This month’s ICM poll for the Guardian is out and has topline figures of CON 39%(+3), LAB 36%(-1), LDEM 15%(+1), Others 10%(-3). ICM normally show the best figures for the Liberal Democrats and the worst figures for Labour for methodological reasons, so it shouldn’t be a surprise that they they putting Labour in a worse position than other recent polls, what is interesting is the direction of travel. While different polling companies have methodological differences that tend to favour one party or another, they normally all show the same direction of travel. YouGov’s daily polls, ComRes, MORI and Angus Reid have all suggested a movement back to Labour in recent polls, this (and Opinium’s last poll) show a movement back to the Conservatives. As such I’ll add my normal caveat – sure, it could be the Tory boost… or it could be normal variation within the margin of error.

For methodology anoraks, it’s also worth noting that this the first of ICM’s monthly Guardian polls to include mobile phones in its sample. I don’t know what, if any, impact that has on results (it may well, before anyone leaps to conclusions, be none whatsoever).

The other questions unsurprisingly looked towards this week’s budget. On which team was best to run the economy respondents preferred Cameron & Osborne to Miliband & Balls by 42% to 25%. On prospective budget measures, 67% of people thought the 50p top rate of tax should be kept, 62% would support a mansion tax and 47% supported stopping child benefit for households with higher rate taxpayers.

UPDATE: YouGov’s daily poll for the Sun meanwhile has topline figures of CON 36%, LAB 42%, LDEM 9%. Clearly no movement back to the Conservatives there.

Inevitably many of you reading this will be asking why the big difference. Well, there are a couple of obvious reasons. Firstly there is likelihood to vote. YouGov do not filter or weight by how likely people say they are to vote. ICM weight respondents by how likely they are to vote, and also weight down people who did not vote in the last general election. This helps the Tories and hurts Labour, and according to the Guardian’s write up without the likelihood to vote weighting Labour would have been ahead in ICM’s poll.

Secondly there is how the two companies treat don’t knows. YouGov’s figures are based only on how people say they would vote in an election tomorrow – people who say they don’t know how they would vote are ignored. ICM on the other hand make an educated guess as to how the don’t knows would vote, assuming that 50% of them will vote for the party they voted for in 2010. This normally gives the Liberal Democrats a significant boost.

Finally, there is normal margin of error, the random variation between one sample and the next that is unavoidable in all polls. Because we get daily polls from YouGov we can be fairly certain that their underlying average is a Labour lead of five points or so, and daily polls fluctuate randomly around that: today’s is clearly a little bit more Laboury than of late. From ICM we have just the one monthly poll, but it’s possible that normal variation has produced a sample slightly more Conservative than usual (it’s equally possible that the sample is more Labour than usual, but that would hardly help explain the difference!).

Beyond this there are all sorts of other factors that produce variation between different polling companies figures (what figures they weight to and particularly what political weights they use, when the weighting data is collected, how they sample, what questions they ask, etc). All of these go together to produce the “house effects” of the different companies – which are more favourable to Labour or the Tories or the Lib Dems. Most of the time (with today being an unusual exception that I expect will vanish soon enough), all the companies show the same trend…and away from elections that’s the thing to watch.

75 Responses to “New ICM and YouGov polls”

1 2
  1. Erm…isn’t the idea of 10% of people in the North of England voting SNP…a bit doubtful in that ICM poll?

  2. Fraser Wight

    Erm…isn’t the idea of 10% of people in the North of England voting SNP…a bit doubtful in that ICM poll?

    The 10% could be descendants of exiled Scots who were left stranded south of the border when it shifted and now wish to reunite with their homeland or it could just be a blip in the polling data!! ;)

  3. KEN

    @”Why isn’t it plastered all over the media ? I wonder if it’s because success in business is just a tad distasteful to our commentators, bearing in mind that most of ‘em are prophets of doom.”

    Yes-some of that certainly.

    But also , I think, a pretty widespread lack of understanding about the scale of e-commerce.

    We are still “bricks” orientated rather than “clicks” when we think of the Retail sector.

    It also leads to the ridiculous criticism of privatising Royal Mail. RM’s business in letters is dying-but it has a great future in e-commerce transportation-parcels & packages.

  4. @ RICHARD
    “That’s the main reason I disagree with ICM because I think they are going on past trends and not taking into account that now that people have seen Lib Dems with some sort of power that people have made a much firmer judgment on them than they have in the past. Ie- no longer a forgotten about party until the election itself.”

    A very interesting thought.

    It’s a pretty key factor to have such fundamental polling differences on.

  5. I know crossbreaks can be quirky.

    But Men/Women are pretty big chunks.

    In this YouGov Poll women VI is 34-46

    In the last one it was 39-39


  6. Fraser – I believe “North” on the ICM tables refers to the North of Britain (i.e. it includes Scotland).

    It’s because they used to just have North/Midlands/South as their only regional crossbreaks, including Scotland in North and Wales in the Midlands. At some point they started putting up Scottish and Welsh breaks as well, but I think they still included them in the previous breaks too, if that makes sense!

  7. The largely academic argument about reallocated “Don’t Knows” and squeezed learners is fascinating for the uber- psephologists amongst us, but I come back to a point I’ve harped on about for some time; which pollster is giving us the most accurate picture of the state of public opinion? That’s the key. The methodology minutiae only really matters in terms of its contribution to the accuracy of its findings; its not an end in itself. This idea that one type of poll “favours” one party as opposed to another poll that “boosts” some other party, smacks of quackery to me. Lets take the recent ICM and YouGov polls that did their respective fieldwork over more or less the same period. ICM tells us that the Tories enjoy a 3% advantage yet the YouGov suggests a 6% Labour lead. Clearly, even allowing for MoE variation, they can’t both be right. One of them, maybe even both, are giving us an inaccurate picture of the current state of public opinion, unless of course it’s so volatile that it’s virtually changing by the hour. The likelihood of that being the case is very low, I would think.

    Of course, one of the ironic delights of all this, and I’m as much a potential victim as anyone else, is that all us loyal UKPR posters are probably expending vast amounts of solemn and earnest pronouncements on opinion poll data that is, for want of a better expression, a complete load of old horlicks!

    Still, it’s all good clean fun – most of the time!

  8. @Crossbat11

    I was thinking about this last night – there are really two uses of polls: predictive and descriptive. Predictive essentially means that the poll is used to predict the likely outcome of an election. This is good because it can be measured, and we can sit around going “ooh, wasn’t ICM terrible this year” etc.

    The other use is descriptive, and ranges from @OldNat and his crossbreak obsession through to more general claims around reflecting public opinion. In theses cases there is nothing absolute we can measure against, so we aren’t able to provide a good assessment of accuracy. Instead we rely on statistics to assess accuracy.

    Unfortunately the statistics we use (MOE being most common) both rely on assumptions that probably aren’t true, and are usually misunderstood. And that’s before we allow people to put their fingers on the scale to rebalance the samples.

    In fact it is entirely possible that both the ICM and YouGov polls are simply being hit by sampling error. The MOE we normally quote only means that in 95% of the cases we would expect the result to be statistically within the bounds indicated. In practice that means that 1 in 20 would be outside that area. I refer you to the case of acne:jellybean causality

  9. The YouGov sample looks unusually dodgy. Only 10% more 2010 Con than 2010 Lab plus lots of unusual crossbreaks.

  10. Crossbath

    It’s not one kind of poll that favours but one kind of methodology.

    Allocating don’t knows is always going be a tricky point as pollster are making a fairly heavy assumption about their behaviour (although based on empirical data, these assumptions may or may not hold, they aren’t like the laws of physics)

    It’s not unreasonable to think that someone who voted party X at the last election is more likely to vote for party X than a person selected from the general population at random. How much more likely is a difficult thing to judge and different pollsters have a difference of opinion about this.

    Taking the Yougov figures, a Tory selected at random who voted at the last election but says “don’t know” now is unlikely to have a 42% chance of voting labour (if he votes) at the next election, he’s still more likely to vote tory than a random person selected at random, the question is “how much more likely”?

    The same question of course applies to each of the main parties and the answer could of course be very different for a lib dem than a labour voter who says they “don’t know” now.

    As the election draws nearer, I’d expect the number of don’t knows to reduce and the polls to become closer to one another (and the real picture).

    I guess there is another complication that the set of people who say don’t know at during the midterm could well be different to the set of people who say don’t know all the way to right before an election, and how they actually vote.

    Pollsters want more than anything to get the election right, if that means making an assumption about the person who says “don’t know” 1 week before an election they will use that methodology to get the best predicted figures.

    If there is a different family of voters who say “don’t know” now but will know before the next election, making assumptions that they behave in the same manner as the first “family” will mean that mid term results will be inaccurate but there will never be any way to tell as this family will disappear in the run up to an election.

    I suppose the figures could be there to analyse if pollsters wanted to compare people who said don’t know midterm but had a preference before the election and people who said don’t know all the way up to the election to see if there was any difference in their behaviour. Ultimately it won’t help them get the poll that matter correct so it’s unlikely they would do that.

    *When I use family here I’m talking more figuratively than about blood relatives.

  11. The Times reporting that Labour are not going to make a commitment to restore the 50 p tax before the next election…IMO,this is exactly what Labour must do in order to keep their centrist credentials and draw the centrist and aspirational voters away from the coalition.

    Red Ed is finally learning his lesson…Move to the left and the government gets a free ride as it did after his conference speech.

  12. Her Majesty the Queen has just announced that she will be running 60 marathons in 60 days to mark her Diamond Jubilee.

    I think this is the only way Price Charles is going to get his hands on the throne.

    It was great when Her Majesty reminded the Houses of Parliament, that there has been 12 PM’s during her reign. The saying ” form is temporary, class is permanent” comes to mind.

  13. Colin,
    I have no hang up about Privatisation per se.
    I thing most have been fine even energy, it is a bad job by the regulator that has been the problem.
    The one big mistake imo was rail which just has not worked and along with bus de-regulation which created a shambles in certain places makes an integrated transport policy virtually impossible.
    I agree re the Royal Mail – I was strongly attacked at a local GC meeting for supporting Mandy’s plans for a part privatisation which he ditched in the end under pressure.(vote was 9 to 1 approx against).
    imo the problems in the Royal Mail in recent years could have been avoided with earlier private sector investment. Rules around deliveries to remote areas etc could have been constructed easily enough.

  14. Re Guardian/ICM poll, etc

    Why don’t polling companies just leave the ‘Don’t knows’ to be ‘Don’t knows!’

    We can understand that they ‘Don’t Know’.

    If I ‘didn’t know’, why should a proportion of my ‘Didn’t know’ be relocated to a party that I my have/did/or won’t support.

    After all, the public can understand if people are undecided. To reallocate the ‘Don’t knows’ based on voting intentions of a different era, quite frankly defeats the object of getting a true reflection of the current state of each the Parties support.

    Just list them:-
    Con %, Lab%, SNP%, Lib Dem%, Don’t know%

    ‘Don’t know ‘ is probably the most interesting figure as it reflects the% of floating undecided voter. But we don’t hardly ever see this figure in the poll headline.

  15. Has anybody pulled together information from the various polling companies to allow easy comparison between the raw data and the “tweaked” voting intention so we can see the effects of their different treatment of the don’t knows?
    I would quite like to see the percentage split for
    Who will you vote for?
    Who do the polling company think you’ll vote for?

  16. @Anthony Wells

    Ah gotcha – yeah that makes much more sense!

  17. rhuckle

    “It was great when Her Majesty reminded the Houses of Parliament, that there has been 12 PM’s during her reign. The saying ” form is temporary, class is permanent” comes to mind.”

    Nothing to do with not having to stand for election, then? If the Royals are so popular they won’t be afraid of a free vote every few years, will they?

  18. Paul Bristol

    Some polling companies choose to handle “don’t knows” in a more sophisticated manner because they believe their models of their behaviour create a better model than simply ignoring them in their predictions.

    If the assumption that a don’t know with a previous vote for a particular party is as likely to vote for any of the parties as a individual selected entirely at random is wrong, then that assumption will lead to an inaccurate prediction.

    ICM seem to have evidence that don’t knows do tend to behave slightly differently than a completely random individual (that they revert to their previous vote more often than switching allegiance).

    Whether their proportion of people who “return home” from a don’t know is correct will only be known at the next election.

    Chances are the truth lies between the extremes of Yougov and ICM. It’s rare, although possible that the extremes of a particular model would reflect reality.

  19. Just watched first half of Daliy Politics and Andrew Neil asked the Labour guest why the latest poll shows the cons 3% ahead….
    Don’t blame Neil who, despite his background is quite balanced imo, but the researchers should do better than that.

  20. JimJam –

    Adam Boulton & I had a slightly heated exchange at EPOP a couple of years back about the impact of daily polls, and how by having so many more polls you are going to inevitably end up with more outliers which the media will puff up into pointless and misleading stories. I said I hoped the impact would ultimately be the opposite – by pushing out enough data the freak results become obvious and eventually the media will learn that the trend is the meaningful bit, not the outlier.

    Adam’s stance was rather “don’t give us daily polls, we can’t be trusted to report them sensibly”. To some degree he is probably right, but I’d like to keep on believing that the media’s reporting and treatment of polls is capable of salvation.

  21. JIM JAM


    I agree- ( again!)

    Rail is still not right somehow. THe fair structure is unbelievably complex & the cost of travel way too high.

    I don’t know what the answer is .

  22. What an enjoyable article this is :-

    Power to the People.

    :-) :-) :-)

  23. “Don’t know”

    That can be an honest answer – but it can also be a form of evasion.

    In many cases it will signify an ambivalence about voting for the same party as last time. In other words – “I’m having doubts, I’m sitting on the fence for the time being.”

    This is the interesting group – certain to vote – but not a clearly defined political affiliation… the type of people we all know who can argue passionately from a contradictory standpoint depending on the day of the week.

    They are concerned about what people think of them (a person who takes the reponsibility to vote serioiusly) and they want to be on the “right/righteous” side – and they are perhaps more influenced by the prevailing media zeitgeist in the run up to an election – Tories are divided/nasty, or Brown is hopeless/a bully for example.

  24. Jim Jam @ Colin

    “Rules around deliveries to remote areas etc could have been constructed easily enough.”

    Could have been? Yes.

    Would have been? Never. Not if the aim was to make the deal attractive for the private sector.

    The fundamentalists who think up these schemes don’t like that sort of complication, know nothing of life in Manchester, far less Caithness, and “know” that competition always brings more choice and lower cost.

    Nearly all the trouble in the world is caused by ignorant funamentalists, and they don’t all live in Afghanistan.

  25. John – which is why it would have better done by a Lab Government imo.

1 2