Monday’s polls

We have three polls today, the daily YouGov poll, the weekly Ashcroft poll and the twice-weekly Populus poll. Topline figures are:

Populus – CON 32%, LAB 33%, LDEM 9%, UKIP 15%, GRN 6% (tabs)
Ashcroft – CON 34%, LAB 30%, LDEM 5%, UKIP 15%, GRN 8% (tabs)
YouGov – CON 35%, LAB 31%, LD 8%, UKIP 14%, GRN 6%

Lord Ashcroft’s poll repeats the Tory lead it showed a week ago, YouGov produce a four point Tory lead, their largest since January 2012. Populus continue to show a one point Labour lead.

All the usual caveats apply, we need to be careful not to overreact to polls that could just be a couple of outliers in the same direction – that said, in YouGov’s daily poll we’ve reached the point that Conservative leads are a little more common than Labour ones. Of the last ten YouGov polls there have been four Tory leads and two Labour ones. I don’t think we can confidently say the Tories are ahead… but I’m certainly no longer confident in saying that the underlying average is a small Labour lead either. I think we can fairly say that the Conservatives don’t seem to have suffered any short term damage from the debate debate last week.


340 Responses to “Monday’s polls”

1 2 3 4 5 7
  1. The Tories have moved ahead in the March YG averages, but that lead only currently stands at Con 34 Lab 33. It may rise; it may fall, but at present we are still taking about a neck and neck race.

  2. James

    The BBC are also trying to contrast their March 2015 Scottish poll with a GB Feb 2014 poll.

    While there probably isn’t much of a difference north & south of the border on the issue, this report is well down to their usual standard of sloppy journalism.

  3. Colin

    YG now put the latest poll at the top right of the Politics page labelled “from the archive” (which, of course, it isn’t in!)

  4. @07052015

    ‘A freudian like the uniquely gifted Leo Abse mp would say that EM standing against DM is seen as the ultimate betrayal and a profoundly anti authority move. ‘

    An insightful analysis and IMO correctly puts the ‘personal into politics’.

    It is clear that the adeptness of political spin and advertising, is dependent on trading on advantageous cultural/psychological schema and preferably in the absence of contrary information. The mass media colludes (intentionally or not) by ignoring all but the ‘battleground du jour’.

    In your example, the focus is on the ‘betrayal’ and totally ignores the very substantial political differences between EM and his brother. Those who say that the ‘right’ brother would have led a more successful campaign for Labour because he ‘looked more like a PM’, ignore the damage the extraordinary rendition and support for Iraq might have had on the Labour IV. However, it is true that parts of the press might have found DM a less threatening prospect.

  5. UNICORN

    Thanks for that.
    Interesting.

  6. OLDNAT

    Indeed !

    Still having their breakfast I suppose :-)

  7. @Colin

    You’re welcome!

    For what it’s worth I suspect we’ll be able to point to a small but reliable Tory VI lead before the Election. But unlike some I’ll be sceptically picking away at the evidence long after it has become common ground amongst the UKPRariat.

  8. UNICORN

    I would expect nothing less from you !

    If this is your profession , then picking away must be difficult to avoid :-)

  9. @Colin

    It is notable that Anthony Wells produces his assessments by eyeballing the data and attempting an impartial overall assessment rather than attempting the over the top statistical assessments produced by others on this site.

    Generally to see significant statistical differences you would need to see a a marked change in VI and/or a change over a extended period of time.

    That’s not why most of us are here on a daily basis. We are looking to identify early trends.

    By the time the stats show any significant changes (eg p <0.05) this would likely be manifestly obvious to the non statisticians on this site given that the data is relatively simple..

    If the statisticians here were really running their assessments for purely intellectual or academic reasons they wouldn't be running them on a daily or weekly basis.

    In particular "there is no statistical evidence…" is misrepresented when AW might say "there is possible evidence of a movement but too early to say anything definitive" (to paraphrase what he has said above).

  10. Note: I posted my previous comment before unicorn’s post above.

  11. For the LibDems to get a decent number of seats we need to make increasingly heroic assumptions.

    I’m assuming labour win where they came second last time, the Nats will win in Scotland.

    We have to assume that the anyone but the Tory vote coalesces around them despite being in government with…the Tories

    I assume they will get a few per cent in each seat, 3%? which means in the seats they currently hold they need near perfect efficiency of votes in the right place.

    The increasing Tory support means they need more votes to win against the Tories.

    Beginning to look to me like they will get less than 20 seats.

  12. JP

    I agree with every word.

    The waffle free zone that is a Wellsian synopsis is just what I am looking for.

    But I can obviously understand that for others , satisfaction is only to be found in lots & lots of numbers & statistical convention.

    Come May, it will all be irrelevant.

  13. “EM himself puts it down to being Jewish”

    But most people who despise Ed think his brother who, one presumes, is just as Jewish as Ed, would have been much better.

  14. Unicorn

    I didn’t realise the quoted confidence intervals were at 90% rather than 95%.

  15. JP COLIN

    You are probably correct that AW’s experienced “eyeballing” is all we need to understand the movement in polls but I like to read the statistical analysis by UNICORN and others and hope it reveals a more. This is one of UKPR unique qualities.

    (Much better than the partisan twaddle that I occasionally cannot resist responding too)

  16. Morning everyone, just a quick update. I had been holding off providing a Welsh projection due to the variability of the Plaid Cymru vote in crossbreaks. Lumping them in with the BNP didn’t seem to be a good idea ;) So, happily the “proper” Welsh poll has come along. On these numbers, PC hold their three seats, but don’t gain Ceredigion (LD win 33-27 on my figures). The Conservatives would gain Brecon but lose Cardiff North, while Cardiff Central would go to Labour from the Lib Dems. So, overall in Wales, not much change: Labour +2, Lib Dems -2, Conservatives N/C, PC N/C.

    So, putting all of that together, and including the Scotland figures from yesterday, my LWVG NowCast ((c)UKPR contributors 2015) is

    Con 282 (-24)
    Lab 290 (+32)
    LD 11 (-46)
    UKIP 2 (+2)
    SNP 43 (+37)
    GP 0 (-1)
    PC 3 (N/C)
    Speaker 1 (N/C)
    NI Parties 18*

    However, these numbers could change a bit once I put in the latest xbreaks in the model.

    *In NI, there will be some interesting and ferocious battles, particularly in Belfast East (DUP v Alliance), Belfast South (SDLP v everyone) and Fermanagh/South Tyrone (SF v possible united Unionist candidate)

  17. Recent Conservative leads have been picked up in the Guardian 2015 prediction site: Cons +7, Lab -4.

  18. Hawthorn

    Typically 90% confidence intervals are two tailed, whereas 95% ones are one tailed.

    It depends whether you want to test y>x or x<y<z. It shouldn't change the intervals, just whether you want to include one or other of the tails.

    It's possible (but unusual) to do a 1 tail 90% confidence interval test but that's often a sign that a 95% interval test for something you wanted to say just failed.

  19. “The waffle free zone that is a Wellsian synopsis is just what I am looking for.”

    The gobbledegook about “Euclidean spaces” is unnecessary.

  20. @Poll Troll

    may2015 believe that a crossover has occurred:

    This summary actually goes beyond what they eventually say. Although they flirt with the idea that the Tories might be in the lead, their final conclusion at the very end of the piece is:

    it would seem Labour have finally lost their long-held poll lead.

    Obviously “crossover” requires a demonstration that the Tories have moved beyond MoE into the lead. As yet, no one has shown this to be true.

    Actually the May2015 statement is almost meaningless as it stands. To make it a testable statement they would have to specify a time interval over which their claim is held to be true.

    I could take the polls since Jan 1 and now and – using this time period – I could demonstrate that Labour are ‘still’ reliably ahead of the Tories. The same is true of polls in February, in a comparison I actually posted her about an hour ago.

    So the question is: What period are they talking about? Are they cherry-picking if they draw the boundaries in such a way as to exclude a couple of awkward looking Labour-favouring results?

    If they specified (say) the last week’s worth of polls, then normal polling variability might make it difficult to prove them decisively wrong. But equally they would be hard put to disprove the competing hypothesis that Labour are still marginally ahead and that polls suggesting otherwise are just the routine manifestation of MoE.

    So, in my opinion the May2015 piece doesn’t add much of substance to the debate.

  21. @Hawthorn


    I didn’t realise the quoted confidence intervals were at 90% rather than 95%.

    5% falls beyond the boundaries at the top end of the distribution and the remaining 5% at the bottom.

  22. Oops! Missed @Alan’s post saying the same thin in more detail.

  23. @LouisWVG

    I’d still be interested to learn more about the Vi-redistributing algorithms you adopt for the different parties, and especially whether they are all the same.

  24. Well I for one quite liked Catmanjeff’s analysis, in particular his Cusum graphs.

    Far more precise than what we normally see, it helps separate the MOE noise from real movement.

    I think there is a momentum effect to movements – those graphs show the direction of momentum for the Conservatives at least is presently upwards, so we can expect increasing Conservative scores until that changes direction again.

    It would be good to compare those with Cusum graphs from other parties to see if we can see for example UKIP or Labour change direction at the same time as we see Tory direction changes.

  25. @Polltroll

    “may2015 believe that a crossover has occurred:”

    If we are talking YG only, it has, and my charts suggest this…if this most recent poll is part of an underlying trend (well that’s an obvious thing to say!).

    http://www.statgeek.co.uk/polling/periodic-averages/six-poll-averages/

    To explain further. Prior to today’s poll, the penultimate data points for Con and Lab had the latter higher than the former. Today’s poll swapped things around so that Con is higher, for what appears to be two averaged data points in a row.

    h ttp://www.statgeek.co.uk/wp-content/uploads/2015/03/compare.png

    So if it is crossover, let’s use the 3/4/5 in a row. I think it’s closer than that at present. For what it’s worth, if you check out the fifteen-poll averages, they are sold on the idea. It’s a question of what group of polls and in what way you want to look at them.

    My 30-poll unweighted MAD system shows: Con 33.0, Lab 33.3
    My 25-poll weighted MAD system shows: Con 33.1, Lab 33.7

    However, the regional total for the latter system (the total of the regional calcs, rather than the calc of the top line data) shows: Con 33.3, Lab 33.9

    So there you have it. If we take a shorter and shorter sample of polls, thing improve for the Conservatives, right up until the 1-poll average, which has the Conservatives leading by a whopping 4% points. :))

  26. @Richard

    I agree that @CMJ’s CUSUM analyses can be very sensitive: indeed, more so than other obvious methods.

    What I don’t get is the rational for the subdivision into the three groups. It would be helpful to have thwt explained.

  27. Edit: The penultimate data point was Con on top in both. The last data point shifted with today’s poll.

  28. Back in Jan 1983 Bob Hawke took over as leader even closer to polling day than we are now. Perhaps Labour needs to look at that precedent!

  29. I am amazed that three people on this site constantly regale us with their
    personal opinions of Ed Milliband ,more useless than Foot etc.I would love to
    let rip with my opinion of Cameron but this is not the place.

  30. @Unicorn (10.24)

    I can’t possibly tell you that – it’s top top secret!!! :)

    Ok, I’ll try to explain it as best I can. First of all, I’ll say that all parties are treated equivalently, the Lib Dems are not a special case (well, they might be, but that’s a discussion for another forum…).

    I’ll define a couple of values: Let V_2010 be the average vote share that a party got at the last election, and V_2015 is the vote share that the polls give. Also, let X_2010 be the percentage vote that the party got last time out in the particular seat of interest. Now, I define the function X_2015 = f (X_2010), such that f(x) = alpha * x^gamma for x1) below V_2010, and then more or less a straight line above. For UKIP, the curve is convex (gamma<1) below V_2010.

    There are two other technical issues. Since the sum of these vote shares does not identically equal 1 in each seat (in principle, this could be another constraint, but I haven't implemented it yet), I have to normalise the values (this doesn't change any prediction, although it does indirectly provide a measure of the softness of the vote in a particular seat). The other issue pertains to the GP and UKIP mainly – in seats where they didn't have a candidate last time. To avoid giving a big fat zero everywhere, I've set X_2010 equal to 50% of V_2010. It generally doesn't make a huge difference, but it seems a bit unrealistic to have lots of South East England seats with 0% for UKIP :) If anyone has a better idea of how do to this, it would be welcome!

    I hope that the explanation is clear(ish), and didn't bore the rest of you to tears (or alternatively drives you away from UKPR to listen to Philip Hammond give out about Russia :) ).

    I can't possibly tell you that – it's top top secret!!! :)

    Ok, I'll try to explain it wi

  31. For what it’s worth, my 25-poll, weighted England and Wales data:

    Con 34.7
    Lab 34.6
    Lib 7.4
    UKIP 15.7
    Green 6.1
    Others 1.5

  32. @Ann in Wales

    Hear hear :)

    I just love the way, according to much of the media etc, a VI of 34% is considered simultaneously Tory success and Labour failure.

    There’s 8 weeks to go. A lot can change. Ed needs to pick a fight. He’s got a good record picking something that will resonate.

  33. There are many comments about EM’s approval ratings compared to DC:
    I thought that this made interesting reading on the connection (or not) between approval and election outcomes.
    http://blogs.lse.ac.uk/europpblog/2015/02/21/do-party-leader-approval-ratings-predict-uk-general-election-results/

  34. Ann in Wales

    “I am amazed that three people on this site constantly regale us with their
    personal opinions of Ed Milliband ,more useless than Foot etc.I would love to
    let rip with my opinion of Cameron but this is not the place.”

    There is a difference. Miliband’s ratings are a material factor in the campaign, and his unfitness for office is one of two key tory messages. Miliband is an entirely appropriate subject of discussion on a website dedicated to polling.

  35. Sorry about the random last lines above – I must be more careful with copy and paste!!!

  36. GRAHAM
    “Back in Jan 1983 Bob Hawke took over as leader even closer to polling day than we are now. Perhaps Labour needs to look at that precedent”
    _____

    It looks like you’re chucking in the towel but who can blame you?

    The big question though! Who do you replace EM with? All the best bits are jumping off the ship at the election, Brown, Haines, Darling and my personal dream ticket Dennis Skinner would probably rule himself out in any leadership contest.

    So I’m really trying to comprehend who could replace EM and the Labour party’s fortunes around before the election!

    I don’t do doom & gloom but if I did then it would look something like the current Labour VI.

    #headsup

  37. @WB

    Interesting read. When I was looking at the post mortem of the Scotland 2011 elections a few days ago, many of the studies pointed to the leadership ratings as the reason for the SNP victory there, but one study that I linked to said that did not really have the impact that other studies said it did – they said

    http://eprints.lse.ac.uk/37993/

    “However, these results belie any suggestion that it was the personal popularity of Alex Salmond that ushered his party into power – because the SNP was in fact slightly more popular than its leader. This is not to say that Salmond’s leadership was unimportant. There may be an echo of Tony Blair and New Labour here. Survey data from the 1997 election suggest that Labour’s success owed surprisingly little to Blair’s personal ratings. Yet the key drivers of that triumph – that voters saw Labour as moderate, united, and competent (especially on the economy) – were obviously connected to Blair’s revamp of the party. The same might be true of the Salmond-led SNP, which has unified around a pragmatic approach to independence. Insofar as the 2011 victory was created by Alex Salmond, it is because of what he had done to his party rather than to his personal appeal per se. ”

    Certainly, the Labour VI is still quite high despite the apparent unpopularity of Miliband, and from what I have seen the Labour PARTY still seems to hold its edge over the conservative PARTY, although I don’t remember recent polling on that.

    I’d like to see some studies that compare party popularilty vs leader popularity vs election outcome.

  38. @LOUISWALSHVOTESGREEN

    It’s a refreshing change to find a Green supporter on here who is not constantly taking a very upbeat prospect of the Green Party’s prospects.

    I disagree with your view that they’ll get zero seats – I think they’ll get one. They ought to hold Pavillion unless the student vote collapses with the advent of IER (possible), Lucas also fails to pick up an incumbency boost (unlikely) and the antics of the local council do enough damage to offset the national polling trends.

  39. WB

    That is the issue, and it inevitably involves voter perceptions of Party Leaders.

  40. “Certainly, the Labour VI is still quite high despite the apparent unpopularity of Miliband, and from what I have seen the Labour PARTY still seems to hold its edge over the conservative PARTY, although I don’t remember recent polling on that.”

    There’s been endless stuff on this. ..the labour party is more popular than the conservatives, but Miliband is a lot less popular than cameron. In many ways, the election hinges on whether miliband weighs down the labour brand sufficiently to hand the keys of no. 10 to Dave and his chums.

    It’s clear to me that the government is not particularly popular…if we consider that the two coalition parties got 60% of the vote last time, and are unlikely to garner 45% this time, you can see how much both parties together have gone down, with the lib dems bearing the brunt of the decline.

    Labour’s inability to harness the anti-coalition parties sentiment has been remarkable, to the extent that the bookies now believe the tories will finish up with more seats than the reds.

  41. @Peter Crawford

    It is one thing to say DM would have made a better leader, and it is quite another to actually support him if he was the leader. Sadly there is a quite significant anti-semitic undercurrent in the UK, and it would probably have come into play for either of the Millibands.

  42. @LouisWVG

    Okay – so do you fit the parameters separately for each party? And if so, using what initial database? I’m just trying to get a feel for how your model would shift VI around.

    ‘More or less a straight line’ ?? With what slope?

    You general approach seems to bear some resemblances to the Strong Transition algorithm used by Electoral Calculus. That, too, divides the VI Into different bands, each of which behave differently.

    It seems sensible to do some patching to handle no-2010-candidate constituencies. But one consequence. – I would assume – is that this puts Clacton firmly back in Tory hands.

  43. ANN IN WALES
    Agree

  44. Regarding the Welsh poll, the full tabs are here:

    http://d25d2506sfb94s.cloudfront.net/cumulus_uploads/document/rh8h1erygs/ITVWalesResults_150309_VI_w.pdf

    It’s very surprising that Roger Scully’s commentary focuses on the Assembly polling when we’re only 8 weeks away from the GE.

    GE polling Lab 39%(+3), Con 25%(-1), LD 5%(-15), Plaid 10%(-1), UKIP 14%.

    Below GB average swing to Labour, but that’s been the case in Wales for some time. Both Labour and Tory up 2% from the Jan equivalent poll, at the expense of UKIP and the LDs. Plaid polling 20% for the Assembly but only 10% for the GE.

  45. Allan Christie

    That and a complete lack of recognition outside those names.

    If Labour were to replace Ed with say Chukka Umunna, the overwhelming response would be “Who?”

  46. THESHEEP

    ” Sadly there is a quite significant anti-semitic undercurrent in the UK, and it would probably have come into play for either of the Millibands”
    ______

    Absolute rubbish. Have you got anything to substantiate this? and I’m not talking about some dire piece in The Sun written a few months back by a crack head.

    Or is it just me who has my back to the wall?

  47. Unicorn/Alan

    Thanks for the clarification.

  48. Hi,
    I was scrolling through some past debates and would like to add an observation, in relation to snp and labour, and the referendum.
    Following the Scottish ‘indy’ referendum, it seems like the losing independence side have kept their votes, or support, and the winning ‘alliance’ have been reduced to sharing the remainder of the potential votes, thus we see the unexpected collapse of Labour support. This was a bitter campaign, polarising and dividing opinion, between families, work places etc..
    If we have a ‘in/out’ referendum on Europe after the coming general election, and let’s say the result is similar to Scotland (ie 55% vote to stay in Europe) can we not extrapolate a similar outcome?
    After a bitter debate, (and what discussions about immigration are not bitter), the winning stay ‘in’ vote will be split between libdem, labour, green and some Tories, while the remainder ‘out’ vote will handed to UKIP, and some Tories. The Tories will have to decide where they stand, but they could have defections either left or right by unhappy members, weakening the party, and leaving just UKIP representing some 45% of voters.

  49. AC
    I have not voted Labour at a General Election since 1992 and have no intention of doing so on May 7th.
    I still expect on the basis of past election campaigns dating back to 1959 that Tory support is likely to peak this month and fall off somewhat during the official campaign. What the peak will be and the degree of subsequent fallback remain to be seen.

1 2 3 4 5 7