Six weeks to go

Here are this week’s polls:

Opinium/Observer (19/3) – CON 36%, LAB 33%, LDEM 7%, UKIP 14%, GRN 6%
YouGov/S Times (20/2) – CON 33%, LAB 35%, LDEM 8%, UKIP 14%, GRN 5%
Survation/MoS (21/3) – CON 30%, LAB 34%, LDEM 10%, UKIP 17%, GRN 3%
Populus(22/3) – CON 31%, LAB 33%, LDEM 9%, UKIP 16%, GRN 5%
Ashcroft (22/3) – CON 33%, LAB 33%, LDEM 8%, UKIP 12%, GRN 5%
ComRes/Mail (22/3) – CON 35%, LAB 35%, LDEM 8%, UKIP 10%, GRN 7%
YouGov/Times (23/3) – CON 34%, LAB 33%, LDEM 8%, UKIP 14%, GRN 5%
YouGov/Sun (23/3) – CON 34%, LAB 34%, LDEM 8%, UKIP 12%, GRN 6%
YouGov/Sun (24/3) – CON 35%, LAB 35%, LDEM 8%, UKIP 12%, GRN 6%
Survation/Mirror (25/3) – CON 32%, LAB 33%, LDEM 8%, UKIP 18%, GRN 4%
YouGov/Sun (25/3) – CON 34%, LAB 35%, LDEM 8%, UKIP 12%, GRN 6%
Panelbase (26/3) – CON 34%, LAB 34%, LDEM 5%, UKIP 15%, GRN 6%
YouGov/Sun (26/3) – CON 36%, LAB 34%, LDEM 7%, UKIP 13%, GRN 5%
Populus (26/3) – CON 31%, LAB 33%, LDEM 9%, UKIP 16%, GRN 5%

It’s been a busy week in terms of voting intention polls – ComRes have now moved to weekly polling for the Daily Mail, Survation did two ones (one for the Mail on Sunday and one for the Mirror) and we got the first UK poll from Panelbase. Five of the polls showed dead heats between Labour and the Conservatives, there were three Tory leads and six Labour leads. The bigger picture remains one of the two main parties being neck-and-neck, but there have been slightly more Labour leads than Tory ones in recent polls, so the UKPR polling average this week has Labour one point ahead – CON 33%(nc), LAB 34%(+1), LDEM 8%(nc), UKIP 14%(nc), GRN 5%(-1).

Scottish and London polls

ICM had new Scottish and London polls out this week. In Scotland they found Westminster voting intentions of CON 14%(+1), LAB 27%(+1), LDEM 6%(nc), SNP 43%(nc), UKIP 7%(nc), GRN 3%(-1), changes are from their previous Scottish poll in December. At 16 points ICM show a slightly smaller SNP lead than some other companies, but there is no significant change from their previous poll, suggesting its something methodological rather than a narrowing of the SNP lead.

This morning ICM had a London poll for the Guardian. Voting intentions for that were CON 32%, LAB 42%, LDEM 9%, UKIP 9%, GRN 8%. That represents a four point swing from Conservative to Labour since the general election – the equivalent of a one point Labour lead in national polls – so again suggests that the swing in London is much the same as in the rest of the country.

Week 12

  • David Cameron ruled out standing for a third term as Prime Minister. Unusual not because of the content – if he wins he was widely expected to stand down at some point after the European referendum anyway – but because he said it, out loud, to a journalist. In terms of public opinion 55% of people said Cameron was right to rule out a third term, 18% wrong. A majority of supporters of all parties – including Tory voters – thought it was the right thing to do. 21% of people said it made them think better of Cameron, 9% worse of him, but for the majority of people it made no difference to how they viewed him.
  • The final PMQs of the Parliament was dominated by exchanges on ruling out tax rises. Asked before Cameron ruled out a VAT rise and Ed Balls ruled out a National Insurance rise, at the start of the week YouGov found 43% of people expected tax to go up if Labour won, 29% expect it to go up if the Conservatives win. Under a Labour government, 43% expected income tax to rise, 41% expected fuel duty to rise, 39% expected national insurance to rise… but only 22% expected VAT to go up. Under a Tory government 34% expected fuel duty to rise, 31% expected VAT to rise, 29% expected NI to rise and 25% expected income tax to rise.
  • The debate debate finally came to an end with an agreement to have four events: a Paxman interrogation of Miliband and Cameron; a seven-way debate between Cameron, Miliband, Clegg, Farage, Uncle Tom Cobley and all; a debate between the five opposition parties and a Question Time special with Cameron, Miliband and Clegg, one after the other. The Paxman interrogation took place last night. An ICM poll straight after the debate found people thought Cameron came out better than Miliband by 54% to 46% – we will have to wait until the weekend to see if it has any impact upon either voting intentions or perceptions of the leaders. Over the last five years Cameron has consistently had better ratings than Miliband, so in many ways a performance that’s pretty even has the potential to help Miliband far more than Cameron. As ever, time will tell.
  • The physical mechanics of the general election have started to kick in. Yesterday Parliament was prorogued, on Monday it will be dissolved and the writ issued and we’ll be off. The start of the formal campaign means various bits of regulation kick in, including the broadcasting restrictions requiring coverage of the main parties and spending limits upon the parties.


The latest forecasts from Election Forecast, May 2015, Elections Etc and the Guardian are below (the Polling Observatory team are doing fortnightly predictions, so nothing new from them this week). Three of the models continue to show the Conservatives with just a few more seats than Labour, but Steve Fisher’s prediction from Elections Etc now has them 35 ahead of Labour. This is due to a methodology change rather than a move in opinion – Steve’s model for predicting the vote shares in England & Wales remains unchanged, but he’s no longer assuming such a big drop in SNP support in Scotland, and has rejigged how he translates projected votes into seats based on Ashcroft and YouGov polling (it’s explained in more detail here.

Elections Etc – Hung Parliament, CON 296(+12), LAB 261(-17), LD 21(nc), SNP 47(+6), UKIP 5(+2)
Election Forecast – Hung Parliament, CON 283(-1), LAB 280(+2), LD 26(+1), SNP 38(-2), UKIP 1(nc)
May 2015 – Hung Parliament, CON 273(-4), LAB 271(+3), LD 24(nc), SNP 55(nc), UKIP 5(+2)
Guardian – Hung Parliament, CON 277(nc), LAB 269(+1), LD 25(nc), SNP 53(-1), UKIP 4(nc)

367 Responses to “Six weeks to go”

1 2 3 8
  1. First

  2. As usual I have compared the VIs in Anthony’s current (unusually large) batch of polls with the VIs predicted on the basis of 2014 data. The only change compared with last week is that for the first time the Greens are now polling reliably below their long-term trends.

    The summary for the 12 different Polling Average batches this year so far is as follows:
    CON ( = + = = = = = = = + + + ); LAB (= = = = = + = + = = + + ); LD (= = + + + + + + + + + +) ; UKIP ( – = – – – – – – – – – –) and Greens (= = = + = = = = = = = –)
    [= implies on trend, + above trend and – represents below trend]

    For the first time since I have been tracking these trends, all five parties are deviating from a linear progression. In four of the five cases (all except Labour) the departures are in the direction predicted by swingback (that is, the two parties above their 2010 share are dipping below trend and the other two are exceeding their trend VIs). So, perhaps we can tentatively say that we may now be seeing a gentle start to the long-awaited swingback. The exception to this pattern lies in Labour’s recent rise above trend. Some possible reasons for this ‘anomaly’ were aired in some comments posted during the week.

    In interpreting the seat projections by the various models it may be worth bearing in mind that relative to the ElectionForcecast model May2015 (and the Guardian model) consistently place more seats in the SNP column, and also consistently understate Tory and Labour seat counts. For the LibDems, there is a tendency for Electionsetc to generate lower seat projections than the other three models.

  3. The models are definitely starting to bunch together in their predictions.

  4. This isn’t any sort of loaded question but does anyone in here think that a Lab + LD + SNP coalition (either formal or informal) is a viable, manageable government which could last more than a year or two before falling apart?

    Reason for asking is that of the 4 forecasts on this article, only 1 suggests that lab + SNP can get a majority together. Just. To me, a con + ld + dup coalition is infinitely more viable as the third party are much much smaller and the largest party will probably be larger

  5. The SNP YouGov crossbreak noise is interesting but seems to contradict the Scotland wide polls and today’s 4 byelections overall would probably represent a good return for SNP.

    I expected an SNP win in Glenrothes although not such a large margin.

    In the other 2 I thought Independents might make it through on transfers but Buckie wasn’t close (perhaps a reverse celebrity effect for Norman fae Bake Off) and Armadale & Blackridge not quite enough Borrowman supporters were willing to back Mackay.

    Uist a Tuath is a rule unto itself but Independents tend to do well in Eilean Siar.

    School closures have been controversial in Eilean Siar with SNP Councillors generally voting against but this obviously didn’t affect the result here where the planned new Uist a Tuath school in Paible seems relatively popular despite the overall reduction in number of primary schools.

    Interesting that the Scottish Government chose not to announce whether it was calling in the closure decision for Lionel School S1-S2 (not in this ward) until after the election. Decision has to be made by early April so can’t be dodged until after the GE.

  6. Unicorn

    To be fair 4 out of 5 ain’t bad.

    It might actually be more likely as there is 5 ways that 4 out of 5 could be arranged.

  7. The DUP are on record as preferring an arrangement with Labour, not the Conservatives.
    It’s very difficult to see how Cameron is going to get past 320 – the numbers just don’t seem to be there.

  8. ADAMB
    Reason for asking is that of the 4 forecasts on this article, only 1 suggests that lab + SNP can get a majority together. Just.

    You should add about 6 SDLP, PC & Green seats onto the Lab side, but with some sources suggesting that SDLP won’t take the Lab whip next time but will give C&S as part of the progressive alliance.

    I have a feeling that LD won’t be able to afford to go into coalition with anyone next time as their survivors will need the “short money” to help re-build their fortunes.

    I can’t see a coalition being considered by any of the parties I list.

  9. David

    It depends what LD get. If 30 seats say then con + ld is viable and potentially they could include dup without much bother. It just seem to me that you could easily end up with lab on say 275, SNP on 40 or something else in the 315-320 range

  10. UKIP is 10% to 18% – very suspicious!

  11. So – I have been studying the graphs of polling in the run up to each election for the last 30 years or so. It is indeed the case that swingback towards the party of government is common. But it generally happens over the last 6 months ago rather than the last 6 weeks. Infact there are multiple examples of the party of government support dipping in the last few weeks 2001 and 2010 are examples. So has the swingback peaked? Was that it?! What is the conventional wisdom on this?

  12. Adam B

    “coalition (either formal or informal)”

    I suppose it’s possible that the LDs would choose to give up the “Short money” advantage of being outwith government, in return for a Ministerial car or two, but for any other party it would be foolish.

    The viability of any arrangement to keep a minority party in government will depend purely on the parliamentary arithmetic that emerges on 8th May – and the willingness of one or both of the largest parties to try and govern under such an arrangement.

  13. The LibDems will get less than 20, maybe even less than 10.

  14. @Simon

    Evidence? (or merely wishful thinking?)

  15. @AdamB

    I thought Tim Farron’s comments in the New Satesman were very interesting where he basically said he thought the Lib Dems should support the side which would be a more stable long term government.

    I therefore don’t think Lib Dems would sign up to something where the numbers only add up to 32x and could be undermined quickly by byelections or single vote defections.

    I think he would back (though likely only via C&S) whichever block would have 33x or 34x seats by the time you add on Lib Dems.

    So for example if it were:
    Con 292
    Lab 260
    SNP 45
    Lib Dem 25
    PC 4
    UKIP 4
    Green 1
    Speaker 1
    DUP 9
    Sinn Fein 5
    SDLP 3
    Sylvia Hermon 1

    He’d view 326 (Con/DUP/LD) as being too unstable and prefer 339-347 (Lab + rest-UKIP/SF) as a better long term bet for government.

    If it were more like
    Con 302
    Lab 250

    Then he’d be looking at Con.

    Given the likely smaller number of Lib Dems I therefore think 300 is the absolute minimum Cameron need to form a government. At 290 the numbers might be there (just) but I’m not sure Farron would go for it.

  16. Simon.
    The wide spread of UKIP polling has been going on for months. Best to look at the individual constituency data from their target seats – that tells a very different story.

    If the Libs have as bad a result as everyone is expecting, then Clegg will be replaced by Tim Farron and he will have no truck with the Tories.

  17. I doubt all these models. For instance I still think that unless the polls show a big increase in LibDems vote share, they will be lucky to get more than 12 seats. I know they are supposed to have special incumbency bonuses, but as things stand they will get about as many popular votes as the Liberals did in 1970.
    Then they stood in just over 300 seats and won 6. To get 12 this times they will have to do four times as well – i.e. their vote will be spread over twice as many seats (and hence averages less in each one) and yet they will have to win twice as many seats.

  18. OLDNAT
    the willingness of one or both of the largest parties to try and govern under such an arrangement.

    I thought there was a real chance that Paxo would ask Ed the question of whether he would try to form a government if Lab were not the largest party, but clearly his disdain for things Caledonian overcame his desire to give Ed a “kicking”.

  19. @ Panther


    With a username like yours I’d be disappointed if you weren’t :-)

  20. @AdamB

    The Lib Dems winning 30 seats?

    If they lose 5% of there vote to the second party, they keep 30 seat (just).

    If they lose 8% to the second party, that leaves them with 15 seats.

    They look like needing a large amount of nose-pinched tactical votes to do as well as 30.

  21. sorry, ‘their vote’.

  22. One thing I can’t get my head round is that whatever the variation in the UKIP vote between pollsters it doesn’t seem to make any difference to the gap between Lab and Con. I’ve felt this all the way through this parliament and another example is the London Polling that Anthony quotes- same swing despite UKIP being way lower in London than elsewhere.

    And yet we have far more ex Tories than ex Labour in the UKIP VI. Maybe Spearmint with her churn can work out why this is.

  23. UKIP pull in a lot of previous non-voters. Could that have something to do with it?

  24. So a relatively good polling week for Labour but they are still playing catch-up on the seat model projections.

    We’ll soon see whether Crosby was right and that the Tories will begin to make things happen after Easter, or whether Labour’s mini-revival continues (or the third and most likely option of polldrums until election day).

  25. @Pete B

    If Ukip is relying on non-voters they won’t win too many votes ;)

  26. @ Catmanjeff

    From previous topic

    It’s the Survey of Carers in Households.

    The Health&Social Care Information Centre has some of the reports ( It use to be published by ONS.

  27. @ Catmanjeff


    Tablet autocorrected.

  28. @Shevii

    Interesting observation! But it turns out that your informal observations are wrong.

    Using all polls since Dec 1st in Anthony’s list (at the top right-hand of this page) following your suggestion I have run a regression of Labour,s margin over the Tories against the Ukip VI, and it turns out that there IS a very reliable positive relationship between the two [t (157) = 3.56, p less than 0.001].

    That is (in line with the pattern you were looking for) the higher the Ukip VI the bigger Labour’s margin over the Tories. This is compatible with most churn analyses, and confirms that when Ukip support rises then this damaged to Tories more than Labour.

    So, no puzzle here really. Exactly the pattern you were expecting. It is just that you have to analyse quite a few polls to detect the pattern in all the noise.

  29. The trends discussed above can be clearly seen in this updated graph:

    Conservatives and Lib Dems are experiencing mild swingback

    UKIP and Greens are clearly declining

    For Labour’s trend, I propose the term “swingforward”

  30. @Northumbrian Scot
    “The models are definitely starting to bunch together in their predictions.”

    Apart from Fisher’s which has acquired a sense of bravado. The question is: is he ahead of the curve of behind it (or the most likely option that his model has simply entered a Pluto-like eccentric orbit).

  31. @Omnishambles
    For Labour’s trend, I propose the term “swingforward”

    Very topical, Sir :)

    However, I believe others got there first with the term “reverse swing” which pleasungly is the polar opposite to “swingback”.

  32. Catmanjeff
    Yes, I said LD to get 30 because, whilst I would never vote for them, I think there’s a reasonable chance that polls are understating their position, simply because it’s unfashionable to say that you’d viote fir them now. I think they’ll l do better in May and ukip worse

  33. @Shev


    The best-fitting slope indicates that there is an increase of 0.26 units in Labour’s margin over the Tories for every one point rise in Ukip’s VI.

    This corresponds to 63% of Ukip losses going to the Tories and 37% to Labour. Equivalently it suggests that when Ukip support rises it draws 63% of its support from the Tories. (Note that 0.63 minus 0.37 = 0.26 which is the original slope figure).

    These figures are fairly similar to those reported by the Churnologists a couple of months ago. More recently, there have been hints that Ukip support comes from the two major parties on a more equal basis. My analysis throws everything together hiding the possibility of such changes over time.

    Interesting to see this corroboration of churn figures without using 2010 IDs…

  34. From Nick Robinson

    Some good gems in there.

  35. Breaking News;

    In Leicester, more leaked emails from the DWP confirm that Richard III has passed his ATOS assessment and has been declared fit for work.

    More on this later.

  36. @raf
    “However, I believe others got there first with the term “reverse swing” ”
    Damn. I have to admit, it sounds better than the ungainly “swingforward”

    BTW I think Fisher *is* onto something, his seat share looks more likely than the others to me. But I agree with what Unicorn et al said about constantly tweaking your model.

  37. Anthony

    “Survation did two ones…” [sic]

    Is this what grammar school education has come to?

  38. @ Alec


  39. Shevii & PeteB
    The UKIP VI variation between pollsters is so large that it must depend on how the pollsters adjust the raw data, and is therefore independent of relatively small changes to other parties’ VI.
    The spread between pollsters is actually less now than at times in the last year when UKIP’s low end was about 10, but some pollsters recorded over 20.
    The alternative question is, with C and L both recording VIs in the low 30s, so with MoEs both about 3, why are the differences so consistently small? If C and L are really running level, there should be occasional differences as large as 4.
    Last summer I remember commenting on the fact that the C VI over several (10 IIRC) YouGov polls varied as it should if constant but with MoE of about 3, while over the same period the L VI varied very little from poll to poll.
    I suspect that polling panels, necessarily consisting of people willing to answer poll questions, actually exclude those extremes of response needed to make the results match random behaviour.

  40. Alec

    Richard has been working hard for Leicester’s tourist industry for a few years already!

    Royals can be economically useful – dead or alive. Indeed if they can’t write letters to Ministers lobbying for their interests, they may be more useful dead.

  41. @Shev

    Intrigued by that analysis, if have just extended it to the greens using the same dataset.

    Against, this shows the pattern you would expect. For each point rise in Green VI Labour’s margin over the Tories drops by 0.13. This slope is reliably different from zero, but only just so (p=0.014). The smallish ratio, suggests that Labour and the Green are not so tightly linked as some commenters would suggest. Sometimes people seem to argue that if Green support drops then almost the entire benefit will accrue to Labour (implying a ratio much closer to unity).

    On these figures, if the Greens fall back to 3% by May 7 the boost to Labour will only amount to a quarter of a point.

  42. @Alec

    Love it. But his qualifications seem a tad limited. Maybe William and Harry need to watch their backs?


    Re LD seats

    I did research based on 2010 LD majorities, by-election results and 2015 candidates for the purposes of placing bets on LD seat bands. I concluded the most likely outcome is 14 LD seats but I will not be surprised if it is even less than that.

  44. Most people I know say they’ll be voting UKIP!

  45. @Omnishambles
    “BTW I think Fisher *is* onto something, his seat share looks more likely than the others to me.”

    I find it difficult to see the Tories only suffering a net loss of just seats. It could happen but it’s close to an ideal scenario for the Tories.

  46. @Omni

    “..a net loss if just 15 seats…”

  47. @Omni

    I’m not correcting my spelling/grammar again! Maybe it will happen after all.

  48. @Dave

    why are the differences so consistently small?

    With a large enough sample you can test if the VIs are clinging closer to one another than they should. The MoE for the difference between two VIs should be 1.414 (I.e., root 2) times as large as that for the individual VIs. If the ratios are reliably lower than this, then that would indicate that the pollsters are ‘adjusting’ their their results. A few weeks ago I did some analyses using several hundred individual polls and found no evidence of such oddities.

  49. @Simon

    Nobody I know is proposing to vote UKIP. We must live in different circles of acquaintances.

    And again, you say you’ve researched Liberal prospects. But where’s the evidence?

  50. Studying the bookmakers is again instructive.

    The bookies odds are more sophisticated than people think, and I suspect they are quite a reliable guide to voting intentions. They perhaps represent that elusive combination of what people think now, and what people think will happen on May 7.

    The odds for current Lib Dem seats seem to accurately reflect some of the thinking and polling that we see on this site. In particular, the odds strongly suggest a Scottish meltdown, but also show a big swing to Labour in the North of England, and urban constituencies. There is marked resilience in ‘Middle England’: where the Tories are the opposition, and the Lib Dem vote appears to be holding up well.

    The conclusion appears to be that the Lib Dems will hold on in leafy England surprisingly well, and might secure a total of 20-25 seats. They will be all but wiped out in Scotland and will lose a number of seats to Labour in places like Burnley, Redcar and Cardiff Central.

    Applying a standard swing against the Lib Dems, and apportioning losses pro rata to the other parties is very obviously a big mistake. The Conservatives are unlikely to make many gains at the expense of the Lib Dems – the gains will go disproportionately to the SNP and Labour.

    Talk of a total collapse ( 5 – 15 seats ) of the Lib Dems is not supported by the bookmakers.

    And it is bad news for the Tories who will not make many gains at their expense.

    I don’t explain, I just follow the money…

1 2 3 8