ComRes’s monthly poll has topline voting intention figures, with changes from their January poll, of CON 41%(+3), LAB 30%(nc), LDEM 17%(nc). The poll was taken between February 22nd and February 24th and the full tables are available here.

ComRes tend to produce the strongest figures for the Conservative party, largely because their weighting is more favourable to them than that of ICM and Populus (equally, they tend to give the poorest figures to Labour), but it remains a substantial lead and the direction of the trend doesn’t support that intriguing YouGov/Economist poll that hinted at Northern Rock’s nationalisation having helped Labour.

UPDATE: The morning after and I’ve had chance to look at the detailed tables. This month ComRes’s weighting isn’t actually unduly favourable to the Conservatives – the target weighting figures that ComRes use are based on a rolling average of only 4 polls, compared to 10 for Populus and 20 for ICM. This means ComRes’s weighting is far more volatile, and this month they are actually weighting past Conservative support to exactly the same level as ICM did in their last poll. Some of the difference between the companies’ figures may still be down to methodology, likelihood of voting, the spiral of silence adjustment, ComRes’s “squeeze question” or Lib Dem weighting, which is different…. but it’s not as simple this month as ComRes weighting the Tories more highly.

41 Responses to “ComRes give Tories 11 point lead”

  1. What could have happened is that when the initial news of about Northern Rock broke, people’s initial reaction was to give some sympathy to the governments position. Now, after a few days of quiet refelection, a differant view is emerging.

    I’m a big believer that you have to wait a few weeks for big events to filter through to the public mood – It could be that ComRes is going to the first of a new run of very bad polls for Labour. I think we’ve got MORI and another ICM to come before the end of Feb? It’ll be interesting to see which way they go

  2. Scottish figures seem equally stark and indeed odd.

    Labour; 31% (+4) Tory; 21% (+10), LibDem;15%(+5), SNP 26% (-19) Other; 7% (+4).

    Given the perils of small samples, I just don’t believe ComRes in Scotland with it’s UK balance and sample of only under 80.

    However the 3% Tory rise seems on the surface of it to come almost all from the SNP who are also down by 3%. Could this poll be an example of a bad sample in Scotland distorting a UK poll.

    The Tories did well in the Scottish budget but so did the SNP, and there is no reason to believe that Labour or the LibDems going up by this amount.

    So I’d be inclined to see this poll as potentially rogue in that if the difference is down to the Scottish voting where there is a different agenda in the news up here then it could distort UK polls.

    Of course it could equally be down to churn with the SNP vote not going to the Tories at all but elsewhere with the Tory increase coming from somewhere else, but all in all I find these results just odd.

    Like I said about the last poll to show a big Tory fall, this big Tory rise doesn’t really seem to have a good reason for it, even with Nationalising NR.

    The slowing economy and house price falls there may be the start of a perfect storm for the Tories to exploit, but this just doesn’t feel right.


  3. On the other hand, Mike “The Oracle” Richardson has been telling you for some time that the Tories are on the move in Scotland, Peter. ;)

  4. GIN,

    Moving is one thing, I have no doubt that they are going up and have said so, but 10% in one moth mostly at the expense of the SNP?

    Remember I am on record as saying we will poll between 25-30% come a Westminster election so I don’t have problem with us being at 26%, but it’s the fact that last month we were on 45% and the Tories on 11%.

    That’s more on the move than an elephant with diarrhoea.


  5. I agree with Peter: the Scottish numbers are nonsense. We wouldn’t even be looking at them if it weren’t for the fact that Scotland’s so woefully under-polled (hint hint).

  6. I don’t believe the Tories are 11% in front but neither do I believe that the Tories are only 3% ahead as in the last ICM poll.

    IMHO the actual situation is somewhere in between.Probably a Tory lead of about 8-9% as YouGov and Populus have shown recently.

  7. I think it seems very clear that we now have an established disagreement between polling companies that has been evident for a long time now. I don’t believe nationalising NR has had any significant impact on any parties – damage to Labour occured when the queues formed outside the banks last year – nationalisation was a very short lived story and the public reaction beyond the inside pages of the broadsheets was a resounding ‘so what’. Cameron’s and Osborne’s slightly hysterical reaction and their inability to promote a coherent solution to the issue over the last few months left them unable to effectively exploit another open goal. Elsewhere we’ve had a major boost for the public accounts with the January tax and spend figures, an unexpected growth in consumer spending, and some excellent figures from manufacturers and exporters – the best for many years. I don’t believe any of this has affected political polls in the slightest, but it may be a sign that the economic disaster is not as certain to occur as some people think, and that would be good for Labour longer term.

    I really think there are problems with the polling methods, and presumably this will run all the way to the election. If ComRes show a 10 point jump for Tories in Scotland, surely anyone must view this as suspicious. Another regular contributor here, (apologies – forgotten who it is), regularly suggests YouGovs internet panel may favour Tories by bringing in more older, internet savvy people into the sample. Whether this is true or not I don’t know, but we have repeated and clear differences between polling companies that must be related to methodology and not real political views or events. This may mean that collectively UK polling is worse now than at any time since 1992?
    I also question NBeale’s use of the weighted moving average and retrospective errors for individual polls. If you construct the WMO from an erroneous set of data, it will not be accurate. Any ‘correct’ polls could show a large deviation from the WMO simply because the WMO is distorted. Unless we know that all polls contributing to the WMO are reasonably statistically comparable the WMO is largely useless.
    If ICM is right, Gordon Brown will remain PM with a good majority well beyond 2009/10. If ComRes is right Cameron should prepare his team for government, probably with a modest but worthwhile majority. They cannot both be right, so what’s going wrong with the polls?

  8. I don’t know how to interpret those tables properly – can anyone interpret the Midlands vote for me, in comparison with the headline 11% lead. Is the Midlands Conservative lead higher, lower or similar to that?

    Many thanks.

  9. Alec , you are totally correct in your observations especially with respect to WMO being not applicable in this case where the differences in the published headline poll figures are mostly down to the different weighting techniques applied .
    On the face of it this Comres poll and the latest ICM poll have totally different result published figures but the raw data is not that different ( this is really not surprising as Comres use the ICM poll centre in Bedford to carry out their telephone interviews ) .
    In these 2 polls the Labour and LibDem raw data figures are pretty close within 1% and yet the published figures differ by 4% , this difference is entirely down to weighting and likelihood to vote adjustments . We do not know whose is the correct weighting if either but they can not both be correct and taking a WMO or any average of the 2 polls would average 1 correct and 1 bad poll if not 2 bad polls and be meaningless . Of the other differences in the raw data , Comres did find a slightly greater number of Conservatives in their raw data and as they always do a higher number of Others ( the latter must be down to the question wording ) .
    Do we have any clues as to whether Comres or ICM weighting is more logical ? I would point to the Comres detailed data comparing vote in 2005 with voting intention now .
    The changes for LibDems are to Labour 6 voters but 17 from Labour net +11 , 12 to Conservatives 6 from Conservatives net -6 , 6 to Others 2 from Others net -4 .
    Overall change is net just -1 voter and yet the published LibDem figure of 17% should give a much higher net negative figure . The equivalent data from ICM is comparable although it is not possible to work out the full interchange with the Others . To me this indicates that the ICM weighting is more likely to be correct than Comres but other commentators can take a contrary view .

  10. Alec-“so what’s going wrong with the polls?”

    Yes that is the question to ask. For the uninitiated like me they are totally confusing.

    In particular, and reverting to ICM vs YouGov, the difference between the two on LIbDem share is stark-and absolutely critical.
    Mark Senior has helped my education on this enormously-but ultimately , a shrug of the shoulders and asigning the difference to “methodology” simply leaves the average reader to fall back on his own hunch…which rather negates the whole point of these polls.

    Yours Sincerely
    Frustrated of Cyberspace.

  11. James , the figures you are looking for are in Table 6 but they are based on very small subsamples and tend to vary wildly from month to month . Others may disagree but I do not put any value in these subsamples . The Nov Midlands subsample for example had LibDems at 22% their best regional figure which is scarcely credible .
    This months Midlands figures are Con 50% Lab 29% LibDem 13% . Don’t forget also other pollsters define regions differently , other pollsters include Wales with the Midlands . Comres include Wales with South West .

  12. Thanks Mark

  13. Mark – there is a stranger thing there actually :) The unweighted figures for recalled 2005 vote aren’t present anywhere in the ComRes tables, but if you look at the weighted and unweighted figures for the Tory vote, there’s the very unusual circumstance that the level of Tory support was reduced by weighting. (Actually having looked at the figures because of your comment, I’m going to have to corret my original article…ComRes normally weight the Conservatives more generously than rival companies, but this month they weighted them to 19% of the sample: exactly the same as ICM)

    Every way you boil it down, on weighting the Lib Dems (or other parties for that matter) it comes down to the simple question of how much false recall you think there is. ICM weight the Lib Dems to 19% of those who voted, ComRes weight them to 17% of those who voted – and therefore as a brute rule of thumb, the 17% score is pretty much what you’d expect from a ComRes poll showing the total of people moving towards the Lib Dems is within -1 of the total of people moving away. On the same figures you would expect to get about 19% in an ICM poll (actually, of course, you’d get higher because of the spiral of silence adjustment, and that’s not to mention whatever likelihood to vote has done).

    Because ICM, Populus and ComRes all assume that some people voted Lib Dem in 2005, but don’t recall that when asked now how they voted back then, the Lib Dems need to get the support of significantly more people in a sample than those who recalled voting Lib Dem in 2005 if they are to match their poll rating from back then.

  14. Anthony , valid comments , with which I largely agree . Any views on Comres consistently finding more Others than the other pollsters .

  15. Mark – that was weighting again. When they first started weighting samples politically they were weighting “others” to a crazy level. Last summer they were weighting them to 7% of the sample (not of voters, of the whole sample), compared to 4% in ICM’s samples and 5% in Populus’s.

    Again, their weighting seems to have come more into line of late. This poll’s sample was 5% people who say they voted “other” in 2005, compared to 4.5% in ICM’s last poll.

    I’m not sure if they’ve overhauled their weighting targets, or whether their formula just happens to have spat out figures closer to those ICM and Populus use by chance.

  16. Yes it is partly weighting which leads to the final others figure but they also seem to be finding more others . The unweighted figure in this poll is 83 people of which 28 are other others , I recall that Populus changed their treatment of the results last year whic reduced the other others figure .

  17. Populus’s change was very subtle. For people who say “other” Populus actually give them an extra question, in the same way YouGov do, that prompts them with the names of various fringe parties and asks which of them they will vote for.

    What they used to do was count anyone who said “other” as an other vote, even if they didn’t know which other they were going to vote for. Now they give a don’t know option in that section question and count those people as “don’t knows”.

    It could be that ComRes are doing the same as Populus used to do – counting people who say “other, but don’t know who” as unspecified others, rather than don’t knows. Another possibility is that it a result of the squeeze question, which pushes out people who’ll vote for obscure strange fringe groups not on the main list.

  18. Anthony,

    ” which pushes out people who’ll vote for obscure strange fringe groups”,

    You better no mean us jimmy….


  19. The Tories will have 11 seats in Scotland just wait and see.

    Scottish LibDems = Meltdown. He He.

  20. Peter – Nah, you’re not even an “other” these days, people prompt for you along with the main parties ;)

  21. Even this subsample of 11 voters out of 53 would only give the Conservatives 5 or 6 seats in Scotland , last month’s Comres subsample would have given them 0 seats still more than they deserve LOL

  22. PETER CAIRNS said :- “Scottish figures seem equally stark and indeed odd.

    Labour; 31% (+4) Tory; 21% (+10), LibDem;15%(+5), SNP 26% (-19) Other; 7% (+4).”

    GIN said :- “On the other hand, Mike “The Oracle” Richardson has been telling you for some time that the Tories are on the move in Scotland, Peter.”

    Both very interesting statements above – Peter seems to have finally come to a sudden realisation that the SNP is not the only alternative to Labour in Scotland and that , as i have stated before – the Scots , like the English and the rest of the UK want change nationally – they know that Labour wins UK elections on the back of the Scottish vote for them – those Scottish figures do I admit seem a little “odd” with the sudden increase in support – but actually reflect a change of thinking in Scotland . They recently removed a Labour leadership at Holyrood and replaced it with a stronger Tory representation and gave the SNP a chance to rule – that will be reflected in the next UK general election in Scotland – 2nd and 3rd place will be fought between the SNP and the Tories – unfortunately for the Liberals they will be pounded into 4th place.

    I stand by my prediction of AT LEAST 11 seats going to the Tories in Scotland at the next UK general election.

    Thank you GIN for your reminder of my past predictions – i did tell PETER CAIRNS previously that he and his party should not be complacent in Scotland about the Tory revival there – he rubbished my thoughts at the time .

    This new POLL is on target for my predictions up to May 2008 – i have suggested earlier that Labour’s core vote UK wise is nearer to 27% – the POLLS to come over the next couple of months will see them getting closer to that figure .

    Once the Liberals and Labour are damaged even more at the forthcoming local elections – the electorate will see that the shift is even more towards change and the POLLING results there after will show even bigger gaps between Labour & Conservative till the next GE.

    Latest “gaffs” by Michael Martin & his colleague about expenses (whether true or false) reflect badly once again on Labour .
    Does anyone know if the latest rumour about the government making everyone buy a £10 proof of age card to buy cigarettes is true ? If it is – I think you will see the 30% of smokers in the voting public thinking twice about voting Labour .

  23. It is simply ridiculous to use 2 surveys both of 50 odd people to try and show that Conservative support is zooming up in Scotland . It is just as likely that the next Comres poll will have them back down in the lower teens and it will be equally meaningless .

  24. 21% is approaching Thatcher-era levels of support for the Tories in Scotland. However I don’t believe this is accurate, however much I wish it were.

  25. Mike,

    “Peter seems to have finally come to a sudden realisation that the SNP is not the only alternative to Labour in Scotland and that”

    Mince, I can only assume that you rarely read what I write.

    I always read yours but mostly because I like a laugh.

    To summaries what I have previously said and what everyone but you seems to have understood.

    The SNP will come second to Labour but on less than 30% of the vote which will probably give us only 10 seats.

    Labour will have their worst result in almost half a century but will still get close to 40 seats.

    The Tories will get close to 20% but probably less and still only get a handful of seats ( that’s FPTP for you, but the Tories support it so can’t complain).

    The Libdems will get 15% or less, but due to a high vote in the seats they hold, will still get close to ten seats.

    My rough rule of thumb for FPTP is to use the proportion of the square of the vote,( It’s not a scientific formula I just use it for illustration).

    In a proportional system if the vote was 40%, 30%, 20%. 10%, then the 59 seats would be 23,18, 12, and 6.

    But if you square the vote and divide by the sum of the squares you get 1600, 900, 400, 100, each divided by 3,000, times 59 to get 31, 18, 8, 2.

    What complicates it in real life is the regional distribution. Labour and the Libdems have pockets of strength ( Central belt for Labour, rural Scotland for the Libdems) where as the SNP and Tories have a more even distribution.

    This makes it harder for the SNP and Tories as the votes that Labour and Libdem pile up in the seats they hold insulate them from a slump in support.

    Therefore on 35-40% Labour will get 35 to 40 seats, On 25-30% the SNP will get 10 to 15 seats, even on 15 to 20% the Tories will get 0 to 5 seats, with the Libdems on 10 to 15% will still get 5 to 10 seats.

    It’s not a precise calculation but it’s a lot more realistic than the Tories on 11 seats. The only 11 seat the Tories will have is on their campaign bus, but if they make 5, which is possible, it will still be their best result in a quarter century.


  26. Sub-samples are of course difficult, hence the difficulty breaking it down into regions in particular – and Scotland is probably a hard country to poll because of 4 parties and very different terrains.

    Of course, the idea is that the quirks one way or another will balance nationally.

    If you add extra weights, you are not making the data more accurate – you’re just making it look like something you want. Weighting should just be a fine tune.

    i stick to my prediction of at least 11 conservative gains in scotland, that would mean that the conservative would have to be on around 26% of the vote or higher and hope the SNP score around 22-24% and no higher, at the present time if the current poll was repeated at the election the conservatives would gain 3 seats off of labour and the SNP would gain 2 seats off of labour on a UNS swing, not 11 seats as mike seams to be talking about, but their is some time yet before the next election a weeks along time in the house of commons anything could happen to any of the parties, and yes their should be more polling in scotland once a month at least and in natinoal polling their should be rules that state that the regions must be split

  28. sorry i stopped their was a local earth tremor in the melton area the regions should be split into different aras in the geographical location, and no it was not a big tremor 2 or 3 maybe.

  29. Do you really mean an actual earthquake?

    I felt one in Cumbria at 1 am – about 3-4 on Richter scale.

  30. Thank you PETER CAIRNS for your response to my comments . If it was taken at face value FPTP in Scotland would give the results you talk about – but there are 11 seats in Scotland which i have listed before are evenly split at the top with the votes – which means that it would’nt take much of an upset or some tactical voting (which i disagree with) to take place to oust a sitting Liberal , Labour or SNP candidate.

    11 is still my figure.

    Yes – a definate earthquake felt here in TEESSIDE at about 12.52am – i thought for a minute it might have been the big financial crash we’re all expecting – but that when it comes will last longer than the 20 second rumble i felt !!

  31. Weighted Moving Average 40:32:18 – ComRes is out by 4. They are by far the least reliable of the pollsters: on average they over-estimate the C lead by 1.2 with a Standard Deviation of 3.2.

  32. Alec: All polling data are “erroneous”. The question is whether there is Systematic Error or not, and if so what to do about it.

    Suppose q is the result you get in the poll, x is the “true” underlying picture and suppose that the underlying bias (Systematic Error) of the pollster is known to be y, then we can write:
    q = x + y + z – where z is the random error
    and we’ll get a better estimate of x by taking the WMA of (q-y) than the WMA of q.

    However estimating the bias of each pollster is not straightforward. It might change over time, and there are sampling problems. Furthermore the data show no real evidence of a Systematic Error: ICM, youGov and Pop have an average error of 0, Comm and I/MORI are +1 and -1 resp. and BPIX is 1.5. These are too small to be Statistically Significant (BPIX is a tiny sample) and pretty much cancel out anyway. In statistics it’s always better to use a simple model unless there is clear evidence for a more complex one, hence the WMA is the best estimate.

  33. The polls don’t seem to be reflected in local by-elections, could the pollsters be experieicning with Labour what they got with tories in the 1980s/90s – i.e. that people currently don’t want to admit they are labour in polls but when they get into a polling station something different happens.

    Has anyone (Thrasher/Rollings) doing any analysis of by-election swings? It will be interesting to see how results in May match the results predicted by the polls (given that turnout is lower than nationals and therefore a biased sample)

  34. Paul Smith – ICM and Populus’s adjustment for the “spiral of silence” works both ways, it was never (despite the popular perception of it) just adding support to the Tories. For the last five years or so it has tended to increase the reported level of Labour support (and more recently of Lib Dem support).

    Rallings & Thrasher still do their local by-election model as far as I’m aware, though I understand they are finding it increasingly tricky because more and fringe parties contest by-elections and their model looks for by-elections contested by the same parties. It’s record of actually predicting election results in recent years though isn’t particularly good to say the least.

  35. So ICM and and ComRes are notable for producing outlier results, whilst YouGov appear to be hitting the average of all pollsters (though may get the end-result – an election – wrong, but we wait to see)! Little sign of humbleness from certain vociferous complainers yet….

    As for the adequate polling sample for Scotland, fifty (ou of a thousand) does seem small, but would eighty be anymore significant? Scotland may be a one of three kingdoms under the Crown but, economically and demographically, it’s significance is less them the sum-of-the-parts of the UK.

  36. Presumably the most reliable test of the accuracy of individual polls is a comparison of their “predictions” closest to a general election against the actual outcome. Here there is a solid piece of data, with no error, for comparison – going back at least to 2005 and 2001. (Although, even here, there will have been changes in the methods of sampling and analysis over the last 7 years).

    No doubt someone has done this sort of comparison – Anthony? NBeale?

  37. John – here’s the final figures from the last election:
    As you’ll see, it’s fair to say they aren’t much help in separating the sheep from the goats :)

    It doesn’t actually give you a very good idea of relative accuracy anyway – obviously you can spot if someone has completely fluffed it, but if everyone is close it doesn’t tell you much at all. There is a margin of error of 2-3% points, regardless of how accurate a pollster is, and beneath that noise it is difficult to pick out the signal. In theory you could have a pollster with perfect accuracy, who through normal sample error gives a result that is, say, 2% too high for the Tories and have a pollster that is consistently 2% too high for Labour that happens, through sample error, to get the answer apparently spot on.

    When all polls have a margin of error of a couple of percent, it’s not really fair to judge them on one poll every five years (though the pollster who does get closest is, invariably, very smug about it :) ). Besides, pollsters do intend tend to alter their methodologies in response to the lessons they learn from elections.

  38. john H yes an actual earthquake i now know it to have been a 5.2 on the scale of earthquackes, but back to the point this poll is bad news for labour but only OK news for the conservative as most of the vote has come from floting voters and the smaller parties labours vote is stable for now so some good news LD vote has also not moved, i think they need the earth to move for them soon or it could be curtains for NC,

    no damage hear from last night earth moving only moderate 3 or 4

  39. Anthony: thanks for the info above – even though, (like much comment and analysis here), it seems to be peering “through a glass darkly”.

    Stuart Gregory: I must admit I was more shaken by the earthquake (which left a couple of pictures on the wall swinging, and me rather bewildered) than by the ComRes poll results!

  40. JOHN H- it’s a few days later now after the tremor but it seamed like it was more powerfull where you live, but all the results are in i was in one of the worst affected areas in your area the tremor was 3.0 hear it was 4-4.5 bgs reedings

    on other news the bbc & sky news reported yesterday that the govenment has lost another disc it was found after being sold on e-bay as a laptop a computer shop found the disc in the back of the computer hiden away, so far as reports say it’s conferdental information ( more damage to the govenment)

  41. Yes Stuart – probably about as much damage to the government as the earthquake did :)

    Let’s be reasonable; with many hundreds of thousands of employees who are lumped together as working for “the government”, and many only remotely connected to the civil service, it is impossible to hold ministers responsible for all shortcomings. The blame in almost all such cases lies with the individuals concerned – or with managers of the particular organisation. This sort of story is dead in the water unless it is very clear that it is due to ministerial incompetence.

    It’s about time we had another poll to chew over!