There have been several polls this week showing both the Conservatives and Labour up at 35%, and given it’s a zero sum game that suggests the other parties are getting squeezed. There is a general expectation of this sort of squeeze as the election approaches – the nature of First Past the Post is that votes for smaller parties don’t stand much chance of being translated into MPs unless they are geographically concentrated and as the election approaches the media coverage almost inevitably focuses ever more upon the main contenders (though more on that later).

The graph below shows the average poll scores for the Greens and UKIP across the nine regular pollsters (for those pollsters who do several polls a month, I’ve taken their average across the month).


UKIP have been on a slow but steady downwards trend since their support peaked after Douglas Carswell and Mark Reckless’s defections. This last month the decline may have flattened out a bit, but that is largely due to MORI having an unusually low score for them last month that jumped back up this month – without that the line would show a smoother downwards trend. The Green party’s advance seems to have halted last month and started to fade a little this month.

Of course, just because they might be getting squeezed shouldn’t distract from the fact that UKIP and the Greens are still doing incredibly well compared to the last election. In 2010 UKIP got 3%, the Greens 1% – both parties could suffer a bit more squeezing and still end up quadrupling the vote they got last time. For UKIP, there is also good news on the horizon, next week the campaign broadcasting restrictions kick in, guaranteeing them coverage as a major party. For most of the last few months UKIP’s media coverage has largely consisted of the latest row or resignation for inappropriate comments. Next week the broadcasters will have to start giving them more neutral coverage alongside the Conservatives, Labour and Lib Dems.

In other news, we had the monthly Survation poll for the Daily Mirror out today. Their topline figures were CON 32%, LAB 33%, LDEM 8%, UKIP 18%, GRN 4% (tabs.

384 Responses to “Are UKIP and the Greens getting squeezed?”

1 6 7 8
  1. forecats=forecasts of course

  2. Omnishambles

    I don’t doubt DC will want all options open and personally was OK with the coalition. The difference between nice to have and needing LD support could mean quite a lot in terms of policy giveaways. I can see a lot of backbenchers getting miffed at an “unnecessary” coalition (despite those particular backbenchers being the reason a larger majority is really needed, even Con 323 would be a nightmare for Cameron, he’d really like to get some other parties onside even with a technical majority).

    Coalition possible but if the numbers stack up I can see reasons for running a minority government.

    I agree UKIP are likely to just vote against everything (including a 2017 referendum) and no deal will be done with them/him I think they will be irrelevant in terms of numbers.

  3. Pete B

    Do you know if they use Scottish cross breaks (of YouGov polls) or only Full Scottish polls?

    I ask because there has been some very odd behaviour in the YouGov Scottish sub samples, earlier this week, that are sufficiently odd to render questionable any forecasts based upon them. Namely: an implausibly large and sharp drop in SNP support. I would discount any and all results based upon said cross breaks, until the issue of YouGov volatility in Scotland has been clarified.

  4. @alan

    Good points

    I don’t have anything else to add to that


    More on Glenrothes candidates from the Courier.

    What will be interesting is the 1st preferences counts for the parties standing. Only 5 this time: UKIP, LD, Lab, SNP & Con.

  6. ProfH
    I’m afraid I haven’t delved into the precise methodology they use, though your theory sounds very plausible. I just publish their figures because they update (almost) daily which makes trends easy to spot.

    They have a graph of their forecasts themselves, but it’s not very helpful. I thought the actual figures might be of interest to people here.

  7. @all

    Guys, guys, there’s UK-wide poll out. THAT’S A UK-WIDE POLL!!!! Has there ever been one before?

    Is this f’real? Or have they just done GB and pretended it’s UK?

  8. “Where nothing’s shown there’s no change. So according to EF’s methodology there does seem to be a slow drift to Labour, especially in Scotland. Does this count as swingback?”

    the “drift” to labour is a function of what market analysts call “time value”…the longer the polls are roughly static, in the sense of the difference between the two main parties, (so it doesn’t matter if the two main parties are level at 33% or 35%), the less impact any imputed swingback to the government will have, so if the polls are the same as they are now in 5 weeks’ time, the labour no. of seats, based on their projected vote share in the election, will be quite a bit higher and the tory score a bit lower than today.

    We should see this is Fisher tomorrow. I expect the labour seat number will be closer to the tories’ seat number than it was last week, which again was closer than it had been the week before…this is because his final projection was for tory lead of 3% on month ago. If the polls stay roughly level, that projected lead will diminish and affect the projected seat numbers accordingly.

    BTW, Fisher’s projection from last week had con on 34.3% and lab on 31.6%, a 2.7% difference. the con score could be right, but the labour score to my mind looks too low. I don’t see the collapse of labour many tories were hoping for and expecting at the beginning of the year, when people credibly believed labour might struggle to get 30%

  9. @syzygy

    The link you provide is an interesting way of trying to quantify when the changes in the polls are no longer due to random fluctuations.

    But their EWMA model is a sequence of constant means with step changes between them. I don’t think this is a very good way of modelling polling data since there are times when there is a steady increase or decline over a period which would presumably result in some kind of staircase pattern, a spurious construct not in the data. In fact you see such a staircase in the analysis of the Green VI data.

    It is worth thinking about how to improve it.

  10. @ Richard

    Thank you – I thought that @Catmanjeff would have been pleased to see their results.

  11. Statgeek: are you the eponymous figure behind

  12. Peter C
    Ok, so the forecasters’ models are effectively correcting for an overly Tory-friendly initial forecast.
    Anyway, the end result is a slight drift to Labour, making even a Tory plurality less and less likely.
    Dave may not get his chance for a second term at this rate.

  13. Pete B

    I have studied Electoral Forecasts’s website, which is not super-informative about how they use crossbreaks, but I think my caveat 6.51pm about the YouGov Scottish cross breaks does apply to this forecast.

  14. Pete B

    I lose track what people mean by swingback, there seems to be multiple definitions floating around.

    There small level of change I entirely consistent with an auto regression model.

    It’s also entirely consistent with the same model with the swingback component set to 0.

    Parties are allowed (in fact should) to bounce up and down randomly (unseen news affecting VI is treated as a random variable). The difference is a slight nudge towards a predetermined ‘mean’. If left to run, you wouldn’t see the values converge perfectly to that mean, but would end up oscillating randomly about the mean.

    The drop in UKIP support is certainly predicted by auto regression, the lack of a drop in their support was pretty powerful evidence against auto regression being applicable the weakness of the model is it assumed that it’d come back uniformly but there are issues with making an already controversial model more complicated.

    Labours recent rise isn’t strictly predicted by the model, but it’s within the bounds of “reasonably might happen” it’s an above average outcome for them, but not by so much it shatters the model.

    Parties are allowed to drift above / below their predicted outcomes in the short term. If everyone followed their predicted path exactly it’d be kind of a weird statistical event. (Like tossing a million coins and getting exactly 500,000 heads, yes it’s the most likely event but in isolation a very unlikely outcome)

  15. @ Hal

    Thanks for your thoughts. The technique reminded me of the sort of analyses that some types of traders use to identify shifts in the markets i.e..when to buy and sell (not that I know very much about it but I’ve seen it used).


    Does Fisher’s projection reduce the size of swingback predicted as one moves closer to the date of the election? There seem to be two ways of modelling swingback: one is to assume if it hasn’t happened by now its likely to be smaller, another is to assume that if it hasn’t happened by now, it will still happen just as much, just later.

    [Yes. Steve’s model is based on the degree of change there is at that point in previous Parliaments – i.e. when he is projecting 6 months out from the election, his model would apply the average changes in the 6 months before an election. If he’s projecting 3 weeks out from the election, his model would apply the average changes in the 3 weeks before an election. Eventually it will be assuming no change at all. I expect his model is projecting very little actual swing now, but remember his model also factors in past inaccuracies in polling, and how the polls have tended to be slightly biased towards Labour, so this will still be a factor making his projection more Conservatives than what the polls are currently showing. Last time round the polls managed to correct this problem and were no longer biased towards Labour, and I hope they’ve solved the problem for good and will get the Lab/Con position right again this time… but I do understand why Steve still puts that in the model. The alternative of “the pollsters have got it right once so the problem is solved!” isn’t exactly robust – AW]

  17. “Ok, so the forecasters’ models are effectively correcting for an overly Tory-friendly initial forecast.”

    it wasn’t “friendly” as such. The dog which hasn’t barked this parliament yet is swingback. There has been a bit, but generally there is quite a bit more swingback to the government as the next election draws on, but this parliament has been very strange in that regard.

    The liberals were on 10% as late as the beginning of 2014…absolutely nobody, and i read this site a lot, and have other sources and friends in the political world- nobody thought that they would actually dip from 10% during 2014, as they had flatlined since the end of 2010. Everybody thought they would get to about 15% in the general election, as a result of swing back.

    The tories seem to be inching forward now, but were flatlining on 32% for 18 months. It’s been a funny parliament. The models which built in swingback have had to readjust their numbers all the time. I prefer models which look at averages from the immediate past, say 3 month averages, as opposed to trying to guess future movements by looking at what happened to VI in previous parliaments. The past is another country.

    I haven’t thought Dave would get another bite of the cherry since the omnishambles budget, though the ineptness of labour never ceases to astonish me.

  18. @profhoward
    “Does anyone know if there will be an “instant” poll after tonight’s debates?”

    Yes, there will. An ICM snap poll.

  19. “Does Fisher’s projection reduce the size of swingback predicted as one moves closer to the date of the election? There seem to be two ways of modelling swingback: one is to assume if it hasn’t happened by now its likely to be smaller, another is to assume that if it hasn’t happened by now, it will still happen just as much, just later.”

    He clearly does…the latter approach you describe is mad, as it would demand you have enormous swingback in the last week, if the polls haven’t shifted the way your model suggests before then.

    We know Fisher’s model adjusts the degree of swingback, because those of us who have followed it from the beginning remember that in october 2013, his model projected that the tories would win 337 seats with 40% of the vote…it was scoffed at then, by many commentators on this site, and the scoffers were absolutely right.

  20. ProfHoward

    I believe so, although it’s not linear. The reasons are a bit like a spring, reduce the extention and you reduce the restoring force. It’s not exactly the same maths but the analogy is similar.

  21. @Omnishambles

    IMO it is very unlikely that LDs will go into coalition with the Tories. If Cons are the largest party it almost certainly means that LDs will be down to 20 – 25 MPS (unless there is a large collapse in the Lab share). With such a low number of MPs it is highly unlikely that Clegg would survive as party leader even if he retains his seat. I doubt that Farron, the most likely replacement, would be prepared to join with the Cons.

  22. Peter Crawford

    It predicted 40% for Con????

    We can infer it’s not a normal auto regression model then. Not sure what else is in there to get that result.

  23. Thanks Peter and Alan. Agree with Peter that the “latter approach is mad” so in a sense I was doing a “sanity check” on their method :)

  24. Why is Bercow hated by his own side? Its odd to have Lab MPs lining up to support the speaker who is a Con MP. Meanwhile Con MPs want to replace him with a Lab MP as the new speaker. Bizarre.

    Some person on here kept saying the LD looked a little high from Late 2010.

    As PETER BELL has said above; if the Lib Dems fall badly it is good for the Tories.

    Feb 74: The old Liberal party strength helped Harold Wilson’s Old Labour Party, as did Ted Heath’s ‘betrayal’ of the Unionists when he insisted on Civil Rights in the north of Ireland.

  26. new thread alert on Panel Base poll

  27. Mikey

    From all accounts he was barely on “their side” and likely would have crossed the floor if not elevated to speaker.

    It’s not surprising he gets a fair deal of support across the parties but certainly the right wing rump of the Tories don’t see him as “one of them”

  28. @ Mikey I have a vague memory (though I have to add I was a child at the time) of the reverse happening with George Thomas who was a very popular speaker with the Conservatives.

  29. @ Unicorn

    Thanks for the answer.

    You have convinced me to place more money on a Tory victory as you clearly state that the Conservatives will win a majority.

    But any chance you could email me your name and address in case your wrong.

  30. JOHNB. At 5.03

    As long as I can remember the young have been predominantly left. Many people move to the right as they grow older. I do not believe there has been a permanent shift.

  31. Chris in Cardiff.

    I’m old enough to remember George Thomas. Perhaps some of his Labour colleagues knew a bit about his personal activities!

  32. Surely we should ignore fishers model?

    he initially projected the tories on over 40% IIRC – and them getting a decent majority. That was widely seen as tosh by many people on here – and so it has proved. I cant remember what he had labour getting – but it was pretty low – and this was before the rise of the SNP (which nobody predicted)

    The fact that model gets steadily ‘readjusted’ in the light of the most recent polls is having your cake and eating it. If the polls are right the model as originally set out has demonstrably failed.

    As for the lib dems – again a fair few of us were arguing over a year ago that their vote was not going to start rising again and that projections of them getting into the mid teens were daft.

    The polls over the past three years have shown the tories marooned in the low 30s, with a slow but steady climb over the past 3 months as UKIP has (predictably) slipped back a bit as the election looms – but nothing suggests that the tories are going to suddenly shoot up.

    Similarly – and also predictably – over the same period the labour vote has picked up as green and some UKIP has drifted back to labour.

    All this says that big swings are very unlikely between now and may 7th and the final result will be something like 35/35 with maybe a percentage point either way.

  33. @Alan

    You are right. There could come a ‘UKIP tipping point’ at which those who are currently expressing an intention to vote for them realise it’s a ‘lost cause’ in seat number terms and flee.

    If / when that happens – and let’s remember people continued to vote in larger numbers for the SDP/Liberal Alliance and Lib Dems than currently do for UKIP when they did not appear to have a chance of forming governments – then much hinges on how those ex-UKIP supporters break.

    Do they go en masse to the Conservatives? Do they abstain? Or do they split more disparately, thereby negating the overall effect on the final outcome?

    For all the churn analysis, second preference, consideration questions etc I genuinely don’t think we have a clear idea of what the consequences of a sharp UKIP fall would be.

    Another element that adds to the uncertainty.

  34. @Reggieside
    “..and this was before the rise of the SNP (which nobody predicted”

    Well no-one except @Couper. She’s now the gold standard for UKPR Scotland Westminster VI predictions.

1 6 7 8