Ipsos MORI’s monthly political monitor is out, with topline figures of CON 31%(+2), LAB 34%(-4), LDEM 10%(nc), UKIP 13%(-2). The three point Labour lead is the lowest we have seen in any poll since a ComRes poll last September (conducted during the Conservative party conference) and the lowest MORI have shown since April 2012.

All the usual caveats about unusual poll findings apply and the full tabs are not yet up on the MORI website, but MORI’s Tom Mludzinski says the change is mostly due to Labour voters saying they are less likely to vote (as regular readers will know Ipsos MORI use the harshest turnout filter, only including respondents who say they are absolutely 10/10 certain to vote. Most other companies either use softer turnout filters, weighting down people who are less likely to vote, or ignore turnout filters completely away from election time).

UPDATE: As with the YouGov/Sunday Times figures for the last few week’s, MORI’s figures also show an increase in economic optimism… or at least, a decrease in pessimism. 30% now expect the economy to improve in the next year, 31% to get worse – a net “feel good factor” of minus 1. This is up from minus 19 a month ago, and the highest since July 2010.

UPDATE2: Full tabs are here. Greens on 6%.


400 Responses to “Ipsos MORI – CON 31, LAB 34, LD 10, UKIP 13”

1 2 3 4 5 6 8
  1. If there were to be the pattering of little Panda Feet this will convert to T+LD> P

  2. @ Steve

    If there were to be the pattering of little Panda Feet this will convert to T+LD> P
    ————–
    I think it could be T+LD<P :-)

  3. Allan Christie

    “What’s the BNP’S policy for Crofting Subsidies on the island of Barra?”

    It will be of similar practical application as those of Labour and Conservative and their policies for Education.

    DC popped across the border as asserted that the Scottish Parliament should copy English changes in education. He didn’t say how many faith schools the should have on mostly catholic Barra.

  4. So-A&E is less busy on a Wednesday night than on a Friday or Saturday.

    Just one abusive incoherant drunk rather than a room full.

    …which actually means that the “nurses” can leave you for even longer staring at the ceiling , because they have more time to chat about The Last Shift, NHS job adverts, The pointlessness of treating determined suicide cases, preferred hair care technology, etc etc.

    Lots of very loud laughing seems to relieve the boredom .

    If you have broken something-or not-it’s great.

    If you need a specialist doctor-its a very noisy parking bay till morning.

  5. @ Turk

    What is the periodical evidence that Yougov polls are any more accurate than the other pollsters.
    —————–
    YG, at this time in the electoral cycle, are about trends rather than greater accuracy (although when compared to actual elections YG polls are usually close)
    —————–
    But other polls seem to be differing to a wider degree of late. Has something changed in the way opinions are being gathered by these companies
    —————–
    ICM reallocate don’t knows based on past vote. Consensus is that, at this point in time, that adjustment minimises Labour & UKIP whilst maximising Tory & LD.

    Ipso Mori downgrade voters who aren’t 10/10 certain to vote. That’s quite extreme at this point in the electoral cycle.
    —————–
    or do Yougov need to change there polling methods
    —————–
    YG do change their polling methods once an election date is known & relatively close. YG make an adjustment for likelihood to vote but not as strict as Mori’s 10/10.
    —————-
    So, if I were you, I’d go by Anthony’s weighted averages but with an eye on the YG polling trends. :-) BTW, Anthony may post you a link of the proper articles which he’s done on different polling company methods. If he does, I’ll be reading it too because it’s hard to remember which does what!

  6. @ Colin

    You must have visited A&E recently as that is exactly how it is.
    I think they need to remember that they are actually paid for their service.

  7. @ Colin

    When you were in charge of an accounting department, were your staff banned from laughing & from talking about anything other than work; & even actually barred from talking about their careers & some work related issues?

  8. @ Blue Bob

    I guess like Colin you’ll be in favour of banning lots of stuff which people in private sector work places do; & in favour of everybody receiving time stamped alcohol tokens so they can’t get blitzed. Maybe just go for prohibition because anything else would simply increase the bureaucracy.

  9. @ Amber

    When you go to work you are there for exactly that.

    Of course having a chat and a laugh with friends is fine but some nurses seem to think that is what they are there for.

  10. I think it is awful how some people denigrate the nurses and doctors. For the money the healthcare professionals earn they do a much more valuable job for society than the overpaid bankers and managers that who have brought the country to its knees. These moaning people should treat themselves instead of going into hospitals if nurses and doctors are so useless.

  11. NORBOLD

    “Subsidies for white crofters only I think”
    ____

    Oooh they might just get my vote. ;)

  12. But bankers aren’t paid by the taxpayer- oh wait.

  13. JOHN B DICK

    “It will be of similar practical application as those of Labour and Conservative and their policies for Education.

    DC popped across the border as asserted that the Scottish Parliament should copy English changes in education. He didn’t say how many faith schools they should have on mostly catholic Barra”
    __________

    Subsidies for white crofters and faith Schools on mostly catholic Barra..The BNP could be onto something!! ;)

  14. @Liz H

    “I think it is awful how some people denigrate the nurses and doctors.”

    Did you watch any of the recent eight-part documentary series about the NHS on the BBC, “Keeping Britain Alive”? It was a timely antidote to the recent kicking that the NHS has received from its habitual and ideologically driven detractors who really came out to play after the Mid Staffs Hospital Inquiry Report was published. The series was also a wonderful piece of public service broadcasting, demonstrating the valuable and unique place that the BBC occupies in our cultural life.

    Maybe a glimpse into our future conjures up an image of Sky, in between it’s 15 minute advertising breaks, documenting a day in the life of a private hospital. Lots of talking heads from Insurance companies extolling the best policies and copious interviews with hospital staff advising on the best process for ensuring quick credit card transactions to ensure that patients get the smoothest admittance to the wards and operating theatres.

  15. @ Blue Bob

    Of course having a chat and a laugh with friends is fine but some nurses seem to think that is what they are there for.
    ————-
    And your evidence for this is what exactly?

  16. @CROSSBAT11
    “Did you watch any of the recent eight-part documentary series about the NHS on the BBC”

    I didn’t but everyone here knows (even if they wont admit it) that the NHS gives a pretty good service considering how little money is spent on it compared to healthcare systems in other countries. It is awful that just to score political points people are prepared to denigrate some of the hardest working , compassionate and caring people in this country.

    Just as well I am not a healthcare professional because if I was I would give the ‘moaners’ the kind of treatment they accuse healthcare professionals of.

  17. It’s fascinating how “ideological/ideologically driven etc.” has become such a ubiquitous phrase. The concept of ideology goes back to Marx: according to him, various systems of thought that seem to be based on a critical analysis of reality are actually products of a deeper economic conflict, which perpetuate particular class systems e.g. Christian theology might seem to be an intellectual exercise, but all MAJOR SHIFTS in the discipline (like the Protestant Reformation) are actually part of class wars. Obviously, explaining something like Protestantism or the Enlightenment as the flotsam of economic substructures has the effect of devaluing it without actually arguing against it.

    Move forward 150 years (and past an interim when ideology was actually considered a very good thing in the communist world- there were even “Ideological Secretaries” in the Soviet communist party!) and the concept of ideology is still with us, but now it’s a term used to mean ‘dogmatic’. Naturally, it’s an irregular adjective: I have principles, YOU have ideology.

    The interesting theoretical framework of Marxism is gone and with it the original semantics of ‘ideology’, but the pragmatics of the word are still useful and get used a lot now that ‘evidence-based politics’ (such as trumpeting every bivariate regression chart that supports your prior beliefs) is very much in fashion and ‘the politics of principle’ is out of fashion. Marx is dead, but his lingo lives on, just as Nietzschean Reactionary Romanticism and Freudianism are dead but we still talk about ‘values’ and ‘lifestyles’ and ‘subconsciousness’. Great Germanic philosophers don’t really die: they just get trivialised.

    And that is why I don’t like it when people throw around “ideological” as an insult.

  18. @Roger Mexico – ” ….we see the percentage and can’t resist it”

    Once or twice I have repercentaged to take account of this, but it’s an awful bother, so I’ll readily allow your criticism – but then I do have my own system for reallocating those who don’t express a voting intention. ;-)

  19. Lizh

    There are certain contributors to this forum who regularly denigrate public sector workers particularly those in the Health Sector. Begrudging them being paid at all and completely ignoring the fact that routinely their private sector counterparts at senior levels are paid on average 5 to 100 times as much.

    As we all know private public Health Care Provision works so much better.

    Just look at the USA

    60 Million with effectively health Service at all.

    The Biggest Cause of Personal Bankruptcy.

    CEO’s of Health Care Insurance companies paid 50 Times as much as the Most Senior NHS Executive for running a business One Tenth the Size.

    Substantially worse clinical outcomes in the majority of areas.

    Huge Disparity in Provision and Prices.

    All achieved for a mere 2.5x as much per head as the useless old NHS!

  20. @TingedFringe / Hal – RE: Omission or inclusion of outliers

    My take on all of this is quite simple. You set an objective for the data, then see if the data can reach it. If not, you reassess your objectives, change your methodology, or ignore the data until it becomes more reliable, making it known that the data is unreliable for the objective required (all in my humble opinion – I have never claimed to be an expert in stats).

    “Otherwise you are just biasing the statistics in favour of your prejudices.”

    If I give you the number 5, 8 and 10 and ask for the integer average (quickly, in your head). Now did you do the sums, or ignore the two outer numbers because 8 sounded about right? The average is 7.66r, or 7.7, or 8, depending on the requirements.

    Would you have been biased for omitting the two numbers, if the answer was correct?

    I think it’s fine to omit outliers in calculations if:

    a) The dataset is made public
    b) The methodology is made public
    c) The objective of the exercise is made public

    All of which give the viewers the chance to challenge a, b or c. Now if we take my ‘MAD’ method, and ask the question:

    “Does the data give an indication as to the voting intention of the UK population?”

    On a UK-wide level, quite probably, but at a cross-break level it’s far less straight-forward. The whole point of grouping the data for MAD averaging is to arrive at a probable ‘guestimate’, given that the cross-break data is far more volatile in nature. Take the two examples:

    h ttp://www.statgeek.co.uk/charts/uk-lib-polls-outliers.png

    5 x 9, 8 x 10, 12 x 11, 5 x 12 (30 values in all). If we took the simple average, we would arrive at 10.6 (1 decimal). However, the MAD method omits the values which are greater than 1 away from the median (11). This means that the 9s are omitted. The resulting MAD value is 10.9, but this is measured over the last 30 polls, so caution is advised. It’s an indicator of the period that the data was generated; nothing more.

    Example 2:

    h ttp://www.statgeek.co.uk/charts/uk-conservative-polls-outliers.png

    This time there are 15 ‘valid’ values, and 15 outliers. This is because the dataset is made up of broader range of values, so more values are more than 1 from the median. In this case the simple average is 30.8, while the MAD is 30.3, which is not too far away. The median in that dataset was 30.5 so 29 or less, or 32 or more are omitted.

    Returning to the three criteria:

    a) Is the dataset available? Yes – I just linked to it.
    b) Is the methodology available? Yes – http://www.statgeek.co.uk/polling/median-absolute-deviation-mad/
    c) Is the objective made available? Yes – Same link, “What is MAD used for?”

    If, however you mean that omission of outliers, then the hiding of them…I completely agree. There will be cases where people accept that tiny fractions of data can be excluded for the sake of cost or simplicity, but that should be made clear.

    In truth, when you compare the MAD results to the simple averages, there’s generally little difference. Most of the time, if you line up the dataset in a chart, as per the images above, people will home in on the centre and arrive at their own guess anyway. It’s not that easy in some of the Scottish cross-breaks though:

    h ttp://www.statgeek.co.uk/charts/scotland-lib-polls-outliers.png

    There we have a dataset which ranges from 2 to 24. With those two numbers, some might guess at the average being 11, 12 or 13 as all seem reasonable. The median is 9.5, the simple average (and by coincidence, the MAD) is 9.9. The more volatile the dataset, the more outliers, and the less accurate the estimate, but it’s pretty close most of the time.

    The method used here isn’t being used to make money, or predict anything. It’s a record of the past, so we can give it a smidgen of leeway when the polls get volatile. As always, the recent 30 polls represent six weeks of data, so 2/3 of it is old news anyway.

    If we look to the Con UK recent 10 polls in that dataset (up to the 10th May):

    Median: 30
    Dataset: 30, 30, 27, 29, 30, 32, 33, 30, 30, 31
    Outliers: 27, 32, 33
    Valid: 29, 30, 30, 30, 30, 30, 31
    Average of valid (MAD): 30

    So the 30-poll MAD is only 0.3 out, and that’s due to 20 other polls. Ten polls is more accurate and faster, but blips in polling throw 10 polls askew almost right away, while with 30 polls the blips take longer or don’t affect things at all (in other words a blip has to maintain itself).

    Long post…to sum up…if the “a, b and c” criteria are met, there’s no reason to treat the dataset as biased.

  21. Steve,

    Are you sure that a sample size of one isn’t an excessively large grouping for judging the merits of publicly-funded private healthcare? Couldn’t you derive the same statistical information by looking at a much smaller number of countries?

  22. Regarding the IPOS MORI weighting of likelyhood to vote.

    In the US Presidential elections Gallup weighed by likelyhood to vote which depressed the Obama vote and enhanced Romney’s – and we know that Gallup was one of the least accurate pollsters in that election.

    So maybe weighing by likelyhood to vote is not a good plan.

  23. I really have to wonder why the Conservative party now seems to be building up a manifesto based on pulling out of Europe, bringing back fox hunting, overturning Gay Marriage and privatising the NHS.

  24. @STEVE

    My dad, who sadly passed away, was in a hospital in London. The award had just two nurses looking after at least a dozen patients. Apart from looking after these patients, they also had to man the phones. It is no wonder the patients were not getting all the attention they deserved. You can’t blame the nurses. You have to blame the hospital for cutting staff and they probably had to cut staff because of lack of funding.

  25. Couper – all US pollsters would have addressed turnout in some way or another, Gallup just presumably did it in a less accurate way than other companies!

    (FWIW, US pollsters have very complicated turnout filters, relying on all sorts of assumptions. UK pollsters tend to just ask people how likely they are to vote. While it may be crude, comparing individual answers to the marked electorate register, it’s turns out to not be a bad predictor!)

  26. “If I give you the number 5, 8 and 10 and ask for the integer average (quickly, in your head).”

    ————-

    Lol. Handy choice there. What if it was 5, 8 and 17?

  27. It’s kinda convenient to say that it’s ok to omit outliers if “the objective is made public”. Well, yes, you could for example have a publically-stated objective to omit outliers!! But what if the objective is to get the most accurate indication of VI?

  28. The BBC have an interesting visualisation of unemployment data up – http://www.bbc.co.uk/news/10604117

    You can see the whole nation’s unemployment rise 09-10. Then we get a slow recovery of jobs, *in some regions*, with other regions actually getting worse. And unfortunately for the Conservatives, the uneven distribution doesn’t appear to be in their favour electorally.

    As I mentioned some time ago, I said that the conservatives needed to show signs of a strong recovery at the start of this year. They haven’t had it, and even if they get a strong recovery over summer, it’ll be too late to get standard of living and employment increases in the regions that need it. Those parts of the economy are the proverbial oil-tankers, very slow to change direction, particularly if only some of their engines are working. “Green shoots of recovery” now are too little too late.

  29. Bill Patrick

    If you can think of One Privately funded and delivered National Public Health Service which has better outcomes for the same per-capita cost as the NHS I will be awfully surprised.

    I doubt if AW would be too chuffed with me producing a comparison chart running to Dozens.

  30. My dad, who is American, had cancer treatment and three operations on the NHS at Leicester hospital in 2010/11 said it was wonderful and the best care he’s ever had. He refuses to move back to the USA because of the lack of a health service.

    Those who attack the NHS won’t know what they have until it’s gone.

  31. steve

    Colin’s posts are mostly a symptom of having a well-nigh pathological dislike of public and civil servants, the NHS and BBC. The proof on this forum is easily found on practically evry thread.

    The irony of a deskbound number cruncher (retired) going on about how lazy nurses and doctors are
    is obviously lost on him. Best just to accept and disagree.

    He’s sort of jolly in his certainty.

  32. Ah nurses doing nothing in A&E.. another mind trick from the Land of Anecdote.

    In this case Festinger’s Theory of Cognitive Dissonance (1957); our natural tendency to rearrange facts to suit our desired outcome.

    You want the nurse to do something.
    The nurse needs to do something else.
    The nurse can’t do what you want.
    Tour priority is not the nurses.
    The nurses priority is not yours.
    You place a higher value on your priority.
    You down grade the nurses priority.

    Therefore;
    Nurses are larking about when they should be treating you.

    Yet again it is, middle income ,middle aged, leaning to UKIP types, that are peddling the myth. Why doesn’t that surprise me.

    UKIP; “The party for Grumpy Old Men”

    Peter.

  33. @Carfrew.

    The point is that if there is a tried and tested method of arriving at a result, use it. I could have said 6,7 and 8 and the obvious average is seven without doing the sums. Does that make the result invalid if the method is tried and tested?

    As long as the method is known, the data is known and the objective is known, the results are open to criticism with scrutiny. If any of the three are unknown, then the results are open to criticism without scrutiny.

    With the case of my setup, the objective is to arrive at the best guess, given the data available, and without excessive manhours being spent. This means that the objective is to omit outliers, but in a reasonable manner, with caveats stated (e.g. this MAD value applies to the most recent poll and the 29 preceding polls).

    My recent UKIP article highlighted two anomalies (a london poll with Con on 19 and Lib on 20, which was outside typical results, and a Scottish poll which had UKIP on zero, while the previous poll, and the two subsequent polls had UKIP on 8). Since I was doing simple averaging for those calcs, it was worth pointing it out, so that the reader can omit said poll in their head and come to a an alternative, if they wished.

    It’s not ‘convenient’ to omit outliers. It’s the point of the exercise when dealing with cross-breaks, if we are to arrive at a more reliable estimate of the reality. We can take the UKIP example:

    8, 0, 8, 8 – Average: 6 (or 8 if you omit the outlier). The poll right after it was 12, so if we then try: 0, 8, 8, 8, 12, the average is 7.2, but the outliers of 0 and 12 omitted make it 8.

    Hence why I do these things over 30 polls. The ‘rogue values’ are just that. They appear rarely, so are easy to omit from the calcs, but are included in the dataset for public viewing.

    There’s no easy way to do it. The easier the method used, the more likely the result will be criticised.

    If I can sum up:

    “It’s kinda convenient to say that it’s ok to omit outliers if “the objective is made public”. – The objective, the method and the dataset must be public. Or are the latter two criteria outliers?

    :p

  34. “I doubt if AW would be too chuffed with me producing a comparison chart running to Dozens.”

    No, I wouldn’t. I don’t think I’ve ever seen a discussion of the NHS in a website comments section that hasn’t quickly devolved into a battle of stupid anecdotes, rabid partisan accusations and false dichotomies. While I’d like to think that this will break the mold, I doubt it will so let’s leave it now.

  35. @Carfrew

    5, 8, 17

    The median in 8, and the valid values are 5 and 8, so the MAD is 6.5. The average is 10.

    A perfect example of why MAD should be used on larger datasets.

  36. In fact, correction…the only valid value is 8, while 5 and 17 are outliers, so are scratched. This emphasises my last post statement on MAD.

  37. MrNameless

    “Those who attack the NHS won’t know what they have until it’s gone”

    That is a certainty. All those complaining about A&E staff chatting & laughing (good grief !!) can always choose to attend a private A&E service where staff will be on hand to attend to their every whim & who most definitely would not be laughing or chatting..

    Oh……

  38. @ AW

    Our posts crossed. Apologies.

  39. Yes I know how to calculate the average. You chose your example as a means of showing it’s ok to go for the middle value and omit the others as outliers.

    I chose another example to show you can’t rely on that. And your response is to cherry-Pick another example…

    If outliers are not so common, it’s not going to save much effort to exclude them. And if you are doing it to save manhours. it is a convenience.

    Also you are assuming they are “rogue”…

    Sure you can give “caveats”. But this does not automatically mean it is more accurate. That’s the matter at issue.

    Can you explain why your method is guaranteed more accurate?…

  40. @AW

    I noticed in the Gallup numbers two sets the first which was the headline was the ‘likely voters’ that was what they were called and the second which was more accurate was ‘all voter’. In Ipos Mori the ‘all voter’ percentage is very similar to YouGov – so I wonder if weighing by likely vote is producing inaccurate polld

  41. @ AW

    No, I wouldn’t. I don’t think I’ve ever seen a discussion of the NHS in a website comments section that hasn’t quickly devolved into a battle of stupid anecdotes, rabid partisan accusations and false dichotomies.
    —————
    I beg to differ. This one didn’t devolve into a battle of stupid anecdotes & rabid partisan accusations, it started with them! Colin & BlueBob’s posts were the epitome of stupid anecdotes & rabid partisan accusations – although I’ll grant you, I quickly piled in with the false dichotomies. :-)

  42. Couper – they have generally made polls more accurate in this country (YG, incidentally, do weight by likelihood to vote during election campaigns. MORI’s filter is particularly harsh, so sometimes makes a big difference. For most companies the impact is much smaller)

  43. @Carfrew

    “Yes I know how to calculate the average. You chose your example as a means of showing it’s ok to go for the middle value and omit the others as outliers.”

    No, that’s not what MAD does. MAD takes the median and uses it to calculate the valid values based on their distance from the median. If their distance is more than the median of the all the distances, they are discounted as outliers.

    Take: 2, 5, 6, 6, 9

    Median: 6

    Distances from the median: 4, 1, 0, 0, 3 (rearranged to 0, 0, 1, 3, 4). The median of that is 1, so all values with a greater distance are omitted (2 and 9). The average of the valid values is 5.7, where as the original average is 5.6 (as usual the numbers are close).

    “I chose another example to show you can’t rely on that. And your response is to cherry-Pick another example…”

    I cherry-picked your (only) example.

    “If outliers are not so common, it’s not going to save much effort to exclude them. And if you are doing it to save manhours. it is a convenience.”

    Refer to my original long post, specifically the Con UK and Lib Scotland data. Outliers make up as much as 15 values of 30 if the dataset is volatile enough. The point isn’t to cherry pick the data we want to see, but to try to get an estimate of what’s really happening. The Scottish Lib data (no cherry picking; this is the up to date most recent 30):

    2 4 5 5 5 5 5 6 6 7 7 7 8 9 9 9 10 10 11 11 12 12 12 12 12 13 14 14 16 24

    For what it’s worth, the MAD is identical to the simple average in that dataset. Sometimes the MAD is superfluous, but occasionally it prevents polling blips from skewing the data. That’s the real point. To try to get a reasonable estimate from small samples and high MoE.

    “Can you explain why your method is guaranteed more accurate?”

    No more than you can for your method. I never claimed it is guaranteed more accurate (I did state in one page that it can be less inaccurate, which is a fair statement). I merely offer up the data and leave others to decide for themselves.

  44. When I said cherry-picked, I referred to your other example.

    Back to the previous example. 5, 8, 17…. and then a few days later… you get 27. Are you going to have to start ignoring some data you used before while reinstating others?

    I understand that MAD does what it does. I am just trying to find out why you can be sure omitting outliers won’t distort things. But you said you can’t give that guarantee, so there’s no point me pressing on it.

  45. @Statgeek

    Post immediately above obviously addressed to you.

    Seems a good moment to go grab an espresso…

  46. “Back to the previous example. 5, 8, 17…. and then a few days later… you get 27.”

    Eh?

    “I understand that MAD does what it does. I am just trying to find out why you can be sure omitting outliers won’t distort things. But you said you can’t give that guarantee, so there’s no point me pressing on it.”

    Out of interest, do you have a guarantee that YouGov’s data in any given poll is representative of the population of the UK?

  47. @lizh.

    NHS funding has been ring fenced and protected under the Conservatives. This is factual.

    That’s my only comment, as I agree with what Anthony said that most of time it’s just anecdotes from left or ring about why it’s great/inefficienct etc.

  48. oops, left or right I mean!

  49. Lol, you didn’t address either of my points. Then you switch to something else..

    I am not seeing how your YouGov example means it’s ok to dismiss data as outliers. Can you explain why?…

  50. @Rich

    However, the government has transferred the cost of portions of the local government social care budget to the NHS, and back benchers are calling for this to be expanded. This is an unusual definition of ring fencing.

1 2 3 4 5 6 8