YouGov London poll

I’m a little late with this, but just to flag up that yesterday’s Times had a new YouGov poll of London. Topline voting intentions there are CON 34%, LAB 42%, LDEM 8%, UKIP 9%, GRN 6% (tabs here).

I sometimes hear a sort of London as the weathervane of the country argument, as London goes so the country goes. It doesn’t, despite a Tory mayor it’s Labour leaning compared to the country as whole, and it doesn’t necessarily swing in the same way as the rest of the country either. In 2010 while the Conservatives were seven points ahead in the country as a whole, Labour won in London by 2 points, and the Conservatives managed a smaller swing in London than elsewhere.

This poll suggests a 3 point swing from Con to Lab in London, the equivalent of a 1 point Conservative lead in national polling, so actually a little less than national polls are currently showing. It could be the Tories are doing a little better in London, or it could just be ordinary sample error – as ever, it’s just one poll and shouldn’t be overinterpreted. A three point swing in London would net Labour three or four gains from the Tories – Hendon, Brentford and Isleworth, Enfield North and, right on a knife edge, Croydon Central.

396 Responses to “YouGov London poll”

1 2 3 4 5 6 8
  1. @Unicorn

    I liked it better when you picked a side :)

    I was hoping some application of stats science could provide a bit of certainty, but it seems that stats is just as uncertain as me looking at 4 polls dropping on one day and saying ‘something has moved’. In stats it just seems you pick different methods and come up with different conclusions, and then debate who has the right method.

    Surely there is something more robust?

    It reminds me of all these statin arguments. One day I am being told the stats people have said they will save my life. The next I am being told that last bunch were being paid to promote the drugs, and in fact if you look at the data they provide no benefit. And here I sit, should I take them, or should I not….

  2. Richard

    I had a post slung in automod presumably because I mentioned D*** H*****. Try again:

    If D** H***** was talking about commissioning polls he was probably trying to get some plumbing done.

    Actually it is possible and perfectly acceptable to do such a thing in private polling when you are trying out various lines of argument to see which affects the vote best and that may be what H***** was thinking of. It’s wrong to then publish the poll or any information from it though, if only because you’ll look stupid when they have to publish the tables.

    As Anthony always says, political polling is a sort of loss leader for polling companies, done to build up exposure to attract commercial clients and nowadays maybe to reassure those polled that you are a ‘proper’ polling company who asks the sort of questions they expect a pollster to. So a company isn’t going to risk damaging its reputation and fiddle the figures to pander to the political whims of one client.

    I looked at the recalled 2010 vote on the South Thanet poll and the pattern looks familiar:

    Con 43% (48)

    Lab 35% (31)

    Lib Dem 8% (15)

    UKIP 11% (5)

    Other 2% (0)

    as usual Lib Dems are too low, Lab too high, UKIP way too high. In some of these polls Con is also inflated but not here.

    Survation don’t weight their constituency polls by past vote, but it probably indicates the sample isn’t too far out given the pattern of false recall we’ve seen elsewhere. In some ways it’s probably more accurate that Ashcroft’s simple weighting to 2010 which may boost the Lib Dems too much for example.

  3. Does anyone know of the potential odds on the following happening in the next Parliament:

    A Liberal Democrat MP defecting to another party.

    Caroline Lucas defecting to the Labour Party.

    Douglas Carswell leaving UKIP.

  4. @ Unicorn

    I get a statitsical drop for UKIP in February, but what I am unsure of is whether that could be sampling errors and/or methodological changes by various pollsters.

    If the number of “refusers” and “don’t knows” rises and a pollster weights those responses using 2010 voting then UKIP support is bound to drop.

    A 1% decline in support is surely still within the margin of error?

  5. Mr Nameless
    1) Depends on how many LDs are actually left in the next parliament

    2) I don’t think that’s too likely but it’s certainly not impossible. I’ll give you 30/1.

    3) Evens.

  6. @Funty

    I run a fairly strict access system on my site. Not paranoid, but I get the hosting as part of a quid pro quo arrangement, with the caveat that I manage the site and server myself. Not wanting to cause a fuss, I err on the side of mega-strict. :))

    Sometimes the odd IP address gets blocked due to it being abroad, or via a user agent that ‘looks’ dodgy (some mobile device user agents (browsers) are not familiar to me).

    Suggest you find out your IP address and post part of it. Usually it’s sufficient to post the first two blocks (e.g. 123.123). I’ll find the numbers in the logs and look into it.

    If you’re seeing this some of the time, it may be because you use more than one device (work machine / mobile etc), or you operate with dynamic IPs.

  7. @Mikey

    “So no Tory crossover just yet with Lab inching ahead in the latest YouGov.”

    Crossover postponed for the sixth month in a row, but fear not, the pet theorists will have their day.

    I predict crossover in June, about 6 weeks after polling day!


    P.S. Thanet poll with Farage way ahead and 13% in tonight’s YouGov; I’d say rumours of imminent death of UKIP are somewhat exaggerated.

    69 days to polling day.

  8. @Roger

    Thanks, so the poll looks ok then, in line with what they normally do with their constituency polls.

    Looking at their performance in by election constituency polls last year, they did seem close enough to the final result, a bit UKIPpy in some, but close enough.

  9. Andy Shadrack

    If the supposed decline was due to sampling error, then it wasn’t a genuine decline was it? In so much that the population hadn’t actually changed with any statistical certainty.

    That’s the point of statistics, to identify which claims we can and cannot make instead of turning the graph through an angle squinting hard and prefacing any claim with “I think…”

  10. Unicorn

    Read your post about treating each pollster independently and so forth. I think that would be a saner way to proceed as if you just ran a regression using all the pollsters in a single regression effectively you are treating variations in systematic bias (“house effects”) as a sampling error increasing the effective sampling error of each sample and reducing the certainty that anything happened.

    Another sane way to proceed is to try and remove the house effect from each pollster before running a regression (election forecast has a set of house effects on their site as at least an estimation of the size of house effect). If pollster x has a house effect 2% higher than pollster y then if you want to run both data sets in the same regression I’d deduct 2% from each of pollster x’s measurements before running the regression (or vice versa as we are only interested in relative movements here, where you set the baseline is irrelevant)

    For interests sake, what where your two standard errors of the intercept, this might give some sort of idea of the size of the effective sampling error.

  11. @ Alan

    I agree. The longer Labour and Conservative remain deadlocked the fimer the UKIP and Green vote will get.

    You cannot squeeze voters by saying vote fore me “I might win”.

    I predict, based on a months intense observations, that the SNP might actually see an uptick, beacuse if Labour looks like they might lose more Scottish voters may decide they prefer to send more SNP MPs to deal with the Tories than Labour ones.

    Also I think that the southeast might become more of a battleground between Labour, Conservative and UKIP if the South Thanet poll is not a one-off.

    But I also continue to see that outside of the 57 seats the LD vote has collapsed. That is evidenced by by-election results and constituency polling, so why does the UK wide polling contradict that evidence?

  12. @ Andy

    I get a statitsical drop for UKIP in February, but what I am unsure of is whether that could be sampling errors and/or methodological changes

    As @Alan has already pointed the application of a statistical test has established that there is a relatively low probability that this particular pattern of results could have occurred merely as a result of sampling error. With an acknowledged false positive rate of 2-3%. Given that it takes two months to gather the data to make month-to-month comparisons, you would get a result of this kind about once in every seven years of poll-watching. You might prefer to believe that the effect is just random: that a once-in-7-years event has just occurred. But the case I am presenting is that this stretches credulity,Manx so it is more plausible that something really is afoot. The argument would be more compelling without the hints of cherry-picking and if other statistical test pointed to the same conclusion. But this is a morsel of evidence that it is not just random.

    Could the VI change have occurred as a result of methodological adjustments and not just because of real reductions in Ukip support? Yes, of course. @AW added precisely this is one of the caveats in his original post. Notice that all I referred to in my own comment above was a ‘drop in UKIP VI’. I was only saying that there was a smidgen of evidence that the polling figures have gone down over this period, and this doesn’t rule out the possibility that part of that change is due to methodological adjustments.

    This is why I’d be more comfortable about supporting the claim if it were corroborated by a variety of different tests, and not just the one I mentioned. For example, as far as we know YouGov hasn’t changed its methods over this period. (They started prompting for Ukip a little earlier). If linear regression on the VIs of their five polls a week had shown a reliable drop, then that would have argued against such methodological accounts.

    At present the evidence is all too flimsy to say much with any great confidence. I just wanted to acknowledge that contrary to what I was writing yesterday, there is at least some basis for saying that there has been a real Ukip VI drop in recent weeks.

  13. very compelling…. labour don’t have the lucidity to point this out in as succinct form as our poster a few hours ago….there’s no doubt that 1 and 3 resonate with ukip voters….labour could make more headway on 2) but lack so much confidence on the economy that they daren’t.

    1) “Cast-iron” guarantee of a referendum – promise broken
    2) Abolish the deficit by 2015 – promise broken
    3) Reduse immigration from hundreds of thousands to tens of thousands – promise broken

  14. @ Unicorn

    Thank you for your patience and explanation. having ruled out sampling error, has anyone reviewed whether the number of “don’t know” and “refusers” has increased.

  15. @Richard

    I was hoping some application of stats science could provide a bit of certainty, but it seems that stats is just as uncertain as me looking at 4 polls dropping on one day and saying ‘something has moved’. In stats it just seems you pick different methods and come up with different conclusions, and then debate who has the right method.

    Yes – that’s exactly what does happen in applications to extremely uncertain phenomena. No one gains in any way by turning a blind eye to uncertainty when it is present. However, once things stabilise a bit it should be possible to use tried and tested methods to confirm what is happening.

  16. @ Andy

    Not that I know of.

  17. @ Alan

    Quick answers…

    Single regression: fair point. Loads of dummy variables given that ‘house’ is not a numerical variable, but that’s not s problem. I’ll give it a whirl.

    Correcting for house effects: I don’t expect most contributors would be all that convince by effects thwt depend of prior data massaging. Better to avoid having to use special adjustments just as it would be better avoid having to use a non-standard form of statistical analysis to extricate an effect.

    Finally, SEs were high. As I keep saying they were questionable tests. If I remember correctly, three of the pollsters published just one poll in each of the months under examination. Within-MoE changes could have completely altered their estimates of Ukip VI drop. We are certainly not talking about robust evidence here.

  18. Alicorn now needs to shift to horizontal position…zzzzzz

  19. Just for curiosity, where can I find historical elections data of London for Westminster?

    Thanks to anyone who would answer me

  20. Scottish crossbreak in today’s YouGov

    SNP 42% : Lab 24% : Con 17% : LD 5% : UKIP 7% : Grn 4%

    Mean of Scottish crossbreaks in last 15 YG polls

    SNP 43% : Lab 26% : Con 17% : LD 5% : UKIP 5% : Grn 4%.

    I have altered the time period, to match that of the Populus analysis (3 weeks).

  21. CB,

    Yes pet theories with mine being a cross over in March I guess.

    Based on no Council Tax and Water Rates being paid in Feb and March which for most people is a boost to disposable income giving the impression of being better off.

    Then we get tax threshhold reductions in April.

    All of which imo will give a modest lift to the Governing Parties, in particular the cons.

    The added ‘help’ that oil price falls have provided for the Tories point to a Con lead in the GE but narrowed by some ABTs moving back to Labour in the final weeks.

    All very rational but then so are other pet theories.

    Another I have is that Tim Sherwood will not be a succesful manager – Villa may stay up this year as they should with their squad but long term you need somebody who can do more than telll people how good he is?

  22. Jim Jam
    “Then we get tax threshhold reductions in April.”

    Umm, are you sure about this?

  23. @richard

    Statistics is fundamentally about uncertainty! I would suggest that most statistical work around polling should really be to say ‘we don’t know’.


    Of course the counter argument is that Water bills are due at the beginning of April, and Council Tax comes back with vengeance. You can still hope for better weather though!

  24. I mean tax reductions due to threshold increases, Mike.

  25. @oldnat

    ”Mean of Scottish crossbreaks in last 15 YG polls

    SNP 43% : Lab 26% : Con 17% : LD 5% : UKIP 5% : Grn 4%.”

    Interestingly compare your rolling average with actual regional list votes for Holyrood in 2011, which were:

    SNP 44%; Lab 26.3%; Con 12.4%; LD 5.2%; Grn 4.3%; UKIP 1%.

    So, is 2015 the year when the ”Scots vote differently at Westminster compared to Holyrood” mantra finally ends?

  26. THe Sheep,

    I agree April/Matt not as positive for the Government in disposable income terms – my main argument is for a March cross over based on a modest number of people swithing their vote.

    It wont be rational (Oh I have more money to spend arent the Tories good afterall but subtle affecting the so called feel good factor or feeling a little less bad factor)

    Pet Theory I know.

    And I did not mention the budget which has the potential to impact a GE result either way more than any for a long time.

    GO has to be careful, though, as enough voters are aware of Election bribes to make it a tricky judgement for him, as more than ever this will be a political budget of course.

  27. @ON

    Thanks for the continuing crossbreaks for Scotland. It is still looking reasonably healthy for the Nats, although if I have understood correctly, JM still seems safe in his seat. So all is not lost for LiS! (Ahem!)

    Labour still seems to be making little impact in Scotland. Is there a lot of work going on under the surface, in the hope that this will produce positive results at the right time – i.e. in the last couple of weeks before May 7, or is panic beginning to set in?

  28. And regarding the thread, only four gains by Labour from the Tories in London doesn’t seem to me to be a good enough target….

    I lived for a while in Croydon, and although Croydon South would vote for a stuffed dummy dressed in blue, and Croydon North likewise for one in red, Croydon Central is one of those seats which can be very indicative of the way things are going… though doesn’t always go with the winning party.

  29. @ROH6

    The ‘mantra’ may well not function this year, but I fully expect the Holyrood and Westminster voting patterns to separate again in the future, once the constitutional situation has been dealt with (for the time being). However, the present proposals are way short of what is needed, and so we may have to wait until 2020 for ‘normal service’ to be resumed. It will be interesting to see what happens at the next Local Authority elections in Scotland. My guess is that Labour will have some positive indications, but that’s a while off yet….

  30. I don’t think Labour is a realistic chance for South Thanet simply because (from memory of all the polls in this constituency) if you divide the Combined Con and UKIP scores by 2 Labour still remains behind in all of them. Perhaps Labour has a chance of picking up a few of those UKIP votes but just as likely to help Tories as it does Labour.

    Had Labour been a clear second they might have been able to do a squeeze on other voters but the picture is too mixed for that to happen.

  31. Not sure if Spearmint has a twitter account but if she does I’m looking forward to her tweet teaser of the churn analysis this month. If she goes with “interesting” then I’ll start to get excited.

  32. The tuition fees announcement today looks interesting. The Tories announced policies benefiting pensioners earlier this week, now we get Labour with policies to help the young.

    Generational inequality has been a bit of a theme recently, and it seems the parties are positioning themselves with the Tories as the party for the old, Labour as the party for the young. We know older folk vote more, and we know the young have recently moved to the Green party or may possibly be signing up to the Russell Brand non participation model. Will this announcement change that at all?

    And we had a new report from the electoral commission earlier this week that showed there were still lots of missing students and attainers from the electoral register

    But they say they have 2m new applications since that report, so it is not clear if that has been corrected or not.

    Page 31 shows some of the places where there are large drops – Reading for example is 8% down, and I think there are some Labour target seats there. Large student population there, so that is probably the reason.

  33. Very interesting 36.27 minute discussion with Keiran Pedley, Feb 11th, 2015, between Rob Vance and Lord Foulkes on PodBean, about the need to better regulate the pollsters and the clear deterioration in polling standards:

  34. @ Alan

    Just a quick update to say thaI have carried out a modified multiple regression including house effects as well as polling date and it turns out that it doesn’t improve things. In fact the p-value for the Ukip drop gets slightly higher.

    It still remains the case that the Ukip VI drop since Jan 1 remains unreliable using most forms of analysis and so far only shows up as significant in analyses treating pollsters as providing independent estimates of the change.

  35. @,John B
    “Croydon Central is one of those seats which can be very indicative of the way things are going… though doesn’t always go with the winning party.”

    I know the constituency well. This should be a comfortable Labour gain. I would be very surprised if the Tories retain it.

  36. Richard
    “Generational inequality has been a bit of a theme recently”

    I think you’ll find that Lab is perceived as string on equality/fairness issues. So, the tuition fees announcement plays into that. As does other recent announcements (eg on tax avoidance/evasion).

  37. Net approval below -20 again.

  38. Never mind, Colin. I’m sure your net approval rating willpick up soon.

  39. It is doing Norman -thanks :-)

  40. @Richard

    Another model with built in swingback.

  41. As a Croydon dweller, I would expect Croydon Central to go Labour given the Ashcroft poll and the 2014 local election results.

  42. @Andy Shadrack

    re Polling regulation.

    As a general rule if Lord Foulkes is in favour of something I’d need to be thoroughly persuaded not to automatically back the opposite option.

    See his 1981 proposed law on “Control of Space Invaders and Other Electronic Games” to prevent youngsters with “glazed eyes” from falling into “deviancy”:

  43. @Richard

    Thanks! Although the best bit is probably the interactive Sankey diagram

  44. @RAF

    “Another model with built in swingback.”

    Unless I’m missing something it’s not adding that secret swingback ingredient.
    Of course, some of the donor polls incorporate swingback in one form or another, but I don’t think the Nudigran is adding extra

  45. Interesting that the article accompanying the new Grauniad model “How David Cameron could win the general election but not the keys to No 10” embraces the discussion we’ve been having on this site for some time about the need for a government to have a coalition that hits the 323 mark and not just parroting the Largest Party Wins narrative.

    Of course the article is written by the data people at the Guardian who probably understand this perfectly well but hopefully the message will filter out to political journalists eventually as well…

    Maybe someone should send a copy of it to Andrew Rawnsley.

  46. @RAF

    They specifically state it is a projection on current polling rather than a Prediction of any future movement.

    Can’t see any element of Swingback built into the modelling at all.

    “The Guardian’s average of polls aims to best represent a snapshot of the polls as they stand. The model aims to translate what polling figures would mean in terms of seats if an election were to take place today (and, of course, if the polls were right).”

  47. Good morning all.

    I’m finding it remarkably difficult to distinguish Scottish polls from Scottish cross breaks…They all look the same!!

  48. Professor Curtice has updated his Poll of Polls for Scottish Voting Intention here:

    Voting Intention:
    SNP 45
    Lab 29
    Con 15
    Lib Dem 4
    Green 4
    UKIP 3

    This translates to a seat projection of:
    SNP 45
    Lab 12
    Con 1
    Lib Dem 1

    “If YouGov can find SNP voters in England Survation can find Cornish Nationalists on the Isle of Wight.. There’s Mrs Trepolpen in Ventnor for a start.”

    And many Grauniad readers who have retired to Tresoddit.

1 2 3 4 5 6 8