After the wild moves in Ipsos MORI’s last two polls, which showed the Tory lead tightening to 6 points and then spiking back to 17, we’ve almost the mirror image from ComRes. Their last poll showed a 17 point, and tomorrow they have a new poll in the Independent that shows a sharp reduction. The topline figures, with changes from their poll just over a week ago, are CON 38%(-3), LAB 29%(+5), LDEM 19%(-2).

Whereas the big movements with MORI were down to the lack of political weighting allowing a sample with a very perculiar amount of 2005 Labour voters, with ComRes my guess it is the rather more mundane explanation of a rogue poll – in hindsight the 17 point Tory lead in their last poll looks wholly anomolous, the only other pollster showing such a low Labour share of the vote was Angus Reid, who seem to consistently show a lower level of Labour support for methodological reasons. If we put ComRes’s previous poll to one side and look at the one prior to that, the shifts are far smaller, with the Conservatives up, Labour up 2 and the Lib Dems down 1 – no significant movement in itself, but chiming with the recent slight strengthening for Labour.

So, as we head to the end of the year (YouGov/Telegraph is still outstanding in theory, but I’m not sure when it will arrive), we still have quite a broad range of polling figures, with leads between 9 and 17 points – from a hung Parliament to a Tory landslide. The Conservatives are in the range 38%-43%, but of the 12 polls in December 9 have put them at 40% or 41% – that’s a noticable difference from November when 7 out of 10 polls had them below 40%. Lib Dem support ranges between 16% and 21%, but mostly between 18% and 20%. The real variation is in the level of support pollsters are finding in Labour’s support, from 23% to 31%. However, the lower figures there are either AngusReid or that single ComRes that appears to have been a rogue, and other figures are in the tighter range of 26% to 31%. That leaves us with an average lead of around 11 points or so – on a UNS right on the cusp of a hung Parliament, though in practice it would probably result in a Conservative majority.

(As an aside, if you are following the polls in the glorious rumourmill of Twitter then the 3 point Tory lead in MORI on Sunday has been followed by CON 40%, LAB 31% in this poll. Boy, are polls in the general election are going to be fun there!)

164 Responses to “New ComRes poll shows 38/29/19”

1 2 3 4
  1. It doesn’t take many chinese tweets for ‘some new account called ‘africansontoast’ to become ‘some people’ to become ‘reliable sources’.

    I’m hearing yougov is 39 30 19. Maybe people feel more charitable and labourish around xmas, maybe I’m just reading my stereo. Maybe I shouldn’t have bought something from a compant that couldn’t get it right the first 393018 times.

  2. do the newspapers have any sway over the polling companies methodology?

  3. On the last part: I confess that was my fault, I misread the blogpost by the Independent as saying the Tory figure was down one, so did the maths from there, it actually stated the lead was down 1 point since last month! : /

  4. This poll is better for Labour than some others recently in terms of the Tory lead but their share on 29% is not particularly good news for them in my view. I think most Labour supporters would have preferred a 41-32 poll if there was to be a 9% Tory lead. As Bob Worcester says the share is the most important thing.

  5. @Andy Stidwill

    Cons low in relation to Lib Dem’s helps reduce the Con gains in the south, so if Labour is in defensive mode, I think they would prefer 38/29/19. Of course it all depends on where the movements occur. As Statto says, swing is the most important thing.

  6. One thing that stood out for me amongst Labour 2005 voters, 10% of them are either undecided or Don’t Know as opposed to 4% for the Tories. Labour could yet get them back.LD has in fact 12% Don’t Know but that is not surprising, eventually they will vote LD.

    Page 9 of Com Res summary

  7. Andy Stidwill: This poll is better for Labour than some others recently in terms of the Tory lead but their share on 29% is not particularly good news for them in my view. I think most Labour supporters would have preferred a 41-32 poll if there was to be a 9% Tory lead. As Bob Worcester says the share is the most important thing.

    Did Bob W really say that as I think it is counter intuitive.

    Let’s say 41-32-17 vs. 38-29-23

    The Labour Tory swings are the same. But LD’s surely will do better with the latter.

    All based on UNS.

    I think, for any given difference in the Con-Lab spread, Labour would prefer a lower Tory figure [ which, of course also means a lower Labour figure ] but a high turnout.

    Labour has far less to lose from a high LD vote than does the Tories. Unless Clegg screws it up, tonight’s debate announcement gives the LD’s one or two points extra.

  8. And… Back to where my trend model was predicting we’d go…

  9. I did some experimentation with estimating seat allocations with improved Lib Dem figures… Basically, no, it’s not bad news for Labour if Lib Dem share goes up, so long as it doesn’t eat into Labour’s share. If they, as I would expect, eat mainly into ‘others’, floating voters, and just nibble at Labour and the Conservatives; then the result is a two or three point increase in the lead that the Conservatives need to get a majority.

  10. The polls seem as reliable as the uk in the snow and economic forcasts remember how everyone predicted growth when we ended up still in recession ?

  11. WMA 40:28:19 as you say this is a mirror image of their last poll, deviation in CLead 3% which is well within MoE. Nothing going on here at all.

    However the wider picture is that the downtrend in the CLead is now essentially non-existent. Over 140 days it has an R2 of .46, but over 2 months the R2 is only .29 and over the last month there is an insignificant uptrend R2=.21.

  12. I think the polls show that the old way of polling does not work anymore

  13. @NBeale

    Conversely, the trend generated from my set shows a significant downward trend at R2=.81

    I suspect it’s because I remove Angus Reid, and use an unweighed MMA, and you don’t.

  14. The cons won by a majority of 21 or so in 1992 and only had 7% lead and now boundarys have changed to favour the tories so surely they should get about a 50 seat majority if they had a 9% lead

  15. Coincidently, an explanation of what an R2 figure is for other commentators. In layman terms, it tells you how well the trend fits the data, and therefore estimates how likely it is to fit future data. The number ranges from 0 to 1. So R2=.01 is a very poor trend from very random data, and R2=.99 is a very clear trend from predictable data.

    R2=.21 is not a very clear trend, but R2=.81 is.

  16. @Paul B

    No, because political populations have moved about quite a bit since 1992. It’s not just the redistricting that’s changed the political landscape, but changes in demographic make up and voting trends of key constituencies.

  17. It’s not impossible that the Tories could win an overall majority with only a 7-8% lead if the swing in the marginals is much larger than in safe seats. In 1992 the swing to Labour was double in the marginals compared to safe seats (4% v 2%) which is why they almost wiped out Major’s majority despite trailing in the popular vote by 7.6%.

  18. I’d also like to mention something else about movement in the polls…

    You can expect up to 4 to 5% movement between polls just from Sample Error.

    So while it’s very very tempting to say that X Poll shows the Result of Y event, if the movement is less than 5%, then you’re probably seeing more effect from Sample Error than you are what ever event. This is why I work from a moving average of polling, as that diminishes sampling error by allowing the error of various polls to cancel each other out. It also allows me to generate a more predictive trend, as my data is less noisy.

    Conversely, my data will only react to an event a few polls after it, so might be slower to adjust that trend to reflect a significant change in the situation. But the last time we had significant poll movement outside of sample error within a short time frame was the expenses scandal. Otherwise, the story here has been steady movement along a trend.

  19. @Andy Stidwill

    But the only data we have on the marginals has been contradictory, and suspect.

    A low sample rate regional poll suggested that the Conservatives might out perform national swing in the northern marginals. But local by-elections saw Labour doing surprisingly well in marginals, and particularly well in London. Neither of these can really be said to be sure predictors of what will happen in the marginals.

  20. So labour voters move but tory voters dont what a load of rubbish your telling Me that you know where people move ? The only way you could be right is if the government has planted voters In places

  21. I agree that the trend is what matters, and the Tories have been about 10-13% ahead for a long time now. Labour only have about 15-20 weeks to change that which will be a difficult task.

  22. I think many lib dems will vote conservative to get rid of labour if Im right it could produce large swings the lib dems vote is down who is taking there vote?

  23. That could be right Paul, but it could well be that such voters are already shown in the polls as intending to vote Conservative anyway.

  24. The LD vote appears to be down most in Scotland but it’s not the Tories who are benefitting from it. Some surveys are showing the Scottish LD vote has been halved from 22% in 2005 to 11% now. Mind you, they must be doing okay in the rest of the country since they’re only slightly down in the national polls compared to their GB vote in 2005.

  25. Andy Stidwill

    I have to admit I’m a natural pessimist!

    While I think that much of the LD vote has gone SNP, I’m well aware that many will make up their minds in the polling booth.

    I’m conscious of the TNS Scottish polls

    Party, Apr 05 poll, 05 election, Oct 09 poll
    Lab, 45%, 39%, 39%
    LD, 14%, 23%, 12%
    SNP, 23%, 18%, 25%
    Con, 14%, 16%, 18%

  26. Kalman filter says Tory lead 10.8%
    Con majority 7.

    Btw, looks to me that the Tories are in a “descending triangle” pattern, which is considered bearish.

    Hung parliament on

  27. The Cs will need a bigger lead over L than in 1992 because the LDs hold a load of former C seats and there are more seats held by others. It’s not just the lead that counts but the share of the vote. John Major got 43% of the vote – if that is share repeated with the current splintering of party political support C would have a huge lead over labour.

    Comparisons with other elections are no longer fully valid because bit bit by bit the political landscape is changing as the two biggest parties become less dominant.

    The next election is likely to produce a C majority but it may well be the last majority government we see for many years.

  28. Why is this poll in particular being spun so much? The BBC picked it out in their first paper review of the morning on the Today programme, and Reuters UK site has the headline “Conservatives poll lead at 6-month low” (although it should be said that the article is not internally consistent, since it concludes with the sentence “recent polls have shown the party’s lead fluctuate between as much as 17 percent and as little as six percent”).

    Somebody, somewhere, seems to be putting a lot of top-spin on this poll which is bizarre given how volatile the overall picture has been recently.

  29. It’s because the press (and press agencies) like uncertainty of outcome, so a close poll is always going to be more widely reported than a divergent one

  30. “I think many lib dems will vote conservative to get rid of labour if Im right it could produce large swings the lib dems vote is down who is taking there vote”

    Lib dem voters that know what they’re doing will hope for the situation that the polls are suggesting, at which point lib dems will hold all the cards.

  31. I have a suggestion that I’m not sure people will make of. Wouldn’t it be better of sampling over 1,000 random people, to instead sample the same number but in chosen constituencies and to change these constituencies week by week? For example, 400 from Seat A, 400 from Seat B and 400 from Seat C? I say this because I see three major advantages: Firstly, It takes the currently undetactable ‘personal’ vote into account, which is a major factor in many seats and bucks national trends. Secondly, it would give us a more indepth understanding of those all-important marginals, in which the Tory lead is said to be greater, but of which we have no definitive proof. And, thirdly, it would be easier to spot a rogue set of results because of an identifiable cause i.e. A wing of a local hospital has been closed. From experience in local polling I can say that particularly in the North of England local issues take far greater precedence over national issues. There can be a psychological as well as an economic detachment from Westminster and the South, that polls do not adequately pick up on.

    The big problem we seem to be having with the polls at the moment is that there are great fluctuations occuring for no apparent reason- For example, a tightening in the polls after the PBR when many were expecting an increased Tory lead.

  32. Further to the impact of polls, we need to keep in mind who actually is influence.

    We can all accept that the man/woman on the Clapham omnibus does not scrutinise and follow each poll as slavishly as us, sad,nerdy people. For him/her, polls register significantly either in an election period or for the impact the polls have on the commentariat in the media.

    We are not in an election period quite yet.

    For the ordinary voter, the polling impact is from how it is playing in the media coverage of the government. Which brings us back to a common theme on these message boards of the ‘self-fulfilling’ poll boost.

    i.e. a couple of polls which aid the government=media coverage of closing polls= media impression of government resurgence= Clapham man getting a better impression of closing polls= Clapham man likely to tell the pollster that the Government is doing better= a couple of polls which aid the government.

    This is all well and good, but it continues to leave the Government at the mercy of events which can change the media trends in a matter of hours. The government gives the impression of being grateful for the breathing space without really knowing why either. Unless Labour’s operation takes a firmer hold of the news cycle, events (as they must) will blow the polls apart once again.

  33. “One thing that stood out for me amongst Labour 2005 voters, 10% of them are either undecided or Don’t Know as opposed to 4% for the Tories. Labour could yet get them back.LD has in fact 12% Don’t Know but that is not surprising, eventually they will vote LD.

    Page 9 of Com Res summary”

    An informative comment from Surbiton, thanks.

    This sheds light on Labour’s agreement to the TV debates. They’re likely hoping to persuade some Lab don’t knows; &/or that Nick Clegg’s inclusion will put Lab DKs into the LD camp, together with firming up LD’s own DKs.

    The polls following the TV debates will be fascinating.

  34. Not much change really with a Conservative average of 40% and Labour Average still below 30% this still points to a Conservative majority in May.

    As I’ve said before this election more than any other will be decided by the English Marginals!

    Con ~ 40-42
    Lab ~ 28-30
    LD ~ 18-20

    Conservative Majority – 20-50

  35. A lot of the commentary in the media seems to be around whether Cameron has ‘sealed the deal’. All the recent polls, despite the swings, suggest that he has: or at least with 40% of the electorate. He can’t get much higher and he doesn’t go much lower.

    The next election hinges less on the Tories and more on the extent to which Brown ‘heals the deal’ with the Labour electorate. The variations in the Labour polling seem consistent with an analysis that suggests that a proportion (perhaps 20%) of Blair-Labour voters really are swinging from Labour to don’t know depending upon current news agenda.

    The Tories need to do three things to this constituency: 1) keep them out of the polling booths by making the Tories less monstrous to them, 2) get them to vote Labour in the south (pulling votes from Lib Dems), and 3) get them to vote Lib Dem in the North (pulling votes from Labour). A delicate political strategy indeed.

  36. @ Cap’n Scooby

    The mainstream media are pretty firmly against Labour. That’s another reason why Labour have agreed to the TV debates, I’d think.

    Regarding events, here’s the 2 that I think will have a major effect on polling:
    1. Employment figures – Lab don’t knows will forgive the deficit if it’s working & employment figures are static or better than forecast.
    2. Ireland – If Ireland comes out of recession sooner &/or has better growth, then the mood will favour cuts in public spending.

    Unless there’s a random, ‘personality’ type event that blindsides everybody, I think the above are the biggies.

  37. In reply to South Londoner –

    In my (far from brilliant) understanding of Polling Methodology, by picking specific areas you risk ending up with wildly unrepresentitive results. Randomness is the key – I believe (despite being 4 at the time) that the 1992 election was predicted so wrongly because random samples were not used. By being too specific as to where you poll, or who you poll, you warp the result. Granted, it would be useful to get opinions from the local area, but i don’t think you could do that by only polling that area.

    I’m also not convinced you could easily identify the cause of a rogue poll, even if there has been a relativly major local incident/event that could be blamed, it would be very difficult to find a causal link between an event and a change in polling.

  38. I believe Ireland has already come out of recession with the latest quarterly figures showing 0.4% growth. I do not think this will aid the hackers’n’slashers though. Ireland is suffering horribly for this anaemic growth, with enormous public sector unrest in the pipleine.

    Figures released in the last week show that the number of those seeking jobseekers’ allowance has fallen for the first time in nearly two years and the unemployment rate showed its slowest quarterly increase in 18 months. Total unemployment figures are at around 2.8 million, well below the 3 million forecast last year by all the doomsayers.

  39. To clarify, those employment figures relate to the UK not Ireland

  40. Has there been a recent analysis of marginal constituencies or an indication that one is to be published soon? I’m just wondering why, when every poll is followed by the qualification “on a uniform swing,” that more emphasis isn’t given to uncovering what we can expect from marginal constituencies.

    Surely a rigorous attempt at concluding whether a uniform swing is likely and to what extent is of more value than a string of wildly different topline figures?

  41. The problem is that as you tighten the geographic area, the harder it is to get a random sample, and the bigger the sample error you get. And getting a large enough random sample of a single marginal seat would cost the poll taker as much, perhaps more, as doing a national poll.

    You Gov have published one regional at the request of the newspaper publishing it, showing Conservatives would do well in northern Marginals, but haven’t repeated that again. And it’s hard to take anything from it, since it’s one poll, with a suspect sample size for the constituencies.

    We did have a lot of local by-elections in marginals recently, which had Labour doing well, especially in London where they out-performed substantially. But Local Elections are not General Elections.

    It’s be nice if we had finer grain regional polling, then we could do monte carlo simulations on a seat by seat basis to generate a much more accurate seat majority projection. But that’s unlikely to happen since we don’t get reliable fine-grain regional information.

  42. If you have ever used one of the electronic swingometers you can see how floored they are if you say Labour have 0% votes somehow they still get a few seats lol I think anything about 5% will give the tories a small majority.

  43. @Jay. How do you get an R2 of .81?? What do you take and over what time period?

    Anthony: Reading the idiotic commentary in the Independent I wonder if we could go to the PCC and get them to ban innumerate journalists who should know better. The massive innumeracy of the Great British Public is a serious enough problem, and journos should not be wilfilly making it worse.

  44. Labour puts drunks and druggies on incapcity benifits even though there unemployed. The inactive rate is about 8 Million the highest it’s ever been.

  45. @TOME

    Thank you for your reply. I’m an amateur at this but just wanted to hear some of the criticisms to the idea, which I’m sure has been thought of before.

    I should have made it clearer, though, that I would combine the results of seats A, B & C to iron out local prejudices, which would hopefully make the poll more representative. Obviously, the more constituencies covered the more reliable the results would be too.

  46. @NBeal

    MMA, 10 poll recent window, trend generated over the time period of the last 20 polls. And I don’t use Angus Reid’s poll results, since I don’t consider them a true voter intention poll.

  47. @ Cap’n Scooby

    Thanks for the update on Ireland; apparently the 0.4% growth needs to be revised downwards as being attributable to Multi-Nationals electing to be taxed in Ireland.

    If the latest employment numbers don’t deteriorate post X-mas; & if they are widely reported, I’d expect to see Lab improvement trend sustaining.
    I really see those numbers as being key to getting Lab influencing ‘don’t knows’.

    @ Paul B

    Your comments on UK unemployment have zero to do with polling outcomes/ predictions.

  48. @ NBeale

    I thought the detailed commentary in the Independent was interesting especially the analyses by age, gender & ABC etc groupings.

    What did you find idiotic about the commentary? Are there conclusions that don’t align with the actual polling data?

  49. What’s going on with the weightings for the site’s average?

  50. there’s nothing more entertaining that reading all of the “experts” here back- tracking and flip-flopping when polls with different readings come out.

    All the reasoning done here is in hindsight.


1 2 3 4