Lord Ashcroft has a new batch of constituency polls out, this time looking at ten Conservative held seats with majorities of around 9% to 12% – seats that would need a Con>Lab swing of around about 5% to fall to Labour. Full details are here.

National opinion polls are currently showing a swing of around 4 percent from Conservative to Labour, so on a pure uniform swing you’d expect these seats to stay Conservative. The average swing in these polls was 3.5%, so very much in line with that. In practice though there is variation between seats, so Ashcorft did find three seats in the sample where Labour were ahead – Crewe and Nantwich, Finchley and Golders Green and Milton Keynes South. This sort of variation is inevitable, and in terms of the overall impact on seats will be counteracted by seats that Labour should be winning on a uniform swing, but where Ashcroft has found the Conservatives ahead, such as Kingswood and Blackpool North.

It is difficult to analyse the variation between seats too much – in some cases they will be genuine, and we can come up with plausible reasons to explain them (are Labour getting a bigger swing in London, for example, or is Crewe and Nantwich seeing a gradual unwind from the by-election?), but remember each constituency poll is just one poll with the normal margin of error. The fact that a poll in one seat is showing a swing that’s three percent bigger than a poll in another seat may just be normal sample variation between polls, and nothing to do with the situation on the ground.


330 Responses to “Latest Lord Ashcroft constituency polls”

1 5 6 7
  1. Apologies…continuing from above. I would spin the line. The high Con /low UKIP pokss are the ones getting it right.
    We are doing even better in the mrginals. Finally Id play the probably.rogue big lead as credible.
    I’m not saying its true but ifout wanted to rally the troops….

  2. @exileinyorks

    Have you found the knackered old keyboard I left up in Yorkshire when we moved?

  3. exile don’t forget mitt Romney and his supporters including conservative commentators here like janet daley were genuinely shocked when they lost in the 2012 American election. if the American republican party ,presumably a much wealthier political party than any in Britain,can get it spectacularly wrong then any party here who thinks they’re confident of the outcome is seriously deluded.

  4. @Charles

    Your formula gives an estimated MoE that is very similar to the figure that @Alan generated earlier using his bootstrapping exercise.

    For the ICM poll that started this discussion the Tories were on 39% – so P1 = .39. For Labour P2 = .33.

    This gives an MoE estimate of 0.052 or 5.2%. For polls with an N of about 2000 this would be reduced to about 3.7%.

    All these calculations emphasise that MoEs are noticeably larger for margins than they are for the VIs themselves.

  5. @ Paul

    What have you seen that makes you think the Tories are being overconfident? The vast majority on UKPR think Labour will win most seats.

  6. Bantams.
    Your keyboard would probably help. Better than using a mobile on a moving train:-)

  7. @ Unicorn

    “For the ICM poll that started this discussion the Tories were on 39% – so P1 = .39. For labor .33”

    Definitely NOT. This the epistemological error I mentioned. p1=.39, thus p2 is .61 (not p1).

    Just look at the consequences.

  8. Regarding MOE between groups in one poll:

    https://abcnews.go.com/images/PollingUnit/MOEFranklin.pdf

  9. @ CMJ

    Thanks. Yes.

  10. Paul

    As I said. I don’t believe that is necessarily true, but it is just about plausible within the data for a bit of wishful thinking amongst those who would like it to be true.

    FWIW it is not the result I am hoping for

  11. @Lazlo

    According to page 3 then you can more that two Candidates, and still use the formula for only two.

    It looks pukka to me :-)

  12. exile I agree. bantam I have no inside knowledge myself I was just responding to exile who said the tories were looking surprisingly upbeat.

  13. I presume that there will be questions in polls shortly on what folk thought of manifesto promises. The very learned discussions such as on here will not be read by most voters, so we will see if voters are as gullible as the manifesto writers take them for.

  14. @Everyone saying the tories are looking upbeat. Coukd it possibly be that they are acting?

    Mad I know, but if they were to come out and say “we’re really worried, things don’t seem to be going our way” it would probably exacerbate the problem.

    Just a thought!

  15. @ Paul

    I think today’s been their first upbeat day, they’ve been pretty negative up to press.

  16. HA RTB

    Dog whistle policy or dogs breakfast?
    ;-)

  17. New thread

  18. Lazslo

    The formula using sqrt(p*(1-p)) is for calculating the variance of a single poll.

    If you want the variance of the margin between two polls you can use the formula provided earlier, the error was due to multiplying probabilities by 100 into percentages which altered one term by a factor of 100 relative to the other. The (p1-p2)^2 term gets negligible for the typical values of Con and Lab we see.

    I reproduced the 5.2% MOE through simulation using the dependent probabilities.

  19. Barbazenzero

    The Con manifesto may just have handed him a lifeline as all the others will be looking to put the boot in on it too.

    Unlikely, but its just possible that each will put over his/her main selling points and then attack the Cons collegiately
    ———————————————————————-

    They could, they can all put a boot in it. Spend the entire evening doing that. Highly entertaining. Is it going to make any difference, considering that, it has not made an iota up to now.

    Miliband Vs Sturgeon and Farage.

    Farage would want to leave that debate with one thing, he has to carry out a character assassination of a pro European, he has to take some blood, Miliband, to make his mark on this election, he does not do that, the Conservatives are going to make a play of that. He is a failed Anti European.

  20. @CMJ

    Thanks for the link to the Franklin paper. Very useful. Same formula on p. 3 as the one posted earlier by @Charles.

    @Laszlo

    Definitely NOT. This the epistemological error I mentioned. p1=.39, thus p2 is .61 (not p1).

    According to p. 3 of the paper p2 is NOT just the complement of p1. Rather, the two proportions are intended to represent the degree of support for two different ‘candidates’. (In our own case, the candidates just happen to be political parties.)

  21. @ CMJ

    The whole thing comes back to: what are you measuring? (this is the epistemological problem I alluded to)

    If you are interested in the performance of parties (relatively independently from each other), then you measure the MoE with the party preference against the party non-preference.

    If your are interested in the changing distribution (but then you have to be very clear that this is what you are doing, in spite of the appearance), you can pile up the party preferences.

    The conclusions are different. The distribution is a zero-sum game of distributing preferences. Preference of the party (and it’s MoE is a positive choice).

  22. New thread polling methods

  23. I recall some time (years ago?) such a discussion took place as now.

    I distinctly remember Charles and I getting involved, and I think Rob S may have provided that link.

    What I ate for yesterday’s breakfast? Clean forgotten…..!

  24. @ Unicorn (and CMJ)

    Please see above.

    As long as you say that the distribution of the votes changed as … It’s fine.

    If you want to say that the preference for party A, the MoE is preference for party A against the rest.

    I’m quite keen on this point, because I think that error rates are misunderstood.

    If you created a graphical representation: let’s say Conservatives with their headline VI around it the MoE in a shade (calculated as Con vs rest), and the same of the other parties, you would see much more in terms of “real” and “imaginary” changes.

    I used the data for the last month, and I can see it on the numbers, but it’s boring, a graph would be nicer but I can’t do it.

  25. Laszlo

    What people are talking about is the SD of (P1-P2) where P1 and P2 are the VI for different parties. Nothing to do with preference vs non preference, just measure the value of P1-P2 and treat as a variable in it’s own right.

  26. ASHMAN
    They could, they can all put a boot in it. Spend the entire evening doing that. Highly entertaining. Is it going to make any difference, considering that, it has not made an iota up to now.

    If the last “it” in your post is “the Con manifesto” then right now you’re correct, but I’d be surprised if it wasn’t used as a sales tool by all of the other parties, including the absent LDs.

    If the “it” was something else please elucidate.

  27. Tomorrows debate:

    It is not in the interests of the SNP or PC to make trouble for Ed. Nicola & Leanne will be focussing on their positive message. Natalie might want to make trouble for Ed but the Greens are running a relentlessly positive campaign.

    Nigel will be kicking the Tories.

    So loads of criticism of the Coalition & the Tories interspersed with each parties positive key messages. I imagine very little arguing amongst the five. Think The Agenda from yesterday.
    It will be win-win all round IMO

  28. 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

  29. COUPER2802

    I can see why cameron didn’t want to do any TV debates, but letting them go ahead and turning up is quite different, he will take much criticism and not be able to respond, how can this be a positive? he will be left to say the next day that all the others are a bit barking so vote for me. and the response will be if thats the case why didn’t you bother to turn up to speak.

  30. Hi all,

    I am enjoying this election and especially the discussions on here.

    I was reading the Daily Telegraph today and this article appeared.

    http://www.telegraph.co.uk/news/newstopics/eureferendum/11537057/Juncker-blocks-EU-treaty-negotiations-until-after-2019.html

    Assuming Its true, what effect do you think it will have on the Tory – UKIP voters as it appears to undermines DC position.

    Could this be a game changer ?

1 5 6 7