Final polls of 2014

We now have the final YouGov and Populus polls of the year (possibly the last two polls of the year, unless something unexpected turns up). Topline figures are

Populus: CON 35%, LAB 35%, LDEM 9%, UKIP 12%, GRN 4% (tabs)
YouGov/Sun: CON 32%, LAB 36%, LDEM 6%, UKIP 16%, GRN 5% (tabs)

The YouGov four point Labour lead is interesting, coming as it does after that odd looking five point lead last week – normally I’d write something along the lines of keeping an eye on the next few polls to see if YouGov are picking up some movement towards Labour…but of course, the next few polls aren’t until January. In contrast Populus aren’t showing any such move, with Labour and the Conservatives neck and neck.

Before signing off for the year, I also wanted to flag up some new British Election Study analysis from Phil Cowley here, looking at where people put the parties on the left-right spectrum in Scotland, something I don’t recall ever having seen before. Essentially it finds people put the two main parties in Scotland in the same place ideologically, on a 0-10 left right scale they rate Labour at 4.1, the SNP at 3.9. Of course, averages don’t tell the whole story as different people see the parties differently – looking at Labour’s Scottish voters, they rate themselves as 3.4 on the left-right scale, the Labour party as 3.4 and the SNP as off to the right on 4.9. For SNP voters, they rate themselves as 3.6 on the scale, their party as 3.8 and the Labour opposition as off to the right on 5.3. Both parties’ supporters see themselves as left-wingers supporting a left-wing party against a more right-wing opposition.

And that’s it – I may do an end of year round up… or may just put my feet up. Either way, have a good Christmas.

120 Responses to “Final polls of 2014”

1 2 3
  1. @ Bramley

    Thanks for the link. The interesting bit is the graph:

    It shows in the cases of two of the three parties (LD and Tory) you’d have been better off taking the polls at the time the predictions were started than you would taking their forecast. Only in the case of Labour could they claim that to date their forecast has been more accurate than polls at the time.

    Con Forecast- Oct 2013 38% Forecast now 33%
    Con Polling- Oct 2013 32% Polling now 31%

    LD Forecast- Oct 2013 16% Forecast now 11.5%
    LD Polling- Oct 2013 10% Polling Now 9%

    Lab Forecast- Oct 2013 32% Forecast Now 31%
    Lab Polling- Oct 2013 38% Polling Now 33%

    We won’t know until election day how those original forecasts pan out but for now you can only say that they have failed miserably.

    To be fair I don’t think they ever claimed this was guaranteed to be accurate- just an academic exercise to test out an idea. Indeed their ranges at the time were so wide that I guess the original predictions still fall within those margins- just not anywhere the middle of those ranges.

  2. @Amber

    “What % of No voters are intending to vote SNP for Westminster?”

    No idea. This article says 14% of SNP voters (does not specify whether alignment, intention or past vote) voted No. With the SNP at 31.5% of VI at that time, say 4.4% of the total votes.

    4.4% of the 55% comes out at 8% of the 55% (4.4 % v 50.6%)

    Of course, none of that allows for those who have no political inclination towards the SNP, but might do to ensure the vow is carried through. To presume that all of those who voted No are of a particular mindset or opinion, when you haven’t canvassed any of them (and Con+Lab+Lib+UKIP is closer to 53% anyway).

    Perhaps the “at least 55%” statement is best left for the referendum. At most, it’s now 53%, and is no longer united, with each unionist party looking for its own May 2015 advantage.

    The 45% as it was called is probably closer to 42.5%, but is united under a single party at present. With 3.5% for the Greens, that would be 46%, but that’s presuming too much of those inclined to vote Green.

  3. “The SNP is clearly dominant at the moment, but that will pass. Which party do you see as becoming the main rival to the SNP – and why?”

    IIRC the Quebec bloc won something like 50 out of 70 at their peak.

    Given that voting for SNP but against independence seems logically like it would be the best way to get the best deal for Scotland inside a UK context I’d have thought a Quebec outcome is the most likely.

    How long it will last who knows but I’d have thought at least one election.

    I can’t see the SNP surviving supporting a minority Con govt. though so I can’t see this helping Con. If anything a large Bloc Jockois would more likely force a minority Lab govt. to tack left.

  4. Amber

    @ Bramley has already answered you question by posting s summary of Stephen Fisher’s most recent forecast.

    Each forecast run is based on (a) the current polling averages (taken from AW’s site, I believe) and (b) the amount of swingback still expected to occur between the date of the run and the date of the GE.

    If you look back over the sequence of forecasts since the launch of the model you will see that there has been a fairly steady decline in Tory-favouring measures (such as the probability that the Tories will have an overall majority, or most seats etc). This is tantamount to an acknowledgement that – to date – swingback has not been meeting expectations, and Stephen Fisher’s own summary of the first year of projections more or less said as much. As far as I am aware this is more than @ Robin Hood has done.

    It is anyone’s guess whether swingback will feature in the final few months to the election, and if it does whether it will replicate historical patterns. Swingback theorists have failed to provide an accurate account of the first phase of the putative process and so are in no better position than others to act as soothsayers about events next May.

  5. @MrJones

    I don’t think it’s an automatic given that Bloc Jockois politics will force Labour to move left. They’ll be entirely in support of Scotland favourable popularist policies, but England can fend for it’s self. The SNP don’t actually get anything for Scotland from, for instance, cutting or abolishing English tuition fees. The “English votes on english matters” and deepened devolution would give the SNP enough cover to abstain from votes that a ‘left wing party’ would be expected to support.

    I’d expect the SNP coalition/supply-and-demand requirements will be “Not be the Conservative Party, and provide Scotland with substantial continuing grants. Also, continue to let us blame you for everything.”

  6. @Unicorn

    It’s not entirely wrong to assume that vote-shares will tighten up getting closer to the election, revert-to-mean is a well known phenomena… But where “Swing-back” models fail is that they look at a small dataset that shows “on average” Government recovers vote share, but don’t consider what the moderating and accelerating effects on such a swing-back are.

    If you widen the scope to look at the purely mathematical behaviour of vote-shares in other countries, you can start to see what these moderating and accelerating effects are. First, is the mood of the country, which can be best gauged by looking at economic confidence. The more confident people are in their personal finances, the more an incumbent government can expect to recover from it’s mid-term low.

    The second, is that inertia builds up around a vote share. Dramatic collapses in a vote share are rare, and tend to be short-term blips that are recovered from, mostly a vote share follows a path of momentum. This can either re-enforce the swing-back if momentum is static or in the incumbents favour, but it also means that swing-back can be diminished or entirely erased with anti-encrument momentum.

    I note that while Fisher recognises his model appears to be failing, he assigns the reasons for that to purely political artefacts around the rise of UKIP. But UKIP is just a side effect, those dissatisfied voters would still be dissatisfied but more likely to protest by not voting if UKIP didn’t exist.

    If you consider that economic confidence is still pessimistic, and that there’s a huge weight of anti-incumbent momentum, the revert-to-mean was going to be hugely moderated.

  7. @Unicorn
    ‘As it happens we are now just a few days short of the half-way point between the date on which he made his projections and the date of the GE. By a nice coincidence it is still possible to write on this very day that:

    “The latest poll of polls shown on the front page gives the Conservatives a lead of -3%.”

    So, if we go by AW’s poll-of-polls, we see zero swingback over the first half of Robin Hood’s projection period.’

    I think you have miscalculated here – so a recount is in order! Robin Hood’s comment of July 7th was more than five and a half months ago. May 7th 2015, on the other hand , is now less than four and a half months away.Thus, we have already passed the midpoint of his forecast!

  8. @ Statgeek

    To presume that all of those who voted No are of a particular mindset or opinion, when you haven’t canvassed any of them (and Con+Lab+Lib+UKIP is closer to 53% anyway).

    I’ve canvassed loads of them; & I’ve also seen post-referendum polling that puts ‘No’ voters who’d vote SNP as close to zero as makes no difference.

  9. @ Scotslass

    I was surprise that she/he was not aware of the YouGov finding since Amber intervenes regularly on Scottish politics with fairly entrenched views.

    Do you always make ad hominem comments about people who don’t share your views? Or was this just a symptom of your “visceral fury” towards LiS ‘No’ supporters?

  10. @Graham

    “I think you have miscalculated here..”

    Oops! Too much solstice mead, methinks..

  11. the fisher model was bunk in its predictive power. he said there would be a tory majority at GE 2015 at the end of 2013- many, on this site, recognised that for the twaddle it was…funnily enough his model is now in tune with what many of us were saying back then….hung parliament, labour most seats.

  12. I do find the ‘search to quantify swing back’ as being akin to trying to make gold out of lead.

    Given the predictions of the 2015 GE, based on historical swings, have radically changed as previously assumed changes in VI simply have not occurred.

    I know that new figures feed into the various models, but how many more times must predictions change before the we treat them as a psephological Trigger’s broom (the same one for 20 years despite having 17 new heads and 14 new handles).

    The complexities of swing have escalated massively with the slow decline of the usual suspects, the sharp rise of UKIP and the SNP in recent times, and the creeping up of the Greens. Each new element adds multiple interactions to the swing model, and once you get to 5 or 6 elements you might as well pack up and go home trying to model it.

  13. @Jayblanc

    Point taken on the risks associated with concentrating on basic VI changes without looking at the factors driving the figures one way or the other.

    This charge could be levelled against the main numerical election projection models. Do you know whether anyone formally incorporates data on factors like economic confidence in modelling projections?

    Judging by his post mortem of the 2010 election “Minority Verdict” I’d be pretty confident that Lord Ashcroft is busy tracking influences of this kind. But it would be interesting to know whether anyone is doing quantified projections on the basis of such data.

  14. @ Unicorn

    Thank you :-)

  15. @Unicorn

    I do.

    Expect a big change in how I represent the data from my model in the new year tho. I’ve adjusted the coalition logic to account for a possible “Tartan Coalition” of SNP+LibDem+Labour which is now a very clear possibility.

  16. Note, that the economic factor is comparatively mild at the moment, with a relatively ‘high’ CCI at -2.

  17. @Scotslass

    Yes, that was fairly obvious, there was no need to shout.

    There is a new poll out (at least I think it is) on Scotland in the Gruaniad.

    The numbers on the Smith Commission are somewhat different:

    30% Not far enough
    26% About right
    13% Too far

    It’s a poll of 1004 by ICM so a moe of about 3%.

  18. Guardian/ICM online (Scotland Westminster):

    SNP 43
    LAB 26
    CON 13
    UKIP 7
    LIB 6
    GRN 4

    Size 1,004
    Usual seasonal caveats…


  19. The Sheep

    Thanks for the info about the new ICM/Guardian Scottish poll.

    Looks like Curtice has had access to the micro–data, so that he has been able to analyse by type of seat and “Curtice reveals Labour’s decline is sharpest in those supposedly heartland seats where it previously trounced the SNP by more than 25 points.

    Whereas Labour’s Scotland-wide vote drops by 16 points, it falls by 22 points in these constituencies while the SNP surges by 26. That combination is sufficient to wipe out majorities that were always assumed to be impregnable, and Scottish Labour’s Westminster caucus is left shrivelling to just three MPs.

  20. @Unicorn
    With regard to your post of today 26 Dec , 8.47 a.m. you ask about Robin Hood’s swing back prediction.

    You ask if he has modified his prediction since the July prediction which you mention. Robin Hood has come back on this site with a revised prediction quite recently that I remember.

    Success! I have found it! Robin Hood also posts on the Labour Target seats site and that’s where you can easily find and read his new prediction.

    Click on “election guide” at the top left of this page. Then click on Labour Target seats. Then you’ll find Robin Hood’s post dated 7 December 5.54 p.m.

    Robin Hood predicts a 3 per cent Conservative vote lead giving a hung parliament with a possible Labour majority of seats now. However, he is not sure.

1 2 3