (-0.3)
No 56.2% (+0.3)

MEAN AVERAGE (not excluding Don't Knows) :

Yes 36.7% (-0.1)
No 47.0% (+0.3)

MEDIAN AVERAGE (excluding Don't Knows) :

Yes 43.6% (-0.2)
No 56.4% (+0.2)

(The Poll of Polls is based on a rolling average of the most recent poll from each of the pollsters that have been active in the referendum campaign since September 2013, and that adhere to British Polling Council rules. At present, there are six - YouGov, TNS-BMRB, Survation, Panelbase, Ipsos-Mori and ICM. Whenever a new poll is published, it replaces the last poll from the same company in the sample. Changes in the Poll of Polls are generally glacial in nature due to the fact that only a small portion of the sample is updated each time.)

The median is now being calculated as the mid-point between ICM and TNS-BMRB.  Having moved into a more middling ranking with their last couple of polls, YouGov have now reverted to their customary place at the extreme No-friendly end of the spectrum.

The No campaign's lead in the Poll of Polls mean average (not excluding Don't Knows) :

Sep 2013 - 21.6%
Sep 2013 - 21.4%
Sep 2013 - 19.4%
Oct 2013 - 18.8%
Oct 2013 - 18.4%
Oct 2013 - 18.2%
Nov 2013 - 18.4%
Nov 2013 - 18.0%
Dec 2013 - 17.0%
Dec 2013 - 16.8%
Dec 2013 - 16.4%
Jan 2014 - 14.4%
Jan 2014 - 14.2%
Jan 2014 - 14.2%
Jan 2014 - 15.2%
Feb 2014 - 15.0%
Feb 2014 - 15.5%
Feb 2014 - 15.5%
Feb 2014 - 13.7%
Feb 2014 - 13.3%
Feb 2014 - 14.2%
Mar 2014 - 14.2%
Mar 2014 - 14.5%
Mar 2014 - 14.5%
Mar 2014 - 14.7%
Mar 2014 - 13.8%
Mar 2014 - 13.0%
Mar 2014 - 12.5%
Apr 2014 - 12.5%
Apr 2014 - 12.7%
Apr 2014 - 12.7%
Apr 2014 - 12.3%
Apr 2014 - 11.4%
May 2014 - 11.2%
May 2014 - 11.2%
May 2014 - 11.5%
May 2014 - 13.3%
Jun 2014 - 12.1%
Jun 2014 - 12.1%
Jun 2014 - 11.3%
Jun 2014 - 9.9%
Jun 2014 - 10.3%

Last but not least, many thanks to Philip Jack for giving me to permission to use the graph below, which illustrates the Poll of Polls trend in visual form.  It doesn't include the last couple of updates, but it'll give you the general idea!

(Click to enlarge.)


  1. Margin of error sampling so not much of change. You will have to see the full data tables for better information.

  2. Yes, although YouGov's datasets are inferior to all the other firms (no raw numbers).

  3. No doubt you're busy preparing an article on today's British Election Studies survey which claims that 'a half' of Scots believe more devo powers will be delivered in the event of a no vote. And no doubt you'll notice that the survey was conducted during February and March, before any of the Westminster parties published their devo plans. And of course you will also notice that in last week's Panelbase poll (11 June) respondents were asked "If there is a No vote, do you think that the Westminster parties could be trusted to deliver any extra powers for the Scottish Parliament?" Only 35% said yes, 43% said no. So the real story is that belief in more devo in the event of a No vote has FALLEN from a half to a little over a third since the parties published their plans.


  4. To be fair though James your approach of analysing the average of all the polls does produce glaringly obvious proof of the swing to the Yes vote. Which won't do at all.

    Now why would you want to do that just because the huge disparity between all the different pollsters is so blatant?

    I admit I too fell into the bad habit of analysing all the polls instead of picking up on one or two. Which is why I predicted months ago that the labour vote would tumble yet again for the EU and local elecions in May while the kippers would come on top. More fool me it would seem.

    Instead of all that boring trend analysis you could always go to the epicentre of the out of touch tory twit at politicalbetting.

    A place where a senile old has-been of a tory like JackW makes predictions based on his own farts. (not a joke for those not familiar with him. That is all he has to justify his 'numbers' as he has been forced to admit several times now) This farting 'pundit' JackW is actually taken seriously by some of the most amusing fools on PB. Which really does say it all.

    PoliticalBetting is of course also the place where the most dim-witted collection of out of touch twits. trapped inside the westminster bubble. try and fail so hilariously to understand scottish public opinion. We're still waiting on Tony Blair becoming the lead spokesman for the No campaign as one of the most witless of the PB tories claimed while he tried to persuade his credulous tory audience. (again, not a joke, that was an actual article)

    After all, the PB tories and doddering old fools like JackW predicted utter doom for the Yes campaign months ago so it must have happened while all this so called proof of the continually shrinking No lead must be a figment of our imaginations.

    It's also not as if there's something called a campaign on the ground still to come in the final few weeks that could conceivably have any effect on the vote, is there?


    So as Cammie used to text so often to Rebekah Brooks..


  5. Somewhere around the last days of February, in UK-wide polls, Yougov's unweighted base numbers for Scotland climbed from ~160 to 231 on average.

    So, now 231 respondents are being shaved to ~173.

    What happened here? Did Yougov start asking more panelists? Did more panelists sign up? Did more panelists just respond to requests?

  6. I'm in Berwickshire in a small town and I am pretty sure that 25% of the town aren't English. It's a smaller percentage than that

  7. James,
    I've had a dig around to try and find out who are the people (owners/share holders) behind the big polling companies. It's proving to be quite difficult to get info about them. I thought it would be good to have some background on these companies to use when I'm talking to people on the doorstep? Do you have anything to hand?


  8. YouGov tables have been released. It looks like they asked Holyrood 2016 voting intention first and found a small Labour lead (~2 points) over SNP in both constituency and list votes.


    It seems a bit odd to me that Labour would be ahead in Holyrood VI if they couldn't win in the Euro election.

  9. Labour 2010 + Holyrood 2011:
    5.3% of the sample up-weighted to 10.1% of the sample.

    What about Lib 2010 + SNP 2011? Why not weight for these too?

    How about Tory 2010 + SNP 2011? Do they not get a shout? Must be some.

    What about Lib 2010 + Lab 2011? Why exclude them in this clever 2010 weighing system?

    If you are getting false recall on 2010 you need to eliminate it completely. Yet Yougov still 2010 weighting...

    Big up-weighting again needed for SNP 2011 from the unweighted base.

    Anyway, Rowlinggate gives Yes a boost when you take into account leaners (change due to Rowling).

    41(+5)% Yes
    51(-2)% No

    45% Yes
    55% No

    Excluding DK.

    A new high for Yes in Yougov.

  10. Scottish Skier, that was your most confusing post ever. Not the slightest idea what any of that meant.

  11. It was about Yougov weighting.

    Pollsters have consistently found that people in Scotland recall how they voted in 2011 well, but when you ask them about 2010 the same sample gives a very different answer; this normally yielding far more people saying they voted SNP and less Labour than was the case.

    It could simply just be false recall due to 2011 being the last ‘general election’ but I suspect it is not.

    SNP were polling 30% ahead of 2010 yet only got 20%. So 10% of the population supported the SNP but tactically voted for Labour or Lib to stop the Tories.

    Those SNPers who voted Labour voted for a party they didn't support and their tactic failed miserably with the Tories returning. Those who tactically voted Lib found themselves to have voted Tory. Nice! So, maybe it’s not quite false recall, but ‘I’m not sure I want to admit what I did’.

    Anyway, all pollsters have found this problem so have stopped 2010 weighting and gone for 2011 weighting if they use past vote weighting (MORI does not for example). Yougov however think they are clever and have now created a group who say they voted Labour in 2010 but SNP in 2011.

    Now, some people will happily admit doing this. They are Labour people who think the SNP are ok in Holyrood but likely are less inclined to independence. The problem is Yougov can’t find enough of them so have to up-weight them to almost double their number. They can’t find enough because the tactical SNPers who voted Labour 2010 SNP 2011 are not saying that’s what they did. So, the views of 5% of the population who are more inclined to the union are now attributed to 10% of the population, with 5% of the population who are more independence orientated being excluded from the sample.

    My point was taking the piss on this. Yougov are still weighting to 2010. If people are not giving the right answer what they voted in 2010 you must exclude it completely. So, me saying ‘What about Lib 2010 + SNP 2011’ is mocking the fact that if you pick out one group of voters for voting differently in sequential elections, why not do it for all if it’s such an amazing method?

    Why in UKGEs do they not start creating groups for e.g. ‘Voted Tory 2010, Labour 2005, Lib 2001, Tory 1979'? Because that's just stupid is why.

    Basically, in addition to having far too many English people in their Scottish base, yougov’s methodology is a pile of dung and it’s obvious.

    MORI = SNP comfortable 10 point lead (no past vote weighting)
    ICM = SNP comfortable lead (2011 weighting)
    Survation = SNP comfortable lead (2011 weighting)
    SNP wins EU elections!
    Yougov = SNP behind Labour (2010-11 hybrid weighting)

    Erm, Aye. Yougov must be right.

    Yougov Rownlinggate:

    Does the abuse directed at JK Rowling make you more likely to vote YES in the referendum, more likely to vote NO, or does it make no difference?

    6% More likely to vote YES 6
    35% No difference - I would have voted YES anyway
    = 41% Yes

    12% More likely to vote NO
    39% No difference - I would have voted NO anyway
    =51% No

    9% Don't know

    =45Y/55N excluding DK

    Rowlinggate boosting Yes according to Yougov.

    Funny that when you ask a different way only 39% are now saying they're firm on No...

  12. Scottish Skier, I think that YouGov weight unsung their panel member's replies to the post 2010/2011 election questionnaires that YouGov send out to panel members immediately after the election, rather than by recalled vote as at the date of the poll.

  13. This comment has been removed by the author.

  14. post 2010/2011 election questionnaires that YouGov send out to panel members immediately after the election

    I've heard this before from Roger Mexico (UKPR).

    Yougov, in agreement with all the other pollsters, obviously recognised they had a problem with 2010 weighting and they've tried to fix it with the 'Labour 2010 + SNP 2011' group.

    If they're are backing up 2010 weighting based on an questionnaire done right after 2010, this should make false recall issues even less likely. If people were asked what they voted in 2010 just after the election, there'd be no confusion with 2011 and not much chance they'd forgotten.

    Thus 2010 weighting should be pretty damn good and no need for 2011 weighting; both should work?Unless of course people lied to yougov about 2010 in the post election questionnaire of course (e.g. tactical voters I described). If you were someone who'd moved to SNP in 2009 and tactically voted Lib only to find you've voted Tory, would you be straight up? Something's up here.

    The fact that people (same group I believe) lied to pollsters for over a year about Holyrood (SNP were up to a solid 40% by 2009), saying Labour when they planned SNP, suggests this group is economical with the truth.

    Very interesting.

    Even in this Yougov poll in the Rowling Q there's evidence that the No is not quite that too.

    Only 39% set on No. Then there's 12% saying they are actually only leaning to No, not definite No.

    Sounds more like they said No, even though they weren't really for it. Do they approve of abusive nats? Oh, erm 'No'.

    It's like TNS. No 12 points ahead on 42 yet <30% definitely voting No? Eh?

    No stonkingly ahead in MORI yet >50% think the Yes campaign are doing a sterling job with just 1/4 approving of Better Together? Whit?

    SNP / leaders getting 50-60% satisfied even though they only poll 38%? (MORI). Eh, WTF?

    Big majority for not joining the union today yet people say they'll not vote to leave it? (panelbase). Come again?

    Big majority think Scotland could be an successful independent country (panebase) but they'll vote for the union? Are you for real?

    What kind of screwed up country is Scotland?

    Only explanation I have is some people telling porkies in straight Y/N. Make it less direct and answers change, with No softening, just like that 'rouge' panelbase for the SNP with some prompting Q's.

    Show them the Tories and suddenly they're Yes!!!! Offer them an egg mayo sandwich a day (SSAS £500 Q) and it's a likely stonking majority for Yes...

    Maybe I'm wrong and Scotland is just the weirdest place on earth and not inhabited by humans.

  15. Was like this in 2009-11.

    late 2009 to early 2011 AS getting >50% satisfied (with ~35% not), Gray struggling, yet Labour building up a huge lead in polls, even getting 49% from TNS?


    Porky pies.

    Those people that gave Labour its mirage of a huge poll lead ahead of 2011 never planned to vote Labour.

    Bettertogetherunitedwithlabournothanks better hope and prey they're not being led up the garden path again by largely the same people.

  16. Thanks, SS, think that's helped.

  17. Thanks, SS, think that's helped.