ICM’s monthly poll for the Guardian has topline figures of CON 37%, LAB 34%, LDEM 21%. The changes from the last ICM poll are Labour up 2, with the other two parties unchanged. The poll was conducted between the 15th and 17th of February.

The poll continues the pattern we’ve seen since September last year of Labour doing comparatively better compared to the Conservatives in ICM polls done for the Guardian than in polls done for other clients. As I said when I first commented on this apparent pattern, I can find no obvious explanation for it, but as the months go past the patten seems to be consistent. The shift in voting intention from the last ICM/Guardian poll, which may be the better comparison, is Labour down 1 and the Lib Dems up 1.

The rest of the poll concentrated on attitudes towards taxataion. Forced to chose between tax cuts and reduced services or sustained spending, 51% said they would chose sustained spending with 36% backing tax cuts. What to make of this question depends largely on the wording – it is implied in the Guardian’s coverage that people were presented with the choice of existing spending or tax cuts even if it meant cuts in spending for services like the NHS. In practice no party will ever go into an election promising tax cuts at the expense of the NHS: parties promising tax reductions will present them as being funded in more acceptable ways, while judging from past election campaigns their opponents will try to paint any promised cuts as being funding out of whatever public spending is most popular. How popular tax cuts actually are will depend upon which of these various claims the public actually believe.


83 Responses to “Tories 3 points ahead in latest ICM poll”

1 2
  1. Anthony – Guardian website has Labour on 35 per cent and Lib Dems on 20. Any idea why there’s a difference?

  2. those are last months ICM/Guardian figures .

  3. P.S. You must forgive a little levity among Labour supporters after endless Conservative gloating. But now, back to the serious comments about rogue polls!

  4. So now we have had multiple polls from YouGov with about a 9 point Tory lead, and multiple ICM polls with leads of about 3 points. I am surprised by this reuccuring 6 point difference….they cannot both be right. Either one is right and the other very badly wrong, or they are both fairly badly wrong. I thought that modern polling methods were too accurate to create these kind of regular idstorions between pollsters…

  5. Once again Guardian/ICM is more favourable for Labour than all other polls including ICM for other papers. Its coming up every month so sorry if I missed it, but do we have any methodical difference between the ICM polls for the Guardian and for others?

  6. WMA 39:33:18. It is straining credibility to the limit to believe that the Northern Wreck fiasco won’t damage Labour further, so I think this poll is suspect (possibly the tax cuts questions framed the response?). However if you look at the retrospective errors both the last ICM polls were out by 3 (2.8 and 3.2 to be precise).

    I suppose it’s possible that net-literate voters who are more exposed to the blogosphere are more against this government, whereas at least some ICM interviewees might get their information solely from the Mirror?

  7. very odd this happening every month, but a patern is being set and that is that both parties are leveling out and the lib dems are still stuck in the mid to high teens, i’ll have to see when the field work was done and where it was done beacuse this affects the top line figures overall a good balance in my eyes would be 250 voters from each region or 500 from each block regeion i.e SCOTLAND,THE NORTH, THE MIDLAND, WALES,LONDON, THE SOUTH all seperate areas not blocked together like some polling done by YOUGOV,ICM,COMRES and POPULUS, all of the above do this apart from MORI and even their region’s are slightly odd.

  8. Could it be that a large section of the public is quite happy to see Northern Rock nationalised? Could they actually be blaming the managers of Northern Rock for the fiasco? Perhaps they are quite happy to see Hedge Fund shareholders walk away with nothing rather than taxpayer’s cash.

    Could the Tories be backing the wrong horse by siding with the “heads I win, tails you lose” fat cats?

  9. Philip – I looked before, but I couldn’t find any difference at all between ICM’s polls for the Guardian and their polls for other media. When I get time I’ll try again and see if there is anything obvious that I missed, but so far I’m struggling to come up with any explanation for the difference.

    NewElephant – the fieldwork was before the announcement, so it can’t have had any effect on this poll.

  10. Isn’t the main difference between ICM & YouGov ( indeed with all other Polsters), the Lib Dem result?

    YouGov produces the lowest Lib Dem figure, and ICM the highest-the difference since end October 2007 being 4 % points.(Average of 9 YouGov polls & 8 ICM)

    The average Lab figure in YouGov & ICM Polls since end October is in fact identical at 33%.

    ICM-and only ICM- have given LibDem 20% or over since end October-in four Polls of eight

  11. Isn’t this difference based on likelihood to vote, with ICM including those who are currently 7/10 – thus increasing Labour’s share compared to other pollsters? I think this is evidence of the Labour vote still largely being there and not having switched to Cameron, just being less positive about voting at this point.
    And no comments please about this being a rogue half term poll – it may have been half term in London but not elsewhere.

  12. Warren – the differences between the pollsters are all pretty clear (the political weighting targets, treatment of likelihood of voting and the ‘spiral of silence’ adjustments are the main ones really). The mystery is why the apparent difference between ICM polls for the Guardian and ICM polls for other clients.

  13. The differences in the LibDem figures between ICM and Populus and Comres are down to the different weightings applied . Some of the Populus and Comres polls would have had LibDem figures of 20-21% if ICM weighting had been used instead of their own versions . It is a matter for debate as to whose weightings are correct ( if any of them ) .
    One of the factors in ICM polls is their likelihood to vote adjustment where they take into account those 7/10 or higher on the scale . The effect of this is easily calculable from the detailed poll data . The 2 previous ICM polls ( detailed data from the new poll not yet on line show negligible effect on LibDem or Others figures but a 2% lower Con figure and 2% higher Lab figure than just taking those 10/10 on certainty to vote and a 2% higher Con figure and 2% lower Lab figure than taking everyone who expressed a voting opinion . It is a matter of personal preference as to which you think is better , taking 10/10 only would imply a GE turnout of circa 50% which to me is unrealistically low and taking everyone would imply a record GE turnout , 7/10 and higher is a reasonable compromise , though personally I would only take 8/10 and higher .
    The differences between Yougov and ICM are much more fundamental and clearly go beyond weighting adjustments . It could be that both are wrong and the truth fortuitously is in the middle amd Mr Beales standard deviation measures he quotes are meaningful but it is statistically more likely that either Yougov or ICM are way out and the other reasonably close . I have pointed out that Yougov samples imply a GE turnout of nearly 80% and that these extra supposed voters are overwhelmingly Conservative and that local election results are more consistent with a LibDem performance a la ICM than a la Yougov . It is up to proponents of Yougov to give some evidence as to why Yougov are nearer the true picture than ICM .

  14. But ICM’s Labour share isn’t more than for other Pollsters is it?-in fact it’s the lowest.

    It’s the Lib Dem share which is different

    Average of all Polls since things changed,end last October:-

    YouGov(9) 42.0/33.1/14.7
    ICM – (8) 38.9/33.0/19.1
    Pop.- (4) 38.3/33.3/17.0
    MORI- (5) 40.4/34.4/15.0

  15. Mark – the third difference the the ‘spiral of silence adjustment’, or more practically how pollsters handle don’t knows. ICM and Populus reallocate proportions of those saying don’t know to how they voted last time, other pollsters exclude them from the final figures. There is also a difference between the way ICM do it and the way Populus do it – ICM reallocate 50% of everybody’s don’t knows. Since the last election Populus have reallocated 50% of Tory and Labour don’t knows, but only 30% of Lib Dem ones and none for others.

    It’s the easiest factor to quantify since ICM put the pre-adjusted figures on their table. Across all ICM polls since October the average effect of the adjustment is to reduce the Conservative share by 1.2%, increase Labour’s share by 0.5% and increase the Lib Dem share by 0.7%

  16. Colin , you would expect ICM to have a higher LibDem figure than Populus and Comres because their weightings are more favourable to LibDems just as Comres are more favourable to Others and Populus to Labour and sometimes Others . Of course we don’t know whose weighting is the correct one but at least we know why there is a difference . Mori is a different animal completely and you have to distinguish between their face to face and telephone polls as the former have around 2% higher Labour figures than the latter .
    The Labour figures of Yougov/ICM/Populus are pretty much as near agreed as you will see ( which in turn implies ICM have got it about right with their 7/10 weighting on likelihood to vote ) .
    This leaves the main difference Yougov’s higher Con and lower LibDem figures especially but not solely compared to ICM .

  17. Anthony , yes the spiral of silence adjustment is quantifiable and one of the things we can debate technically as to whose his better . I am still puzzled by the last Populus poll where the weighting adjustments which in the previous poll moved the LibDem figure up but in this poll boosted the others figure instead . This meant that the topline LibDem figure went down although the raw data figures were marginally better .

  18. This discussion is very informative when it comes to how the organisations weight, particularly Anthony’s post @10;55. Knocking off 1.2% of Tory support seems an odd outcome as it is widely believed they are the more likely to vote and the most likely to intentionally keep quiet.
    We shall see. There is no ‘weighting’ up or down on election day.

    Guardian phenomenon is intriguing.
    We are always told to attatch significance to the trends in polling, rather than the headline figures, but on this basis, the Guardian ICM variation is one of the most consistant ‘trends’ in polling over the past few months.

    If this tendecy continues over the next few months, where does that leave this polls credibility?

    Should we look at the trends in ICM Guardian polls in isolation from other ICM polls?

    Do polling organisations tell their interviewees for whom the poll is being conducted?

  19. Mark – the effect of weighting can work both ways. There is a tendency for the raw samples to be “too Labour” and for weighting to therefore help the Conservatives and Lib Dems, but it doesn’t always happen like that and it happens to varying degrees. Normally pollsters will have to weight up people who say they voted Lib Dem in 2005 to get the desired proportion, but it doesn’t have to work that way – fate could deal them a raw sample with too many 2005 Lib Dem voters and they’d end up weighting them downward.

    The large majority of people who vote Lib Dem in polls at the moment are people who voted Lib Dem last time round. In the Jan Populus poll the raw sample had 10.2% people who voted Lib Dem in 2005, and they weighted it up to 12.6%, so those people were given a weighting of 1.23 (on top of weightings for all the other things Populus weight for). In Feb their raw sample had 10.7% people who voted Lib Dem in 2005, so was higher to start with, and the target for weighting had dropped to 12.2%, so the weighting applied out have been only 1.14.

    Or did you mean something else. Looking at the last two Populus polls one striking thing was that far more 2005 Lib Dem voters said don’t know in February than in Jan.

  20. Sally – careful not to confuse things, the 1.2% is the average effect of how ICM re-allocate don’t knows. Their weighting is an entirely different thing (and, incidentally, is beneficial to the Conservatives. At some point I intend to start doing a nice series explaining all these things that I can link to as a FAQ down the sidebar.

    The Guardian thing is weird, since there is no obvious explanation. They appear to be done using identical methodology, so we should treat them as a single data series… but the trend has been consistent over recent months. As far as I am aware interviewees will not be told who the client is for a poll, at least not before they answer the questions.

  21. Anthony.
    Para 1. OK. Gothcha. Thanks. It’s obvious now you mention it. Was not thinking.

    Para 2. Does that mean it is left to the discretion of the pollster? – They can do it if the client requests it or if they think the results are more likely to result in repeat business? Or can merely vary procedure in line with the chosen method of the person in charge?

    I am not aiming this at ICM/Guardian in particular. I understand that there is a voluntary code of conduct for all, which surely covers thing like this. Any ‘pre’ questions/statements could give rise to a shift in response.

    Mike’s PB site recently ran a comment from a pollster [MORI/IPSOS – I think] when their poll results ‘felt wrong’ and had been rechecked.
    Obviously, they were aware of a situation that might draw comment and so sought to address it.
    Would you expect any comment on this situation from this pollster as it increasingly draws comment? Mike S ran a thread when the last ICM for G published.

  22. Gotcha not..

  23. Normally it wouldn’t be done because it could skew the answers of a survey, if you were doing a survey commissioned by Coca Cola and it asked you to list soft drinks you liked, you’re quite likely to be influenced by knowledge of the client.

    In 1998 ICM’s standard script was “Hello, I am telephoning on behalf of ICM the independent social research organisation. We are conducting a research project which requires us to talk to a representative sample of people throughout the country on issues that affect all people. We have selected your telephone number purely at random and would greatly appreciate your help for a few minutes to answer some simple questions.”

    that’s a decade ago so it may have changed slightly, but I would be very surprised if they mentioned the Guardian to respondents because it would indeed have the potential to skew the answers.

    I might drop a note to Nick Sparrow and see if there is something being done differently for the Guardian polls.

  24. Anthony. Have you asked the pollster to explain the anomaly? If their integrity is at stake, they would want to clear it up.What mathematically are the odds that this is a chance event?

  25. Who the client is if told could be an even subconscious factor, people associate the Guardian far more with the left and the Telegraph with the right after all. But if its not told then it shouldn’t be a factor.

    Has anyone from ICM commented on this trend? Could you or somebody ask someone from ICM for a comment if not? ‘Watch the trend’ is a good motto but this is standing out as a bigger trend than almost all others.

  26. Sorry, I wrote that post without having seen your one at 12:40

  27. Thanks…

    [Have you broken down the Scots results?
    Can’t belong before you are have to hand in your homework….]

  28. Sorry if this has already been asked, I haven’t had time to read all the posts yet, but do ICM say who they’re doing the poll for when they ask the questions? Perhaps people are more likely to say they vote Labour when polled by a more left-wing paper even if the actual question is identical?

  29. Oops – Sally did already ask that, sorry.

  30. Steven – the effect would be more likely to be a non-response bias: people with right wing views thinking ‘I’m not doing a survey for that bunch of pinkos’. That said, I would be very surprised if ICM were giving the name of the client in their interviews.

  31. Philip and Collin –

    I’ve just emailled Nick Sparrow at ICM about the difference and he assures me that it is indeed pure co-incidence. It’s a bizarre pattern, but there goes: strange things sometimes happen.

  32. Mark Senior:-

    “This leaves the main difference Yougov’s higher Con and lower LibDem figures especially but not solely compared to ICM”

    Thanks Mark.
    It sure makes one hell of a difference which of these two is nearer the truth!
    Labour supporters would appear to know, with some confidence, where they are from either YouGov or ICM.
    That is not the case for Cons & LibDems.It would appear that they have to choose one or the other.

  33. These postings make for better & interesting reading – no more running for the hills at a rogue POLL – purely examining the methodology used for the outcome. Very sensible !

  34. I’m not happy with Osborne’s performance on Northern Rock. It is so obviously political mud slinging when we as the responsible opposition and a feel for financial matters should be trying to make sure a success is made of this.
    I find it no surprise we are slipping in the polls with these school boy attacks, and urge the party to put Michael Fallon or Dominic Grieve in his place.

  35. Joe, I agree with you. I do think Osborne has miscalculated. I can’t know what the post-nationalisation polls will be like but the sense I get is that most people are understanding of the government’s position even if they do not necessarily agree with it and appreciate it is a difficult situation. I get the feeling that the Tories are going over the top in their criticisms of the government generally at the moment and that it may rebound on them. We shall see in due course!

  36. Paul – good points.

    They think they’ve finally got something on the competence card and are going on and on about it in a highly personalised way. I think it’s could backfire – somewhat.

    Michael Fallon spotted serious flaws in the run up to this crisis, but I think would take a responsible line on “where we are now”. Osborne is out of his depth.

  37. Joe, it is the same on the latest data fiasco. Everything is being portrayed as the ultimate incompetence by the government. Clearly the government must take responsibility for what goes on under its watch but the vehemence of the attacks on Brown, and other ministers whenever some clerk has messed up just doesn’t seem credible. The truth is that if we have a Tory government they too will be victims of officials cocking up from time to time

  38. Does anyone have any local council by-election figures. My impression is that the Tories are struggling to make gains.The LibDems might be slightly improving recently.I think many Tories look at Vince Cable with envy.

  39. I think my objective answer is they are very patchy. I can give you Central Offices summary (as a Tory) but they put their gloss on it – if they don’t say what the change in the vote is it’s a fall, if they boast about an increase, it is a good result.

  40. Wolf, very good council by elections results for the Lib Dems on 7th Feb they took three seats from the Tories, with nearly 60% in Tavistock (West Devon BC) and 50+% in Portsmouth. I think the 21% poll rating is about right.

  41. I have been polled before, by MORI, and I was told first the company it was on behalf of. I’m sure it’s the same for newspapers, and yes my experience of that company had an effect on my answers.

  42. 5 months in a row now? Or longer? At what point does co-incidence become symptomatic?

    Of course there are bigger coincidences in politics and we can always find some often when there’s nothing there in truth. But the odds of this must be quite low, but then again so many things happen that the odds of some quite low odds things happening is actually quite high ;)

  43. I suppose it is the same sort of likelihood of tossing a coin and 5 heads in a row coming up or 5 reds in a row on a roulette wheel .

  44. Devonian:-

    “I think the 21% poll rating is about right.”

    which would mean-in the context of previous discussion-ICM are right & YouGov are wrong…which would mean the 9 point Tory lead is non-existent-it’s more like 4

    This really matters.

  45. Mark,

    It’s 5 polls of each so isn’t it more like getting ten heads in a row (or 10 alternate heads and tails in a row really). I know it’s dodgy maths but that would be roughly a 1 in 500 chance. I’m finding it really difficult to believe that it’s just coincidence.

  46. That’s very dodgy Maths actually. We should tske into account that if any of the other pollsters had made similar predictions we’d think it was just as strange. We also have to take into account that we’ve have been looking at hundreds of polls so odd patterns are bound to turn up from time to time.

    Hmmm… I think I’m going to make a complete U-turn and say it is just a coincidence. Lucky I’m not a politician.

  47. Steven , on refection you are more correct , as the first “head” had already occurred the odds are then the equivalent of 9 more alternate tails and heads ie 1 in 256 . I agree if I were ICM I would be looking for a reason more than pure chance .

  48. Mark: not really. A red is pproximately 50:50 (excluding 0 and 00 in the US), a heads is 50:50.

    The deviation of the Guardian ICM polls can be far down the margin of error which is much less likely than a 50% chance.

    So 1/256 when the chances were even for an individual one is much less likely when the probabilities of each are plugged in too.

  49. Anthony.Could be coincidence. However, if it continues, at some point that explanation is unacceptable.For you, can you say when this point would be reached?

  50. The full ICM poll data is on their website . This month the likelihood to vote 7/10 filter has negligible effect on the Labour figure but adds 2% to LibDem and subtracts 2% from Con figures compared to the 10/10 figure .
    The net changes of those who voted in 2005 is LibDem to Con 8 voters , Lab to LibDem 3 voters consistent with the overall published LibDem figure of 21% .

1 2