ComRes have an interesting post over on their site about differences between online and telephone polling so far this year (as well as making some extremely sensible points about the polls not being all over the place). As they correctly say, telephone polls this year have been showing a tiny Conservative lead, online polls a tiny Labour one. It’s only a small difference, but it’s there and it is not new – at the start of the year I produced a chart showing house differences between the different polling companies over 2014, and even then an online vs telephone tendency was observable: the two most “Toryish” polls were Ipsos MORI and ICM, both done by telephone. The most “Laboury” polls were TNS and Opinium, both done online.

Look a little closer though, and things are not quite that cut and dried. There are many causes of variation between polls, telephone or online fieldwork is just one of them. There is variation between different online companies and between different phone companies. Last year ComRes’s telephone polls actually produced some of the more Laboury figures, the online Populus polls tended be on the Tory side of average. Below is the average for each company so far this year (given the polls have been pretty static in 2015 I haven’t worried too much about timings of different companies polls, it’s just a straight average).

phoneonlineleads

So all three companies who have been showing a Tory lead are done by phone, all the online polls have been showing an average Labour lead. But note the variation – MORI use the telephone, but they are showing a Labour lead on average. Two online polls (YouGov and Opinium) show barely any Labour lead at all, Survation, TNS and Panelbase average around a 2 point Labour lead. This is because there are plenty of other reasons for variation between pollsters too, different approaches to weighting, turnout, don’t knows and so on – I summarised lots of them here. Just looking at one can sometimes be misleading, for example, ICM and Ashcroft also reallocate don’t knows by past vote, which normally bumps up the Tory position by a point or so, so that will also be a major part of the difference between them and companies showing worse results for the Conservatives (one should also bear in mind that the monthly polling companies have only produced 3 or 4 polls this year – so a single odd poll like ICM’s this month has a large impact on the average).

I’ve no doubt that telephone vs online is one of the reasons for differences though, especially when it comes to UKIP. The graph below has even starker differences. With Labour vs Conservatives the difference between phone and online polls is a matter of a few points. With UKIP there is a vast gulf between the figures from different pollsters…

phoneonlineukip

The companies showing lower UKIP scores are all telephone. The companies showing higher UKIP scores are all online. While there is little difference between the phone company showing the highest UKIP support (Ashcroft) and the online company showing the lowest (YouGov), there is a gulf of 9 points between the highest and lowest ends of the scale. Why there should be such a difference between online and telephone polling of UKIP we cannot tell – some of it may be an interviewer effect (people being more willing to tell a computer screen they are voting for a non-mainstream party than a human interviewer), some of it may be sampling (some online samples getting too many of the sort of people who vote UKIP, or some phone samples getting too few, or both). Until the results are in we won’t really know.


566 Responses to “Phone and online differences”

1 2 3 4 5 12
  1. @ Ashman,

    If this is the response from the Labour manifesto launch, one point movement, not very aspiring. Or have I got something wrong ?

    The idea that either manifesto launch is likely to move VI, for a start.

    But also the idea that the polls are moving Tory-ward. YouGov is static with a slight Labour lead, Ashcroft had an unusually Labour-leaning poll this week, Populus had a Tory-leaning-for-Populus-but-within-normal-variation one, so they’re probably static as well. So we’re left with ICM and TNS.

    TNS has been giving strange results all Parliament. Maybe you thought they were magnificently accurate when they were the only pollster showing 7-point Labour leads when everyone else had the lead around 3%, I dunno. My view is I didn’t rate them then and I don’t rate them now.

    ICM is reputable, but that doesn’t mean they can’t occasionally throw up an outlier, and I suspect that this is what happened this week. Maybe I’m wrong- we’ve got almost continuous polling now, so we’ll soon see if other pollsters have the Tories on 39%. My advice is to watch YouGov for movement, because even if their absolute numbers are off they should show shifts. If they say everything is static, then all these exciting Survation and ICM results are probably noise.

  2. @AW

    Thanks for your answer to my earlier (9.40pm) question.

    I think someone has mentioned this before. Your often very helpful responses to requests for information can often be missed when you insert the information in the original message. Exchanges move on and the UKPR pages pile up and it is not the most natural thing for your guests to go back and re-examine a string of comments already perused a few hours earlier.

    I suspect that some of your answers are missed just because the tide has move on in this way.

    I very much appreciate your finding the time to dig out this information. Frankly I have no idea how you find time to monitor comments and pick up questions like this as well as compiling your informative blogs and maintaining the site plus the little matter of keeping up with your day job. I’m sure your swelling band of followers tells you all you need to know about how perfectly your efforts hit the mark.

  3. @ KeithP,

    Could that possibly be UKIPers wavering?

    It’s probably not wavering so much as sample variation. If you were an older working class man who voted Tory in 2010, the odds are pretty good that you may have switched to Ukip but the pollsters have no idea whether you have or not; everyone in that demographic looks identical to them. So sometimes they poll a lot of Kippers and sometimes they poll a lot of Tory loyalists, and Tory VI appears to be bumping up and down even though the individual voters aren’t changing their minds.

  4. My earlier question must look totally dumb!! Am I right in thinking that YouGov has always been purely an online pollster?

  5. @ Spearmint

    What does that mean in practice? Does it mean the true UKIP picture is the average?

  6. @ Alan, CMJ, unicorn, charles

    All were very interesting …

    Still, what is the question (I read your point Alan, and I got it)?

    Here’s a problem-wise similar, content-wise unrelated problem (and hopefully it shows what my problem was – though Unicorn’s summary was very convincing):

    http://www.nature.com/neuro/journal/v14/n9/full/nn.2886.html

  7. MartinW

    I have certainly thought it has always been online.

  8. @ MartinW,

    Well, I hesitate to say “true picture” because obviously some pollsters are right about Ukip and some are totally wrong, and we can’t know which is which until the election results come in.

    But if an individual pollster has Ukip on 12% one week and 16% the next and then back on 12%, I think you’d be pretty safe in assuming Ukip is really on 14% and not bouncing up and down like a yo-yo.

    TNS is a classic example of this: 16% then 19% and now 14%. Was Ukip really on 19% last week? I doubt it. There is compelling evidence for a long-term Ukip decline, but they haven’t jumped 3 points and then plummeted 5 in the space of a fortnight, even taking into account the publicity boost from the debate.

  9. Ta, Spearmint.

    I meant “true picture” for a particular pollster (if that doesn’t sound too silly!)

  10. For the record, I asked whether there was good independent evidence that telephone polls are more accurate than others and Anthony provided a link to 2011 paper by pickup, Matthews, Jennings, Ford and Fisher.

    This looked specifically at the serious overestimation of LibDem support just before the 2010 election, and in the final sentence of the Abstract the authors concluded:

    We find robust evidence of an over?estimation in Liberal Democrat support, but do not find evidence to support the hypothesis that the polls erred due to a late swing away from the party, nor that any of the methodological choices made by pollsters were significantly associated with this over?estimation.

    As ‘methodological choices’ included factors such as use of telephone vs online methods, I think this suggests that the jury is probably still out wrt to the verdict of whether one method is more accurate than the other.

    So, for the Tories telephone polls may place things in a slightly more favourable light. But we don’t really know whether this provides false optimism or a better guide to the outcome of the election.

  11. @ Martin W,

    I meant “true picture” for a particular pollster (if that doesn’t sound too silly!)

    Nope, makes perfect sense to me!

  12. Using black and red (Conservatives ahead, Labour ahead) at this point seems to be good predictions at this point – the polls seem to be a “fair coin” – well, with built in, not too important biases (that seem to even out).

    I’m more and more intrigued by the question, what is the point when details within these polls become interesting (percentages, crossovers, etc).

    However, if the rouge et noir were true, then important changes should be picked up earlier.

    So, with the increase of the number of polls: in one polling company movement to party A is +1. Summed up to all polling companies. Having stable or growing sums (over, let’s say 2 days) suggests movement in the polls that warrant a statistical analysis with the assumption of: movement happen and we want to know: the magnitude (and not if the movement happened as it is in the assumptions).

  13. @Laszlo

    Misinterpreting interactions is, indeed, endemic in academic research.

    Here on UKPR comments fall into this trap most often when they claim to have seen an improvement in a party’s margin over another’s over a period of time.

    An example, is a recent set of claims that over the last week the Tories have increased their margin relative to Labour. This basically amounts to claiming that with VI as the dependent variable there is a significant statistical interaction between two factors: Party and Time Period of polling. Most of the time the evidence isn’t there to back up the statements.

    But people will mostly see precisely what they want to see.

  14. I have managed to write a largely incomprehensible post just above.Apologies.My excuses are time and the still working cogwheels of my mind to grasp and properly express my thoughts.

    So apologies if I wasted anyone’s time.

  15. Unicorn + others interested in MOE of a lead.

    OK, reran it with 1000000 trials from a true multinomial distribution. (each with sample size 1000). Pretty much get the same result, 95% CI of 5.19, which matches the analytical value. Checked correlation and sd and they both fit the analytical values.

    I ran it for various VIs from (5%, 5%) to (40%, 5%) and always got a normally distributed gap. It’s nice to know that for anything we are interested it, the gap with be distributed normally.

    Overal between 60% sum of parties and 80% the MOE rose from 4.8% to 5.54% for all intents and purposes linearly.

    A rule of thumb for the CI of a gap would be 4.8 + 0.037% for every % the sum of the parties is greater than 60, I was going to tabulate them but with such a linear result (R^2 = 0.9996!) there’s little point.

  16. @ Alan

    So, after the collapse of the LibDem vote, UNS is actually a pretty good indicator for the vast majority of .seats.

    The question mark is with seats where surviving pockets of LibDem votes and/or UKIP votes are important as we can’t control for the false negative.

    DoI understand your conclusion from the trials to the data co4rectly?

  17. @Spearmint

    I agree, over the same period trends on YouGov are a far better indicator of underlying movement than anything that ICM can come up with in their occasional polls.

    In the fortnight between each ICM poll of 1,000, YouGov will get responses from around 20,000 people in 10 polls of 2,000.

    Furthermore, ICM pick up a higher proportion of DVs/DKs than YouGov (not necessarily bad in itself but it further curtails the sample size) and then go further and discount by 50% others who didn’t vote in the GE (again shrinking the sample). All that means that the effective ICM response rate is particularly low. So the ratio of effective total sample size between YouGov and ICM is much more than 20 to 1 over a fortnight. Ironically, I suspect that ICM polls would be even more volatile but for the controversial reallocation of don’t knows according to the assumption that these will revert to set and therefore stable 2010 voting patterns.

    That means that in the period that ICM take one polling shot with a scattergun musket, YouGov are taking 10 polling shots with a rifle. The trend in the rifle shots is the one to watch. That will be the case even if ICM had the better methodology such that there musket was aiming in the correct direction with random inaccurancy and YouGov’s rifle had a persistent slight bias to the left or right.

    One other thing. For the purpose of argument, even if ICM’s methodology is assumed to be better, if they are to get the more accurate result in their final poll they’ll be keeping their fingers crossed that it isn’t another one of their many outliers.

  18. LASZLO

    Agree, the point raised in that paper is the very same we are essentially discussing.

    Very much a case of good science spoiled by bad statistical analysis. A little knowledge is a dangerous thing, and running around wielding (p < 0.05) without sufficient knowledge can lead to dubious levels of certainty about "facts"

    For a Lab + Con of 70, the 95% CI that one party is in the lead is 5.2% for a sample size of 1000 and 3.7% for 2000. It's very hard to tell anything on one poll alone (hence the strong caveats) and frankly when looking for "aha! a lead" in the middle of nothing happening, looking at 50 polls and finding 1 or 2 evidences of "a definite lead (p < 0.05)" is to be expected.

    Clearly we need suitable ways to aggregate the polls if we are to reliably identify a small lead.

  19. During the Heywood and Middleton by election I worked on a telephone poll on behalf of UKIP. We identified ourselves, from the start, as representing UKIP. The reaction was amazing and very close to the actual result.

    Approach has a deal to do with the inaccuracy of telephone polls. Just as with face to face interviews people are reluctant to state their view when UKIP is continually defamed by the press, media and unscrupulous politicians.

    In the end I anticipate even the online polls will be significantly understating UKIP support. In the secrecy of the polling booth the pencil mark cannot be intercepted by any electronic or other means.

    The massaging down of UKIP support by the polling organisations is deliberate. They cannot produce any evidence to support their reasoning. They systematically reduce the UKIP share by a quarter and increase the Liberal share by a third. The increase all parties except UKIP… Funny that!

  20. @Alan, @Laszlo

    Nate Silver is scathing about using confidence intervals as a means of trying to identify a departure from the norm and recommends instead using a Bayesian approach. Both Bayesian and Fisher based approaches are dependent on assumptions, the fact that they are transparent in the former but hidden in the latter is actually a further merit of the Bayesian approach.

  21. LASZLO

    This is nothing to do with how the LD vote is distributed in various seats.

    Essentially what I am simulating is a fixed population (so fixing the true population VI) and taking many samples of size 100 (typically 1000000 samples) from that population. From there I look at the values of (P1 – P2), the lead, and studied how they are distributed.

    The mean of the lead comes out as P1-P2, as expected.

    To a first order of magnitude (assuming the gap between the parties is relatively small) the sd of the lead is sqrt((P1+P2)/n).

    I was also checking that the lead was distributed as a normal distribution as I see warnings in the literature that at low p the distributions can become asymmetric. These warnings were largely unfounded as I couldn’t generate any asymmetry for reasonable values of P1, P2.

    This all stemmed from the discussion about a 6 point lead and how likely it was down to pure sampling error that it occurred, given a proposed reality that the parties were equal. The statistics show it is about a 1% chance and given the amount of polls we look at, I don’t think even that is significant alone. It might well be the one poll in a hundred that shows a 6% lead when the truth is a dead heat.

  22. UKIP

    There is a big difference between online and telephone polling of UKIP due to people being reluctant to tell a stranger on the telephone that they intend voting UKIP because of the racist tag the Establishment have tarred UKIP with.

    UKIP is probably well over 20% in England (excluding London) and Wales. As reported here, a poll a few days ago put UKIP on 24% in Kent (17 seats).

    An analysis of the 9 April TNS poll, which put UKIP on 19%, concluded UKIP could win 42 seats:

    http://ukipdaily.com/tns-bmrb-poll-brings-hope/

  23. Simon

    Something to chew on from today, from TNS, just so you aren’t falling into the trap of cherry picking polls.

    TNS GB poll

    LAB 32% (-1),
    CON 34% (+4),
    LIB DEM 9% (+1),
    UKIP 14% (-5),
    GREEN 5% (+1),
    OTHER 5% (-2)

    I can’t comment on the 42 seats as details of the analysis were non existent.

  24. Apparently YouGov’s next poll (Wed) has a 2% lead for Labour. So whatever these other polls seem to show – it’s really just a blip.

  25. UKIP’s higher rating in Internet polls is purely down to the phenomenon of “internet trolling” whereby people offer outrageous opinions purely to gauge a reaction, even though the opinions they offer are not genuine.

    As such, the telephone polls offer a more accurate representation of the truth.

  26. (By the way, my post was a joke in response to Simon’s utterly ludicrous comment above)

  27. UNICORN
    “…… with VI as the dependent variable there is a significant statistical interaction between two factors: Party and Time Period of polling. Most of the time the evidence isn’t there to back up the statements.”

    In current research I am examining the correlation of eco-agriculturalamd market factors by location and the incidence of off-farm employment among ethnic minority groups in southern Laos impacted by hydropower dams. I explain first that in my methodology I wish to strengthen the database and hypotheses, since I’ld like readers, if any, to understand that I don’t, scientifically seek to present my conclusions as “fact”. My purpose, I explain, is to provide a better basis of information for the restoring livelihoods in community relocation, but that doesn’t stop the approach being scientific. The statements I make are heuristic, intended to push the work of enquiry along, rather than to be persuasive, except ins so far as others can follow my workings. Ditto I would have thought, your and others’ admirable explorations of VI and its correlations.

  28. Good early morning everyone.

    No breakthrough for the Tories to a parliamentary majority, yet, although the LD figure on the YG Poll is still high; maybe face to face interviews rather than internet or phone polls would reveal the true figure.

    Beautiful New Dawn here.

  29. Mike Smithson [email protected] · 5m5 minutes ago
    Twice as many, 17%, of 2010 LD voters telling YouGov that they’re still undecided compared with LAB & CON ones.
    A lot votes there

  30. ALEC

    @” I can, on behalf of all at UKPR, say (very politely) you were wrong.”

    Something , in respect of which, he has subsequently become a minor celebrity . :-)

  31. “Exchanges move on and the UKPR pages pile up and it is not the most natural thing for your guests to go back and re-examine a string of comments already perused a few hours earlier.”

    ————

    I go back and re-examine all the time, and would urge others to do the same, for the advantages are legion. For a start, one hates to miss anything peeps may have posted. Especially one’s own posts. Also, stuff may get modded some time later. Bit like pollsters changing methologies…

    Sometimes one has to try and reconstruct what really happened in a thread from the bomb craters and tell-tale fragments left over after some enthusiastic modding. I do enjoy this challenge: it’s like trying to crack Linear B, especially due to inconsistencies in the moderation methodologies, and frequent revisions. (Cracking the methodology is important, because it can give clues as to what is currently verbotten today, and may become verbotten tomorrow. Like, the other day, after making a few posts concerning popcorn, I discovered they had all been vaped, bar one. I am still trying to figure out what was special about the orphaned popcorn post; it is unusual for a moderation opportunity to be missed, so it likely has some deep significance).

    It’s also useful for when peeps try and change the story. Which only happens several times a day…

  32. And of course many of the undecided 2010 voters are women -but will they become decided ?

    Labour has a lead amongst women on you gov most days ,tories tend to lead amongst men.

    Whilst i wouldnt normally put much store by tns ,yesterdays had some VI figures by age and gender.

    There were three groups on 40per cent and above -women over 55(tory)
    men 25-34 (labour).And I will return with the other.

  33. @Alan

    Thanks for carrying out (and sharing) those late-night calculations of yours. It is good to nail what MoEs we should be using for assessing margins between VIs. Even here on UKPR many have been rather cavalier about assumptions being made, apparently not having been aware that MoEs for margins are larger than those for individual VIs.

    Clearly we need suitable ways to aggregate the polls if we are to reliably identify a small lead.

    Have you had a chance to look at the details of EWMA Chart analysis as introduced here recently by @Peter Ould?

    This is based on Gaussian distributional assumptions (albeit apparently being fairly robust) and so your checks last night seem to make it suitable for use for detecting small leads.

    Given that we now have a good estimate of the SD of (say) Con/Lab margin distributions, we have all the figures we need to plug into the MoE formula.

    The normal usage of the charts is to detect any kind of change over time. But by setting the upper and lower bounds around zero, it seems to me that the method provides a good way of aggregating polls and hence a sensitive way of detecting the appearance of a lead one way or the other.

  34. Age and gender 40percent and above on yesterdays tns

    Men 18-34 (labour)
    women 35-54 (labour)
    Women 55 plus (tory)

  35. @The Truth
    “By the way, my post was a joke in response to Simon’s utterly ludicrous comment above).”

    It is often wise at times like these to fall back on the sage advice of the great philosopher, artist and philanthropist Michael Jackson in his famed opus Billie Jean:

    “Be careful what you do/Because a lie becomes The Truth.”

  36. @UNICORN

    “Thanks for carrying out (and sharing) those late-night calculations of yours. It is good to nail what MoEs we should be using for assessing margins between VIs. Even here on UKPR many have been rather cavalier about assumptions being made, apparently not having been aware that MoEs for margins are larger than those for individual VIs.”

    ———–

    Thing is, peeps don’t have much choice to be a bit cavalier, else one winds up just shrugging one’s shoulders and saying “dunno”.

    The vagueries of polling are such that one is not exactly going to be attaining five sigma levels of certainty. You can almost always stick a caveat in there. Take a single poll, and it could be an outlier. Take a sequence and average it, and you may be averaging out a genuine brief change etc.

    So peeps try and use a mix of things, build in circumstantial evidence etc., which makes the system more nebulous, and drives the mathematicians crazy.

  37. @Carfrew

    Were you a fan of Stavros?

  38. AW
    A week or so ago you mentioned that the change in YG methodology to include likelihood to vote had not made any difference to VI at the point the change was made.

    Is this still the case?

    And can you please explain why it didn’t have (and perhaps still hasn’t had) the expected effect of reducing Lab VI relative to Con VI?
    Thanks

    [Pretty much – it’s making a minute difference, a tiny fraction of a percentage point, so it’s rarely making any difference to the headline figures. Reason is very straightforward – at the last election Tory voters said they were more likely to vote than Labour voters, this time the two sets of voters rate their likelihood to vote as about the same – AW]

  39. @RAF

    You had me wondering there for a minute. I was like “Stavros? Never been to the place.”

    Innit.

  40. Very interesting article on EM by Raphael Behr today.
    Chris Lane,
    Good morning from an equally sunny South Wales.Interestingly there is a very
    beautiful rose called The New Dawn,symbolic perhaps?

  41. Re Ashcroft marginals yesterday.

    In all Libdem VI has been decimated.

    But ukip VI remains squeezable apart from finchley.

    Cleethorpes ukip 15 (tory 38-lab 36)

    Crewe 13 ((38-41)

    Dover 20 (40-32)

    Dudley S 18(39-35)

    Finchley 6 (41-43)

    Harlow 16 (44-34)

    Milton K South 10 (37-39)

    Rossendale 8 (42-42)

    South Ribble 12 (39-39)

    Farage to say manifesto is fully funded today.
    Clegg we are the party of coalition.

  42. @Carfrew
    “@RAF
    You had me wondering there for a minute. I was like “Stavros? Never been to the place.”
    Innit.”

    Lol. Your use of the colloquialism “peeps” in several posts above caused me to consider the possibility.

  43. @RAF

    That’s why I said “innit”.

    As in “innit peeps”

  44. Thanks for the graphs, Anthony. now I know which polling companies I can view as trustworthy and reliable and which I can write off as vile propaganda.

  45. Farage to say fully costed

  46. Oh Dear.

    It looks like ”fully costed” has joined the growing lexicon of fatuous phrases used ad nauseam through this election campaign.

    The more often it is used the more clear it is that any party using the phrase hasn’t a clue how to fund its commitments.

    Perhaps a pollster will do some scientific research to establish the public’s least favourite / most annoying buzz phrase of the election.

    ”Fully costed” can join ”our long term economic plan” and ”saving our NHS” on the list, along side many others.

  47. ANN IN WALES.
    That was a brilliant piece of writing on Ed Miliband. Thank you

  48. How can one say fully costed when the key issue for UKIP is withdrawal from the EU – the place where half our exports go. What makes UKIP believe we will have free access to the EU after withdrawal? I see a strong likelihood of tariff barriers (EYU- you dont want to pay to be members of our club, but you want all the benefits? Get lost…) and the consequent destruction of our export trade – and many manufacturers leaving the UK and going to the EU. Result? destruction of the Uk economy… That needs to be costed into the UKIP manifesto…

  49. I am still having a hard time discussing the notion of “squeeze” when it comes to UKIP and Green in comparison to 2010, as in the latest Lord A polls there continues to be a swing away from LD in particular, but also from both Conservative and Labour in a seat like Dover, for example:

    UKIP +13.5%
    Green +4%
    Labour -2%
    Conservative -4%
    LD -13%

    And the question has to be asked, in earlier times would Labour have taken this seat from Conservative without UKIP presence. Thus UKIP may not be winning many seats, but how are they influencing the outcome of this election in every seat?

    In the North Labour is definitely making votes gains, but so is UKIP in particular, noting I am only including the swing to UKIP after factoring in the previous vote for BNP, NF, English Democrats and various socially conservative Christian parties:

    Crewe and Nantwich

    UKIP +8.2%
    Labour +7%
    Other -1%
    Conservative -8%
    LD -11%

    Another 1% is up for grabs as there is no Green candidate in this seat and again could Conservative have held this seat without a UKIP presence?

    Cleethorpes

    UKIP +8%
    Labour +3%
    Green +3%
    Conservative -4%
    LD -12%

    Is the presence of UKIP and Green the difference between Labour taking the seat or does their presence make this seat more of a marginal?

    South Ribble

    UKIP +6.3%
    Labour +4.3%
    LD -6.1%
    Conservative -6.5%

    Again no Green candidate here, but has the presence of UKIP made this seat more or less likely to become a Labour seat?

    Rossendale and Darwen

    Labour +9.8%
    Green +2%
    UKIP +.9%
    Conservative +.2%
    LD -14.1%

    This is a seat in the North, in Lord A’s sample, where UKIP growth is minimal and one is left wondering how many more like that exist across the country?

    I am really disappointed in Lord A for choosing Dudley South, rather than re-polling Dudley North:

    UKIP +10%
    Green +4%
    Labour +2%
    Conservative -4%
    LD -13%

    Again it surprises me that half the decline in LD support, on the surface, appears to be swinging to UKIP – unless of course it is going to Conservative and Labour and lessening the bleed off to UKIP from those parties. Again one has to ask if this seat would have gone to Labour without the presence of UKIP and Green.

    Next there is Harlow and Milton Keynes in the southeast, in which Lord A has again avoided a UKIP heavy seat in his mix:

    Harlow

    UKIP +8.4%
    Green +2%
    Labour nc
    Conservative -1%
    LD – 11%

    Milton Keynes South

    Labour +7%
    UKIP +4%
    Green +2.6%
    Conservative -5%
    LD -10%

    Again, in terms of Harlow, one is left wondering if Labour would have taken this seat without the presence of UKIP and whether the swing from LD is to Conservative to prevent Labour winning the seat.

    It seems to me that between now and E-Day that this seat could still go either way, again leaving me asking what would have happened without the presence of either a Green or UKIP candidate.

    Next the lone London seat in the mix, Finchley and Golders Green:

    Labour +9%
    UKIP +4.3%
    Green +3.4%
    Conservative -5%
    LD -11%

    This seat appears to be operating like a classical Conservative-Labour marginal, in which one might ask does the presence of UKIP and Green make a difference and the answer I come up with is no because of the collapse in the Liberal vote appears to mitigate their presence.

    Lastly North East Somerset where there are number of factors to discuss:

    Green +6.7%
    UKIP +6.6%
    Conservative +2.7%
    Labour -3.7%
    LD -11.3%

    First off I observe that LD have retained 40.4% of their vote, which is a good sign if they want to retain their 15 seats in the southwest. Next there has clearly been a swing from LD to Conservative.

    Next, and I will be probably be criticized again here, I looked at the 2014 Bath and North East Somerset European election results:

    Conservative 27.2%
    UKIP 24.8%
    Green 16.3%
    Labour 15%
    LD 14%

    Even though the turnouts and the voting system is quite different I find it intriguing that Green is still in contention for a third place finish in a Conservative-Labour marginal and that the UKIP vote has so spectaculary declined in the Southwest. This is not the first time I have seen this as several pollsters are showing this pattern in their cross-breaks as well.

    The latest Lord A poll ,for example, has the split in Wales and the Southwest as:

    Green 11%
    UKIP 11%
    LD 9%

    The rogue ICM poll for the South shows:

    Green 8%
    UKIP 8%
    LD 7%

    The latest Wales and Southwest Populos:

    Green 9%
    UKIP 9%
    LD 6%

    Only the latest YouGov has the standard pattern in Ros as:

    UKIP 17%
    LD 10%
    Green 6%

    So I’ll be watching the Southwest as the vote results come in, not just for the GE but also in terms of what happens in the local elections as wel.

    And of course I wish I was a fly on the wall in terms of the various campaigns in Bristol-West. If the LD vote holds up then they could hold it, but then again given Lord A polling in LD-Labour marginals one would think Labour will take it, but yet again if Green are stronger in the Southwest maybe this is a seat they will take.

  50. @Roger Mexico at 11.18 on accuracy and precision.
    Bravo! also @Spearmint on UKIP (and others) wobbling up and down due to sampling not mind changing.
    I’m still puzzled as to why some VIs do not vary as much as they should if the polls are subject to random errors.

1 2 3 4 5 12