I wrote last November about the dangers of cherrypicking out figures in crossbreaks to come up with sensationalist stories that don’t actually reflect the truth – and I spend an inordinate amount of time nagging about not paying too much attention to regional crossbreaks. Nevertheless, they never seem to go away.

On Friday, for example, the New Statesman was getting overexcited about the crossbreak for under 25s in the most recent YouGov poll, which showed Labour 47 points ahead of the Conservatives amongst young people. The figures were based on a sample of only 71 people, so the margin of error was about 12 points (in fact, given that the figures were re-percentaged to exclude Don’t knows and Won’t Votes it was actually even lower – only 45 people under 25 actually gave voting intentions, giving a margin of error of plus or minus 15 points.)

If the New Statesman had taken the time to look at other recent cross-breaks for young people it should have become clear that (a) the figures were very volatile, as you’d expect from such a small sub-sample and (b) that this was an outlier. The average figure for the rest of the last week was CON 24%, LAB 49%, a lead a little over half of Friday’s (this is still a very large Labour lead of course, but not unsurprising given they have a 12 point lead nationally and there tends to be a correlation between age and voting intention, with young people more Labour and older people more Conservative).

Another example this week was David Skelton at Platform 10, citing regional cross-breaks from Populus polling to demonstrate that support for gay marriage isn’t just amongst a metropolitan elite, but is actually higher in blue-collar Northern areas. Now, while I suspect David’s ultimate argument is correct (after all, it’s not like only Southern middle class people are gay or get married), the evidence he cites doesn’t really hold up. 81% of respondents in the North East did indeed tell Populus that they supported gay marriage… but it was on a sample size of 45 people, giving a margin of error of 15 points and meaning support for gay marriage in the North East was not actually significantly different to that in London.

Here’s what to remember about cross-breaks

1) Cross breaks often have small sample sizes and are not internally weighted.They are hence very volatile and imprecise, especially for things like age and region where some sample sizes are below 100, and very little weight should be given to them. For a sample size of 200 the margin of error rises to plus or minus 7 points, for 100 it rises to plus or minus 10 points.

2) Where you have a regular tracker such as the YouGov daily poll, the sheer volume of data means it is inevitable that volatile crossbreaks with large margins of error will sometimes produce results that look extreme. However odd these look, unless there is a sustained pattern they are not meaningful. If the actual figure is 50%, but you’ve only got 70 respondents, then you ARE sometimes going to get results showing 62% or 38%… purely from random variation.

3) All this goes double or triple for voting intention polls! For most polls the precise figures don’t matter – it is much the same story if 30% of people support a policy as if 40% do. In contrast, there is a world of difference between Labour being at 30% and Labour being at 40%. When it comes to voting intention, crossbreaks in a single poll should basically be ignored.

A couple of months ago Lewis Baston asked me an interesting question on Twitter. Given that regional cross-breaks on polls are so consistently misrepresented and misunderstood, should pollsters publish them at all? It does make me ponder. My starting point is always that it is good for pollsters to be as transparent as possible, unless there is a good reason not to be open, we should be.

Some crossbreaks are very useful in understanding and interpretting polls – think, for example, of how much voting intention cross-breaks help our understanding of leader approval ratings, best PM figures or my bete noire of “would policy X make you more likely to vote Y” questions. Sometimes they do show interesting things (look, for example, at the huge gender contrast you find in polls on nuclear power or nuclear weapons), or many issues where there is a clear correlation with age. Regional cross-breaks are, admittedly, less obviously useful but there are many instances when cross breaks are extremely beneficial to our understanding of polls if looked out as crude indicators of trends and correlations, rather than taken out of context.

I wouldn’t want to see pollsters stop giving out data, even data of limited use, for fear of it being misunderstood. The solution is really for political journalists to better understand polling and statistics. Some people will always misunderstand or misrepresent polls…but political journalists shouldn’t, they are too important a part of politics today.


73 Responses to “A reminder about crossbreaks”

1 2
  1. Well said – and thanks for the mention. You’re probably right about it being better, if in doubt, to publish the data however flawed the use that might be made of it.

    Another question, if I may – is it methodologically OK to average successive crossbreak numbers (e.g. for regional voting intention) in the YouGov tracker? And if so, how many polls should one average to get a meaningful figure for each region?

    I’ve experimented a bit with this, and doing a five-poll average seems to produce smoother lines that ‘feel’ right, although I would be interested to know your view on whether this is a valid approach.

  2. Lewis –

    It’s better… but still has limitations. Polls are weighted to be representative at the GB level, the sub samples are not internally weighted. So it is, for example, possible for a poll to have too many Labour people in London and too few in the South-East, and still be OK at the GB level.

    Of course, in an ideal world these biases would cancel out over time, but that is assuming no systemic bias. If there is a systemic sampling bias in a particular regional sub-break aggregating up lots of polls won’t address it.

    To give a couple of examples, up until mid-2011 when they adjusted their weighting to treat the SNP & PC separately from other smaller parties YouGov’s Scotland cross-breaks didn’t used to have sufficient SNP support compared to proper Scottish polling. No amount of aggregating data would have solved that.

    Another example is that YouGov’s London cross-breaks in GB polls are often less good for Labour than dedicated London polling – partly because London polls are weighted separately for ethnicity which makes a difference there. Just aggregating up London sub-samples wouldn’t address the need for ethnicity weighting to get London right.

  3. On Friday, for example, the New Statesman was getting overexcited about the crossbreak for under 25s in the most recent YouGov poll, which showed Labour 47 points ahead of the Conservatives amongst young people. The figures were based on a sample of only 71 people, so the margin of error was about 12 points (on top of that, polls are only weighted to be representative at a GB level, the sub-samples are not internally weighted).

    I’m sorry Anthony this is wrong. You have forgotten that of the 71 under-25’s in the sample 37% said they didn’t intend to vote or didn’t know how. So the percentages are only based on a sample of 45 people.

    This is even shown up in the tables because the figures have gone into italics which indicates a sample size of under 50.

  4. Roger – indeed, you are correct!

  5. I don’t ‘get’ why YG can’t incentivise more young people to join their panel & complete political surveys. Why not, e.g., pay them more?
    8-)

  6. Good piece again Anthony.
    I have always been uneasy about aggregation (Lewis’ averaging query same I guess)) as the unrepresentative weightings will lead to incorrect results regardless of sample sizes.
    My question though is, will the level the weightings are unrepresentative by be much different over a series of polls; ie. whilst we should not pay too much attention to aggregate headline numbers can we give any credence to any trends suggested.

    So is it meaningful if 20 YG dailys aggregate to give xzy in for over 65s and the next 20 give x-10, y +8, z +2;

  7. I also don’t ‘get’, in this time of internet panel polling, why the regional x-breaks in monthly political polls can’t reflect GB demographics. They’d still be open to the criticisms mentioned in Anthony’s above post to Lewis but surely it would be a step in the right direction.

    Political seats are determined by geography so regional x-breaks which were just a bit more accurate would be eagerly pored over by poll users.
    8-)

  8. I wonder if by adding together successive cross-break figures might be self-defeating, by the time you get enough to have a worthwhile sample size, the figures might have missed a blip or dip that the main poll detects.

    Perhaps they should be for curiosity value only.

  9. Jim Jam – yes. That should be meaningful. There is still, as I mentioned to Lewis, the potential for systemic sampling bias… but as long as that bias remains constant then long term trends in the data should be meaningful.

    Amber – cost benefit I expect. It would quintuple the amount of time needed to weight and process polling results, for negligible gain, as regional breaks would still only have a sample size of 200 or so and there’s only so much one can do with weighting… you can’t polish a turd, as they say)

    The best one can hope for is for companies to occasionally aggregate up their data over a long time period and then weight the aggregates for each region separately.

  10. Sorry Anthony if that came out a bit snappier than I meant (I’ve also discovered that blockquote doesn’t work). I couldn’t help noting that the Statesman column missed out the best bit which was that the Greens came second in that tiny sample.

    Of course if instead they had looked exactly a week earlier:

    http://d25d2506sfb94s.cloudfront.net/cumulus_uploads/document/co56g0frvp/YG-Archives-Pol-Sun-results-070612.pdf

    they would have found the Conservatives leading Labour among 18-24s by 42% to 28% – surely a massive endorsed of the coalition’s policies among the young.

    David Skelton’s is a particularly blatant piece of cherry picking based on Populus’s large number of regions, but I also think that there may be an accidental truth in it. I’ve noticed in the past that it is YouGov’s North region (which is usually around 400-450 people) which tends to be the most pro-gay – against all stereotype.

  11. Amber

    It’s not just that regional samples won’t be weighted to reflect the GB demographics, it’s that they really should have to reflect the regional demographic. For example London has less over-60s than the other regions.

    The classic example of sub-samples not having the same make-up as the sample as a whole is our old friend the Scottish VI crossbreak. YouGov’s figures are weighted so that there are the expected number of Conservatives in the GB sample as a whole. But this means that for Scotland, where the percentage of Conservatives is much lower there will probably be ‘too many’ Conservatives. Hence the reason why their figures in the Scottish cross-breaks are nearly always a bit bigger that you expect and that dedicated Scottish polls show.

    This sort of bias is systematic – the point about it is that you can’t reduce it by increasing the sample size – another good reason for being careful about cross-breaks.

    Also it must always be remembered that the margins of error that Anthony quotes are minimums. It’s the error you would still get even if everything else is perfect, just because you’re taking a sample. Nothing you can do can make it any smaller. In reality however it will be bigger than that because we don’t live in a perfect world. To make things worse we usually don’t know how imperfect it is. So always mentally add on a bit before you use the figures.

  12. @ Anthony

    It would quintuple the amount of time needed to weight and process polling results, for negligible gain, as regional breaks would still only have a sample size of 200 or so and there’s only so much one can do with weighting…
    ———————————-
    Thank you for replying. I genuinely did want to know what are the road-blocks to better x-breaks.

    Having thought about your reply, I think better x-breaks will happen; it’s just a question of when. The cost to do it will come down because apps will be written to do more of the number crunching. Polling companies will find ways of getting more responses & getting them more quickly.

    I am thinking it is not so long ago that daily polls like YG’s would have been unthinkable due to a combination of cost & logistics. :-)

  13. Thanks for the timely reminder, Anthony.

    Publishing crossbreaks is important despite being subject to regular abuse. They allow greater insight into the reliability of polls and also intercompany comparisons.

    Here are some examples.

    Knowing the 18-24 split is important because it is regularly subject to such extreme upweighting by multiples of 2 to 3. So if the normal pattern of the split (a big Labour lead, but not one of silly proportions) is way out – say a Lab lead of 47% net, or a Con lead, you can reasonably conclude that the anomaly has been exaggerated by a factor of 2 to 3, which won’t have been compensated for elsewhere.

    The split between 2010 Con, Lab and LD voters seems important, as while false recall might distort the picture a bit, you’d hope that the proportions are reasonably close to the actual GE result – as a rule of thumb, Con about 1/4 more than Lab, Lab about 1/4 more than LD. If the proportions are well out, then it raises a few doubts as to the credibility of the poll and a nod in the direction of sampling bias, although that
    must remain only a possibility.

    If the subbreaks over all of the various demographic categories all seem plausible, that gives a bit more confidence in the poll’s findings and in my mind reduces the risk that it might be a rogue poll. Conversely, the 13th June YouGov that had Labour 6% ahead of the Cons in the SE must surely raise a few more question marks over that poll than some others.

    I find the split in the 2010 vote quite informative, especially when averaged over time. It’s clear that a significant part of UKIP support is coming from 2010 Cons, and that Labour are picking up roughly as many (or possibly more) 2010 LD voters than the LDs themselves. Likewise, without subbreaks we’d be unaware of the pattern that UKIP are consistently picking up a disproportionate share of their vote from 60+ voters. Broad trends like this are worth knowing, even though we should be cautious over stating the precise %s with any certaintly for the reasons that you state.

    And the split is useful when comparing polling companies – YouGov seems to find consistently more defecting LDs than ICM, for example. This is worth knowing, because it raises the question of why. My own pet theory is that it could hint at a problem with false recall for companies such as ICM which are trying to retrospectively establish 2010 loyalties, as opposed to companies using loyalties established largely in 2010 on the panel then used.

    So, in summary, if we were to be asked to vote, mine would be to keep the crossbreaks. But as their abuse seems inevitable, perhaps the terms of Lord Leveson’s inquiry could be widened even further to cover such abuse by political journalists.

  14. @ Roger Mexico

    I read your comment, it was a really good expansion on Anthony’s comment to Lewis. I do understand & appreciate your points & especially your illustration using the Macbeth sample.

    In defence of my proposition, I can only say: There’s always a great many reasons for not doing something; it won’t be perfect is always high on the list. ;-) I think in the not too distant future, there will be a firm which decides to go ahead & do it anyway.

  15. Why exactly are the demographic categories 18-24 (7years), 25-39 (15 years), 40-59 (20 years) and 60+ (30++ years)?

    Given that the 60+ grouping may well include parents and offspring from the same family, I’m thinking that useful attidudinal differences could be gleaned by breaking it down into two or three categories – as some other polling companies do.

    Would it be an advantage to shift the spread of age breakdown – would this lessen the need for weighting and produce more accurate polls… or is there something specific about the 18-24 category which needs to be retained?

  16. Anthony

    While taking heed of Roger M’s points about the limitations of any weighting scheme, nonetheless I consider that there are two areas where the weighting schemes of all of the polling companies could be improved.

    The first is to weight seats by the tactical situation i.e. to get the correct % of the sample from Con/Lab and, Con/LD and Lab/LD contests (at least in England, where complications with the SNP/PC wouldn’t arise) . Read any of the Butler/Kavanagh/Cowley election studies and you’ll see patterns of variation in the election result depending on the tactical situation. In many current LD held seats, the Lab vote has been squeezed to single figures to an extent that a model based on current weightings could never come close to replicating.

    As an example, to focus on just one element in this tactical mix, I’d suggest that a robust model ought to have about 6% of its sample from LD seats where the Cons are 2nd, and about 3% from LD seats where Lab is second, to better replicate the national tactical situation. The fact that your own study last year of the Scottish Parliament results highlighted the importance of incumbancy and its absence in determining the scale of the LD collapse reinforces the need to get the right scale of LD incumbancy in polling for Westminster seats.

    The second is to weight by ethnicity. I say this from experience of canvassing/door knocking from a few years back. It didn’t matter how leafy the street or how posh the house, one with Singh/Kaur for example was almost always rock solid Labour, such that we’d often do blind knock ups for uncanvassed households just on the strength of being able to identify ethnicity on the household name. I know that you’ve responded negatively to me on this point before, but I feel that you’re underestimating the scale of ethnic diversity in urban areas outside of London. Given YouGov’s undoubted success in London, I think that should be cause to consider extending your weighting for ethnicity across GB.

    Both factors are, IMO, far more significant factors driving voting habits than the factors that are currently weighted for, with the exception of newspaper readership.

  17. Amber

    Sorry if you thought I was criticising you – it was just that I wanted to point out the difficulties.

    As Anthony suggested, the best solution would be to take the YouGov daily figures for a month or half-month and re-weight them, but that spread in date would assume that the VI wasn’t changing and of course the newsworthiness of polls is that they do change. In addition the decline of the regional press means it’s very difficult to get someone to pay for it (look how hard that is even in Scotland).

    You’re right about YouGov’s particular problem with the under-25’s – as you’ve probably noticed I’ve been nagging poor Anthony about it for months. Some research companies such as panelbase seem to manage a very young profile, but they tend to be very consumer orientated and it may be the panels with a more ‘politics’ image (even if most of their work is otherwise) recruit fewer under-25s for that very reason.

  18. Of course there may well be an “Elephant in the Room”.

    It could be that most journalists writing political stories based on polls are well aware of the limitations on cross breaks and quoting an individual poll out of context but simply don’t give a monkey’s.

    Why let the facts get in the way of a good story, or even a mediocre story if you have space to fill and a deadline. Look at the kind of stuff that fills the headlines.

    Today four out of six Scottish papers are leading on “SchoolDinnerBlogGate”

    I’d be surprised if papers who , as an example, have a particular line on the EU care one iota that they are misrepresenting the data a poll gives them.

    You don’t really expect Fleet Street to act responsibility and give the public an accurate account do you?

    Peter.

  19. Roger – depends how fussy you are about the under 25s in your samples. Graduates and students are easy to recruit. Under 25s who left school without much in the way of qualifications are harder.

    I suspect getting under 25 samples that were too well educated was one of the things that contributed to the over-estimate of Lib Dem support at the last election.

  20. Anthony,

    I share your concerns about the dangers in pushing recruitment to make up the sample.

    The elections in both Iran and Venezuela were ones where commentators in the west got it very wrong.

    One reason was that much of their information came via the Internet and was from people who were both middle class and not particularly well disposed to the eventual runaway winners.

    It is better to have a small sample of the rift people than a large sample of the wrong ones; GiGo…Garbage in Garbage out!

    Peter.

  21. Curse of the iPad again, that should of course be the “Right” people not the “Rift” people, although some in the press to talk a load of…….

    Peter.

  22. “….journalists to better understand…….”

    Four little words that tell me we will still be getting plenty of misinformed comment in the press on cross breaks, and much else besides, despite the noble work of AW.

  23. Anonio: Mark you this, Bassanio: the devil can cite Scripture for his purpose.

    Merchant of Venice….

    Says it all I guess Anthony!

  24. John Murphy,

    Indeed, from today’s herald under the headline;

    ECONOMISTS WARNING ON SCOTS FUNDING

    The first paragraph is in bold.

    AN independent Scotland that ran into financial difficulty might have to be bailed out by the rest of the UK, one of Britain’s top economists has suggested.

    Below that in normal type;

    Alternatively, Robert Rowthorn, Emeritus Professor of Economics at Cambridge University, also suggested a prosperous independent nation would not need outside help and would be a valuable trading partner for the rest of the UK.

    It’s how you tell them!

    Peter.

  25. AW what is your opinion of the summation of cross breaks over several YouGov polls, as IIRC, was done prior to the last GE.

    I rather thought that technique was held to be a good guide, for instance, to Scottish VI.

    BTW others here regale us with views on football (I have them too as it happens) but I am fully engaged with live TV of the Le Mans 24 hours.

    The official web site has the possibility of choosing one’s various individual in-car coverage this year.

    Magnifique! Mind blowing stuff.

  26. I wonder what the hyper-sensitive young Labour supporters found in crossbreak world, will think about their party being bought by a billionaire ? According to Michael Meacher, and he is supported by the GMB, Unite, and Unison, Lord Sainsbury, by bankrolling, ‘Progress’ is trying to buy a party. Apparently this could derail the Labour conference, since Progress is, ‘a party within a party’ according to the Unions. Far be it from me to judge, but surely this is a serious issue, Ed M could be ‘tainted’ by accusations of a sellout to the Blairites ? :-)

  27. Ken – stop trolling.

  28. The Times today carries a page 6 article titled, ‘ Unions take on Blairites in fight for Labour’s heart’. I thought that battle had been won, and Ed M had everything under control, seems not, unless of course Rupert is starting a fight back. :-)

  29. AW………I think this is a serious matter, and will affect polling, mark my words. :-)

  30. Perhaps it would be helpful to publish the crossbreaks with a prominent health warning, preferably in big red letters, headed “FAO POLITICAL JOURNALISTS. YES, YOU IN PARTICULAR.”

    I have no doubt that crossbreaks would still be overinterpreted by journalists but at least they couldn’t plead ignorance if they were pulled up about it…

  31. @Ken

    Cameron will have to throw more dinners ;-)

  32. PAUL BRISTOL………Throw as in ‘Bullingdon club’ or throw as in, ‘ Host’ ? :-)

  33. @KEN

    Host more `Country suppers` perhaps.

  34. SMUKESH……Even someone of my advancing years can see the attraction, all that red hair ! :-)

  35. Czech/Poland game – I’m supporting the referee!

  36. @KEN
    It flashed red alright,someone wasn`t taking notice

  37. Steady 10 point Labour lead in tonights ComRes poll. Virtually no change on previous ComRes. Labour up 1

  38. Another poll out today for the IOS

    http://politicalbetting.com/

    Labour 42% (+1)
    Tories 32% (NC)
    LD 9% (-2)

    Also questions asked about George Osborne.

  39. Would the Hunt vote abstention explain the fall in Lib Dem VI?

  40. I really like this entry here because people often will delve too deeply into the crosstabs when they’re not reliable overall. They give off bad numbers but journalists, who are either lazy or uneducated about polling, will give them away as key numbers. I think crosstabs are most successfully analyzed not for the numbers contained within them but for how much of a polling sample they comprise. Those can help you guage the accuracy of a given poll.

    I doubt that Londoners are less supportive of marriage equality than northerners are.

  41. SMukesh – unless there is a sustained trend, the most likely explanation for any movement of one or two points in a single poll is random sample variation.

  42. @AW
    Thanks

    Would be interesting to see Youguv.

  43. The polling of the Lib Dems in my opinion will not start to improve until they are given credit for any popular policies implemented by the coalition. For example the increase in the personal tax allowance, which is something people may associate with the Lib Dems, has not helped their polling.

    I just think that the purpose of the Lib Dems as seen by many people, being the third party, helping create coalitions when required, is something people are not comfortable with. Perhaps it was the fall out from the tuition fees, which has led to them not being taken seriously. Anything good they have to say, is not considered by people. Yes it is not fair, but any party that helps the largest party form a coalition, may have the same issues.

    I really think Nick Clegg and senior Lib Dems need to think about their strategy going forward. If they wanted to see Hunt face the independent inquiry into breaches of the ministerial code, they should have voted for it. Abstaining, was perhaps not seen by people as a commonsensical thing to do. Normally if you are in favour of something, you vote for it, rather than debate it and then don’t bother voting. Not everyone polled, would understand parliamentary ways.

    In regard to Lib Dem strategy, they need to show a little more independence and talk about issues where they have a different opinion. They should be prepared to vote against anything, that is not really part of the original coalition agreement. If the back 99% of what people consider are Tory policies, then people would then think, what is the point of voting Lib Dem. The Lib Dems have their own policies, which differ from the Tories, but can they afford to wait until 2015 to talk about them ?

  44. ComRes tables are here:

    http://www.comres.co.uk/polls/IoS_SM_Political_poll_17th_June_2012.pdf

    For some bizarre reason (maybe to keep Anthony happy?) most of the questions on Osborne don’t have the Agree or Disagree options. Instead they have the even worse ‘Does apply’ or ‘Does not apply’, which can be full of misunderstandings and even more implicit bias.

    Anyway apparently Osborne:
    Has made too many mistakes to be taken seriously 48% to 25%
    Is too posh to understand the financial pressures on ordinary people 55% to 23%
    Comes across as arrogant 52% to 24%
    Is out of touch with the public 59% to 20%

    But it ‘Does not apply’ that he:
    Is leading the country’s economy in the right direction 25% to 46%
    Is an able politician 31% to 39%
    Should be replaced by someone like William Hague 21% to 34%
    Is doing a good job in difficult times 27% to 46%

    ComRes have obviously tried to balance the questions to avoid the ‘tick to first box’ problem, but I don’t think it helps the problem that the ‘Apply’ wording is confusing

  45. @LEWIS BASTON

    “I’ve experimented a bit with this, and doing a five-poll average seems to produce smoother lines that ‘feel’ right, although I would be interested to know your view on whether this is a valid approach.”

    I found that six poll seperation were good. The average unweighted sample for Scottish crossbreaks is 160 ish, so six polls is almost a thousand, which is a lot closer to an ideal sample. It also reduces the effects of an outlier poll (being 16.67% rather than 20%).

    I started by experimenting with mean absolute deviation using 60 polls, then switched to median absolute deviation, then to 30 polls. Thirty polls is enough data to generate some meaningful averages without being so far into the past that things like veto bounce and the like are affecting the data two months later. YG have five a week, and I keep them in sixes, fifteens and thirties for comparison.

    With the med abs dev, I get Excel to serve up the most recent 30, then sort each party’s 30 from smallest to largest, then apply the MAD calcs to get the outliers omitted, then the remianing polls give me my averages:

    h ttp://imageshack.us/photo/my-images/85/madx.png

    Then I record the MAD after each poll, and keep a running total of each. From this some reasonable trends emerge:

    h ttp://imageshack.us/photo/my-images/266/scot.png

    That’s the Scottish crossbreak MAD data since 27th March. Each dot represents the MAD of 30 polls (I removed the Con, Lib, UKIP, Green & Others to help highlight the trending).

    It’s not perfect, but it does reduce the clutter somewhat from the pure VI graph:

    h ttp://imageshack.us/photo/my-images/515/scotvi.png

  46. Why don’t we have a bit of fun and try to concoct the most ludicrous headline from the poll that Roger has linked to?

    Here’s your starter for 10

    “100% of Plaid voters are social class DE says poll”

  47. Anthony (or others)

    ComRes tables p19 & others

    “Base: All those likely to vote Cons/Lab/Lib Dem (5-10) OR those who are certain to vote UKIP/Green/SNP/Plaid Cymru/ BNP/ Other (10)”

    Am I reading this properly? Do they mean that they count only those respondents for the GB “minor” parties if they are absolutely certain to vote, but count respondents for Con/Lab/LD even if there’s a good chance they won’t vote?

  48. Oldnat – you are indeed correct. ComRes started doing this a few months ago, presumably because they felt they were overestimating support for minor parties.

    It does seem somewhat odd to include the SNP & PC – mainstream parties but with a limited geographic spread – with smaller parties, but there goes.

  49. A reasonable check on the accuracy of your calculations (at least for Scotland, Wales & London where there are occasional proper polls) is a comparison with these polls.

    For example – YG Scottish poll 21/5/12 for Westminster VI

    Lab 40% : SNP 35% : Con 14% : LD 5%

    If there is significant variation from these figures in your data, then there be a systemic flaw in YG methodology which your averaging can’t pick up.

    Indeed that variation does exist. In your graph, the Lab and SNP figures are both depressed – though there seems to be a similar 5% Lab lead.

  50. Anthony

    Thanks. The limitations of the metropolitan mind are always interesting to see in action.

1 2