Sunday, April 27, 2014

Another example of what gives YouGov such a bad name

This question appears in today's YouGov poll commissioned by the Sunday Times, and it's certainly not the first time we've seen a variant of it -

If there are to be TV debates at the next election in 2015, which of the following would you prefer?

a) Straight debates between David Cameron and Ed Miliband, as they are the only two leaders who stand any realistic chance 

b) Three-way debates, like last time, between the Labour, Conservative and Liberal Democrat leaders

c)  Four-way debates, to include Nigel Farage, leader of the UK Independence Party, as UKIP is now more popular than the Lib Dems

d) Don't Know

If I had been responding to that question, it would have been literally impossible for me to give an honest answer.  I'd have had to say "don't know" to avoid appearing to support one of three options that I strongly oppose, but it's not true that I don't have an opinion - I have an extremely firm view that the SNP and Plaid Cymru should be represented in any leaders' debates that may take place.  Even some respondents in England must have struggled with the question, because there will be many people who think that the Greens and/or Respect should have a place in the debates - after all, unlike UKIP, both of those parties are already represented in the House of Commons.

Listen, YouGov.  If you want to arbitrarily restrict the ability of your respondents to express their positive preferences, that's one thing.  But what is so hard about giving them a "none of the above" option to at least allow them to feel that they aren't having words put in their mouth?

*  *  *

Because of the absurd weighting scheme that YouGov use for their Scottish respondents in GB-wide polls, it's quite unusual for the SNP to hold a clear lead in the Scottish subsample for Westminster voting intention, but today is one of those rare occasions -

SNP 36%
Labour 29%
Conservatives 17%
Liberal Democrats 9%
Greens 3%

Heaven only knows what the position would be if YouGov were using a sensible weighting procedure, because as happens day in, day out, the voices of people who identify with the SNP and Plaid Cymru have been massively downweighted in this poll.  51 SNP and Plaid identifiers were interviewed, but they were downweighted by a whopping 40% to count as just 31 'virtual' respondents.

As you'd expect, the SNP also hold the lead in the subsample for European Parliament voting intentions, although curiously that lead is somewhat more modest than the one for Westminster...

SNP 29%
Labour 25%
UKIP 18%
Liberal Democrats 10%
Greens 8%
Conservatives 8%

Much has been made of the fact that UKIP have now taken a slender European election lead over Labour in this poll across Britain.  Unlike some people, though, I don't think that's good news for the Yes campaign.  Voters in Scotland will certainly be repulsed by a UKIP victory down south, but the biggest side-effect is likely to be a temporary UKIP surge in voting intentions for Westminster, which will come largely at the expense of the Tories.  That will disguise the overwhelming polling evidence that the Tories are heading for a general election victory next year, thus potentially giving soft No voters the comfort blanket of thinking that they aren't necessarily voting for David Cameron to remain as Prime Minister (which is of course exactly what they're doing).

However, the only real alternative to a UKIP victory in May is a Labour victory - it's highly unlikely that an incumbent governing party in Westminster could come out on top in a relatively unimportant mid-term election that offers the opportunity for 'free hit' protest voting.  So whichever way it goes, there may well be a scenario created that will give Labour a misleading temporary boost in the GB-wide polls, which is the last thing we need with just four months to go until the referendum.

We can't control any of that, so the most important thing is to ensure that the SNP win the popular vote in Scotland, and if possible an additional seat.  The momentum generated from that could prove invaluable.


  1. I think it's worth remembering that UKIP got 17% (and 2nd place) in the 2009 Euro elections, but still only 3% in the 2010 general election.

    We're fond of pointing out the "sophistication" of the Scottish electorate in tailoring their vote to different electoral systems, and I'd be wary of the lazy prejudice that suggests UKIP voters aren't capable of the same.

    Despite everything, they know FPTP is stacked against them. When it comes to the crunch they won't vote in such a way as to let Ed Miliband become PM when Cameron is promising a referendum and Miliband isn't.

    Of course, that doesn't preclude their vote still showing as high in opinion polls, but there's no way a big UKIP result at the Euros will work in favour of Labour.

    It might even see more Labour voters defect to UKIP from poor areas where there's strong working-class anti-immigrant sentiment. You'd have to suspect most of the disgruntled Tories who might move to UKIP have already done so.

  2. There has been a sea-change since the last Euro-elections, though - UKIP have made a breakthrough in local elections, they are now regularly outpolling the Liberal Democrats in Westminster voting intention surveys, and they even topped their first Westminster constituency poll recently (in Eastleigh). So it's no longer fanciful to think that UKIP could make some kind of breakthrough at a UK general election, albeit they'd obviously still be polling at a significantly lower level than in European elections.

    "You'd have to suspect most of the disgruntled Tories who might move to UKIP have already done so."

    Unfortunately I don't think that's true - what you'd be likely to see in the post-May polls is a large number of the Tories who voted UKIP in the Euro-elections also saying that they will now vote UKIP for Westminster, perhaps taking UKIP up to around 18-20%, and indirectly increasing Labour's overall lead by a few percentage points. That will only be a temporary phenomenon, but with the referendum only a few months away we'd need it to recede as fast as possible.

  3. I'd focus on the SNP share of the national total in Yougov UK-wide polls. It is at least partly weighted and represents what they believe the SNP should get in a UKGE. 3-4% (36-48% in Scotland) is the norm recently and rising. Agrees with Scotland only polls such as Survation.

    Yougov group SNP with PC in outputs, but you can largely ignore the PC element due to how small the PC vote is in terms the total UK share. That and Yougov down-weighting of SNP respondents.

    Of course the Yougov Scottish subset does still show Labour falling and SNP rising and has done for some time. Crossover soon.

    Agreed that UKIP will do well in the Euros but have zero chance in the 2015 UKGE. Jeez only 2% of people think a credible economic policy is a reason for voting for them (ICM). All the evidence says a Tory win in 2015 as UKIPers return home having made their protest.

  4. I don't know whether I should be concerned by this, but I've noticed that the figures for MORI and ScotCen research's figures for "definitely decided" are both Yes: 20% No: 42%. Could that potentially back up MORI figures? TNS had Yes: 21% No: 30%. Are TNS underestimating the No vote somehow?

  5. Do you have a link for the ScotCen figures, Calum?

  6. (on page 4)

    Even though the figures are quite old now, it's quite disconcerting to me that they show similar results. It could be seen as the old ScotCen figures backing up MORI, although why TNS as an offline pollster would be finding such different results for no support is a mystery.

  7. Thanks for the link. We shouldn't forget that at the time that poll was in the field, TNS were showing a much bigger No lead than they are now. So while the ScotCen figures might be said to back up Ipsos-Mori to some extent, they don't really contradict TNS in the here-and-now.

    The Ipsos-Mori findings are a concern, though, there's no getting away from it.

  8. The most accurate pollsters for 2010, 2011 and the AV referendum were telephone or face to face pollsters, which doesn't help my concerns either.

    But looking at it from that perspective, the difference between Ipsos and TNS is very mysterious. I can't think of a reason why contacting respondents through door knocking would find fewer strong NO voters than by telephoning (or why ScotCen's pre arranged interviews would appear to back up Ipsos).

  9. The two obvious differences between TNS and Ipsos-Mori is that the latter appear to restrict themselves to interviewing people who have a landline phone (we don't know that for sure, but that's the strong suspicion), and that TNS weight by past vote recall.

  10. I wouldn't count on the tories being hit hardest by another kipper surge since the polls seem to be playing out almost exactly like they did last May with labour getting hit very heavily as May approaches.

    Take a look at the average trend line of all the polling on Wiki for VI and you'll see what I mean James.

    It may sound counterintuitive but Labour really do take a bit of a VI pasting at May elections when the kipper vote peaks. The tories go down but not as much and they pop back up again pretty sharpish.

    The labour tory polling is close right now for a reason and it will likely stay close or get even closer from now on.

  11. MORI's samples do not represent the demography of Scotland at all, both in terms of national identity and country of birth (try comparing with SSAS data and the census). If you re-weight on these you end up with something much closer to other pollsters such as ICM, panelbase and survation. It is a product of the limitations of landline telephone polling in terms of getting a good sample, a lack of past vote weighting and shy factors.

    It is important to remember that small differences in things such as natID and CoB can make big differences to the final Y/N numbers given that we are dealing with a binary question highly sensitive to these factors.

  12. Mick : According to the YouGov poll, 46% of people who voted Tory in 2010 are planning to vote UKIP in the Euro elections, compared to just 17% of people who voted Labour in 2010. In terms of current Westminster voting intention, 32% of people who plan to vote Tory next year will nevertheless vote UKIP in the Euro elections, compared to just 8% of people who plan to vote Labour next year. So if a wave of publicity for UKIP follows their victory in May and some of their Euro voters temporarily transfer to them in Westminster polls as well, it's hard to escape the conclusion that Labour's lead will probably increase - unless of course the UKIP surge is much greater than most people would expect, and they end up overtaking the Tories in Westminster polls for a time. Now that really might help the Yes campaign, but it's not very likely.

    Scottish Skier : The question mark I have over the national identity figures in the census is that the numbers in England were so similar to the Scottish ones - which on the face of it contradicts what we know about how the concepts of 'English' and 'British' are regarded as being more or less interchangeable south of the border. I'm wondering if the way the question was posed (or more likely the way the answers were set out) led to a misleadingly 'un-British' outcome.

  13. I did say it was counterintuitive James and I'm certainly not arguing with the makeup of kipper switchers or that come the GE the tories are going to absolutely get hit hardest.

    I'm also not really making a prediction, I'm saying labour is getting hit right now and the all polls VI trend backs that up.

    The 'why?' is interesting because it can't all be down to the kippers. It's likely little Ed and the spectre of the last time labour crashed the economy. Little Ed really is one of the most unimpressive labour leaders we've had and the only reason that doesn't get more attention is that Cameron and Clegg are also comically inept. Naturally those inside the Westminster bubble are hardly likely to be too keen to highlight that inconvenient fact as the referendum approaches.

    As to how the aftermath of the May elections will play out, we will indeed get Eurosceptic tories running about like headless chickens for a few weeks. However, after that there will be a tory and media omerta on Europe, immigration and Farage with the kipper vote plunging pretty damn fast and the tory vote picking back up just as fast. Why do I think that? Because that's exactly what happened last year when Farage had his best set of locals ever.

  14. "I'm wondering if the way the question was posed (or more likely the way the answers were set out) led to a misleadingly 'un-British' outcome."

    SSAS 2000-2013 averages:

    29% Scottish not British
    33% More Scottish than British


    62% 'Scottish' only in the census.

    Your 33% are just 'passively' not British. They have a British passport, so yeh, they're British citizens but they don't attach much 'value significance'# to that and you'll not catch them waving a union flag at least not with any fervour. Jeez, they support devo max and are listening to indy arguments. However, saying 'I'm not British' implies that they might not like English people or something, hence the split in the SASS you don't see in the census.

    Also, check out this:

    "Imagine yourself meeting someone from overseas for the first time. Regardless of how you plan to vote in the referendum, would you feel more proud introducing yourself as Scottish or British?

    62% Scottish
    19% British
    18% Would make no difference"

    There's yer 62% Scottish again. This time of course it's not really clearly saying there's anything wrong with British, as per the census.

    Your 19% British will be your 17% 2000-2013 SSAS average who answer 'British' in Scottish-British-other' forced choice natID. Folk like Darling ;-)


  15. Compare and contrast.

    Forced choice English or British:

    Kinda even huh. Aye, because British largely = English.

    Scottish doesn't equal British though, hence:


    17% English not British
    15% More English than British

    Aye, because British = English generally.

    All here:

    So the census data looks similar on first inspection, but not when you dig a little deeper.

  16. The strange thing is that TNS seem to back up Ipsos and SSAS on the YES vote, but back up the online pollsters on the NO vote. Trying to make sense of divergent results isn't easy, and why (assuming Ipsos and SSAS are correct) weighting by past vote would reduce the NO vote only is the mystery.