Lord Ashcroft has published a new batch of constituency polling. I hesitate to call it marginals polling, since we’ve moving up into some less marginal territory with today’s polls. Ashcroft has polled four different groups of seats in this set (all the tabs are here.)

The first is the next cohort of Lib Dem -v- Conservative marginals, this group are those seats with a Lib Dem majority of between 9% and 15% over the Conservatives, so we are no longer looking at ultra-marginals. The average swing from the Liberal Democrats to Conservatives in these seats is 2%, nowhere near enough to win seats like these. However, as we’ve seen in previous Lord Ashcroft polls of Lib Dem marginals there is an awful lot of variation between individual constituencies – some seats (Carshalton & Wallington and Thornbury & Yate) are actually showing swings from Con to LD. At the other end of the scale two seats are showing large enough swings for the Conservatives to win the seat (North Devon and Portsmouth South, which has a chunky 9 point swing from LD to Con, presumably at least partially connected to the scandal around Mike Hancock).

The second group of seats consists of two more Lib Dem seats with Labour in second place. Lord Ashcroft’s previous polling in LD v Lab seats essentially showed a complete Lib Dem collapse, raising the possibility of an almost complete wipeout for Liberal Democrat MPs where Labour was the main opponent. One of the seats here – Burnley – follows that pattern, with a ten point swing from LD to Lab. The other, Birmingham Yardley, represented by John Hemming, bucks the trend. There is still a 2.5% swing from LD to Lab, but it is smaller than we’ve seen in other LD -v- Lab seats and would be small enough for Hemming to hold on.

The third group of seats is two unusual seats – the close three-way marginal of Watford, and Wyre Forest, an Independent seat between 2001 and 2010. Neither of these really fit into any broader category, but looking at them as individual seats Watford shows little relative movement for the three main parties – all are down a little, UKIP are up a lot but still in fourth place, meaning the Conservatives retain a narrow lead. Wyre Forest was held by Dr Richard Taylor between 2001 and 2010. He’ll be standing again come the next general election for the National Health Action party, but I think under the same Kidderminster Health Concern label that he won on in 2001 and 2005. Ashcroft’s poll currently has the Conservatives holding the seat on 32% with UKIP in second on 27%, Labour 16%, Lib Dem 7%, Green 5%, Other 13%. The others aren’t identified in the poll, but is presumably largely Dr Taylor’s supporters.

Finally Ashcroft polled three of the four seats that will be contested by the main party leaders come the election – Sheffield Hallam, Doncaster North and Thanet South (presumably he didn’t do Witney because he thought it would be too boring… it would seem there comes a point when even Lord Ashcroft saves his money!). Party leaders normally do pretty well in their own seats. It is extremely rare for them to lose their own constituency and they very often outperform their party nationally. Such is the collapse of Liberal Democrat support however people have seriously raised the possiblity of Clegg losing his own seat – Ashcroft’s poll has it very close. Clegg is on 31%, Labour on 28%, just three behind (and this is on the question prompting people to think about their own constituency, the standard voting intention question had Labour a point ahead). Ed Miliband’s Doncaster North seat is traditionally a very safe Labour seat that should pose no concerns for him, but there was some speculation about how well UKIP might do. The BNP have held their deposit there at the last two elections and their was some significant support for the English Democrats too, with the far-right parties now collapsing and UKIP hoovering up that right-wing protest vote it looked as if there could be some potential. In fact Ashcroft’s poll did find UKIP in second place in Doncaster North, but 12 points behind Ed Miliband. Finally Thanet South, the seat where Nigel Farage plans to stand at the general election. Current figures there are CON 34%, LAB 26%, LDEM 7%, UKIP 29% – so UKIP in a strong second place, but not currently quite enough to send Farage to Westminster.

529 Responses to “Ashcroft polls of the Lib Dem battleground and leaders’ seats”

1 6 7 8 9 10 11
  1. Richard @ Bill Patrick


    I’ve been conducting the same exercise!

    Actually, the accuracy of the statement that Bill was challenging depends on the definition of age.

    If we define the geriatrics as those aged 40+ and exclude them from the calculation, then Yes won by 53% to 47%.

  2. @Oldnat

    I assume from your question to Amber, that your hospital stay is imminent. Hope the operation goes extraordinarily well and all are astonished by the swiftness of your recovery.

    All the best :)

  3. Syzygy

    Thanks. I go in on Monday for operation on Tuesday. I’m hoping that there will be wifi once I’m back on the ward after Intensive Care – otherwise I’ll be bored out of my mind. :-)

  4. OLDNAT…………BLimey, it sounds as though you’re in for a bit of a stressful time, for what it’s worth, UKPR wouldn’t be the same without you, so do us all a favour, get well soon. :-)

  5. Ken

    You’re about to find out how good it can be without me for a week or so. :-)

  6. OldNat, Richard

    Oi! That was my calculation! :D

    Apologies to Bill for giving the wrong age range (I gave figures for 65+ not 60+ as I said – how dare YouGov not use their usual age-bands), but it is pretty clear that age was a big factor in how people voted. Not that it means that independence is inevitable: views may change as people age; more non-Scots may move to Scotland; other changes may make going alone look less favourable. But the pattern was there.

    Midlothian don’t seem to have the transfer report up yet, but David Boothroyd seems to have got rough figures which seem to indicate that Labour just got in with Conservative transfers:


  7. @oldnat

    It mentions wifi in this story


    But also says it is expensive!

    Best of luck with your op and we expect to hear you talking about Scotland again soon!

  8. It mentions that they use a hospedia system in that story, their site is here


    It tells you how to pay, but not what the prices are

  9. And to use wifi you need to download an app first


  10. @old nat – very best of luck, I’m looking forward to reading your wifi enabled posts.

  11. Richard


    Web access only on their most expensive package!

    Bloody PFI !!!!!!

    In properly funded hospitals like Ayr, the wifi is free!

    I suspect Charles (thanks) is going to have to wait till I get home.

    Still, I’ll no doubt have a spell on morphine, so actually won’t give a damn! :-)

  12. OLD NAT
    Good luck with the op. Get well and home soon – they like to shove you out as soon as they can, so back to some good, bad-tempered posting in no time.

  13. It is wonderful to receive a new batch of polling data and I wholeheartedly endorse those who have expressed appreciation of Lord Ashcroft’s decision to share this valuable material.

    Summary of this post for the busy reader: In the new Ashcroft data the Proportional Swing Assumption fares much better than Uniform Swing.

    For the more leisured, details follow here:

    A few days ago I posted a piece in which I argued that given public access to these marginal polls there is now a prospect that we can improve on existing methods of predicting future polling VI, actual election VIs and, indeed, future seat counts – all of which are presumably a matter of interest to UKPR folk.

    In essence what I was suggesting was a method for implementing churn analyses. The basic idea was to use regression analysis on Ashcroft constituency data to identify evidence-based patterns in voting shift.

    The new batch of constituency polling data offers an ideal opportunity to test the effectiveness of this move. The evidence suggests that it has worked rather well – with some (important) caveats mentioned below.

    Much of the most recent batch of polls focused on how the LibDems are doing and how many seats they are likely to retain after May. Given this, I’ll concentrate here on projections of LD VI. (In all cases, these analyses are based on the use of *corrected* VI taken from Question 2 (Table 2) in the Ashcroft tabulations. These numbers are converted to %VI by expressing them as a proportion of the total number of votes for all parties.)

    If fully developed, the Regression-based (RB) approach would basically set out to perform the same task as the UNS model and other similar models: namely to predict General Election seat tallies for each of the parties. It is essentially a modest elaboration of UNS that assumes that instead of swing being uniform across all 650 constituencies it varies across them in a mathematically predictable way. Regression analyses are carried out on the Ashcroft data and used to extract patterns of VI change, and thereafter to convert these into seat projections.

    Using Ashcroft constituency data collected between August and October, I concluded that the (then) best formula for predicting LD vote share was:

    Current LD VI share in Constituency X = 0.47 x (LD vote share in X in 2010) – 3%

    Using this formula it is possible to “predict” the LD vote share that would emerge in the LD-held marginals reported in Thursday’s batch of polls. (The inverted commas are by way of acknowledgement that I didn‘t do these calculations in advance of the report: I didn’t know where then next set of polls would be. However, I posted the formula a few days ago – prior to poll publication – and in this sense the projections were determined in advance.)

    As a comparison the UNS “prediction” is:

    Current LD VI share in Constituency X = (LD vote share in X in 2010) – (Drop in national VI since 2010)

    (UNS then goes on to use this figure to calculate seat tallies in a notional election held today).

    How do the two models compare in anticipating the LD VIs in the 14 LD-held seats just polled by Ashcroft?

    Column 1: Constituency
    Column 2: Actual Ashcroft LD VI share
    Column 3: UNS prediction
    Column 4: Regression based prediction
    Column 5: Which model “wins”

    Birmingham Yardley 16.40% 24.56% 15.60% RB Model
    Brecon & Radnorshire 17.85% 31.16% 18.69% RB Model
    Burnley 8.29% 20.68% 13.77% RB Model
    Carshalton & Wallington 21.87% 33.30% 19.70% RB Model
    Cheltenham 22.39% 35.50% 20.74% RB Model
    Colchester 17.32% 33.01% 19.56% RB Model
    Devon North 20.25% 32.36% 19.26% RB Model
    Hazel Grove 21.10% 33.80% 19.93% RB Model
    Kingston & Surbiton 22.13% 34.78% 20.40% RB Model
    Lewes 21.64% 37.00% 21.44% RB Model
    Portsmouth South 15.04% 30.85% 18.55% RB Model
    Sheffield Hallam 16.77% 38.44% 22.11% RB Model
    Southport 19.39% 34.61% 20.32% RB Model
    Thornbury & Yate 25.67% 36.91% 21.40% RB Model

    As you can see, in many cases the regression formula could have given the answer before Ashcroft collected the data (and much more cheaply!). Indeed, the majority of the projections are accurate within the normal polling MOE. The model also ‘beats’ UNS hands down, basically substantiating my argument that we can improve our ways of dealing with these matters.

    Does this matter?

    UNS consistently and massively overestimates where LD VI now sits in these seats (with a 77% overestimate averaged over the 14 seats). (The corresponding figure for the RB approach is just 6%). This could cause the model to generate misleading predictions of at least two different kinds:
    (1) It may cause poll-watchers to overestimate how many seats the LibDems are likely to retain after the GE (more on this below)
    (2) It may distort the influence that ex-LD shifters have on the VIs for other parties. Churn data indicate that Labour benefits more from this group than the Tories do. Seat projections could go awry if, in marginal seats, the model overestimates (or underestimates) the number of LD votes that are available for shifting. I have previously argued that as a result of this the UNS model may be overoptimistic in in its GE predictions for Labour.

    In response to the suggestion that some apparently safe LibDem seats may be at risk, @Roger Mexico and @ Statgeek have pointed out that the Ashcroft headline figures seem to indicate that the reverse is true. It seems that, for good reasons, Ashcroft highlights the figures for VIs when respondents are asked to consider their own constituency (i.e, responses to his Question 3 and not to Q2 – which I have been analysing). This preamble presumably reminds people of their local candidate therefore producing an incumbency effect. Arguably an ‘own constituency’ VI is for the LibDems a much more relevant figure to try to predict, than is the unprompted response to Question 2. After all, voters are given lists of their own local candidates to annotate in the polling booth. There may be a case for repeating the exercise with Q3 VI as the dependent variable and looking to see how the two models compare here. Alternatively, an adjustment could be made to the projection by using existing polls to estimate the average Q3 incumbency VI redistribution and applying this transformation to the predicted VIs for each constituency with a sitting candidate.
    Note that questions about raw materials do not affect knock-on influences of the kind mentioned in Point 2 above. The LibDems currently hold fewer than 10% of all seats so in the other 90+% the LD VI should presumably not change systematically between Q2 and Q3.

    In summary, I would argue that there is enough order in the patterns of VI shifts to warrant setting aside the assumption of uniformity of swing in seat projection models. To the extent that uniformity of swing is demonstrably false, predictions should be improved by making changes of this kind.

  14. @ Old Nat

    I’m sorry to say, I don’t know whether there’s wifi at the Royal Infirmary or what it costs. I hope that you will be all sorted out & home so quickly that you won’t need it!

  15. This idea that once all the old codgers pop their clogs we’ll have EU Membership/Scottish independence/Labour government in perpetuity seems to be dependent on the assumption that people don’t change their political beliefs as they get older which seems quite shaky indeed.

  16. Oldnat

    All the best for you plumbing job and we’ll be waiting to see you back here multi pronto.

    I don’t know what phone you have but my smartphone has a wizard app that turns it into a wifi hotspot so I basically get free wifi anywhere in the UK – just a thought.

  17. Which app is that Guy ? I’m always looking for wifi connections when out & about because my mobile data connection is rubbish :(

  18. Guymonde

    I have an i-phone, so I’ll look into that. What’s it called?

  19. Unicorn

    That looks a big piece of work – I’ll let others critique it.

    Have you compared you method with the academics who are also trying to produce better predictions?

  20. If you have a smart phone go into settings and turn on the personal hotspot. You don’t need an app. Of course, you do need mobile phone reception and a package with a few GB of data…

  21. @oldnat

    Go into settings, select “Personal Hotspot”, and turn it on. It will generate a password for you the first time that the tablet or computer you want to surf on should pick up, and you instruct them to accept. After that it should connect automatically in future whenever you turn Personal Hotspot on.

    You need Bluetooth and wifi enabled on the iPhone, which you can also turn on in Settings. Alternatively, you can post on the site directly from the iPhone browser S”afari” at a pinch…

  22. @oldnat pt2

    A few extra bits…

    – personal hotspot can drain the battery pretty fast: you’ll need your phone charger!!

    – this practice is known as “tethering”, and not sure that all mobile providers allow it; sometimes they may charge a bit extra, you may need to check your phone plan.

    – on an iPhone, “Settings” is an app on the main screen. Sometimes the “Personal Hotspot” disappears as an option on the main settings page. To get it back, while still in settings, go into “General”, and then “Network”, and select Personal Hotspot from there, whereupon it will also reappear on the main page.

  23. @OldNat

    “Have you compared you[r] method with the academics who are also trying to produce better predictions?”

    Not as yet: partly since it is not entirely obvious how to do this. The two standout examples are electionforecast.co.uk and Fisher’s Swingback model.

    Both publish predictions about the GE itself rather about the outcome of a set of constituency marginals. So, it is not easy to see what they would say about an interim event of this kind. That said, for any particular electionforecast run, I don’t think there is any basis for VI estimates to change over time. So, until they have fed in the latest Ashcroft data I think their published predictions for the GE are effectively the same as they would be for an interim poll.

    So, I must rush and download their present predictions before they are removed from public view forever. Thanks for prompting me to do this!

    My understanding is that Fisher’s model does nothing more than adjust for Swingback over the approach to the election. If it were predicting any set of results for tomorrow, the Swingback adjustments would be set to zero (as no measurable Swingback can occur overnight). As its seat projections are made using the UNS model, I presume it’s predictions wouldn’t differ from the UNS figures I posted a few hours ago.

    On all these points I stand ready tobe corrected by anyone with more intimate knowledge of the workings of these models.

  24. ALEC

    Plasma Wars.


  25. OLDNAT

    Best wishes and hoping all turns out well.

  26. @Jim Jam (from yesterday’s 9am post)

    “Having read Roger over the years his analysis makes more sense to me.
    Intuitivelly the LDs vote holds up more in seats they are defending, in particular where the MP is restanding.”

    Yes, I’m sure you are right. In the Ashcroft polls the VI difference between Question 2 and Question 3 makes it clear how significant an incumbency effect there is in LibDem seats in particular.

    Without taking account of this my regression based approach does a very poor job of predicting Q3 VIs – much worse than the UNS model does.

    One way of handling this would be to use past Ashcroft polls to estimate the average size of the ‘incumbency boost’ and then to add this figure to the VI for the seat-holding party when making projections for future polls. I suspect this would do a lot to improve the Q3 projections.

    As an aside, I find it ironical that it is for entirely ‘accidental’ reasons that the UNS model outperforms my non-uniform account for Q3 projections. It mispredicts the LD VI shift for all seats. But in current LD-held seats the error just happens to act in the same way as an incumbency boost (even though the model is not furnished with any information at all about seat occupancy). So, rather like a stopped clock being right twice a day, it comes good by making an error that just happens to help in this particular kind of seat.

  27. Best of luck OldNat and I hope the Op goes smoothly.

    And don’t you worry about your Scottish crossbreaks- we’ll take good care of them and they’ll get spoiled rotten. If they look up at us with those mournful eyes I can’t promise we won’t put a bit of extra Labour into their diet but you can always wean them off that when you get back.

  28. Colin

    Re “ridiculous parody of Holmes ”

    Totally agree about that, stick to the original and Jeremy Brett was the master. However I do think Cumberbatch produced a very credible portrayal of Turing as I said earlier on the thread.

  29. @unicorn

    An archive of fisher’s model predictions can be built by looking at the entries on electionsetc dot blogspot dot co dot UK and the entries on electionsetc dot com

    However, the election forecast dot co dot UK model does not have a complete archive available online. Their twitter feed is partial – they have a full set of entries, but each entry is not necessarily complete! You will have to contact the academics and ask them, or create your own archive on the fly (but that latter option can’t be backdated)

    You can get old versions of websites via the Way back Machine (www dot archive dot org) or you can take archive copies yourself via archive dot today.

    Two questions for you.

    1) how do you measure the accuracy of a model? What statistic do you use to measure the error of, say, 35/33/15/9 vs a result of, say, 37/29/10/5?
    2) do you have any archives yourself you would like to share with the group?

  30. Someone commented about Ashcroft’s findings that a huge percentage if Sheffield Hallam voters had been contacted by the LDs in the past few weeks.

    I’ve just experienced their tactics. We’ve just had a local magazine pushed through the letterbox. Big header “Hallamshire LIFE”. Picture of Clegg and Miriam feeding school kids on the front cover, with the tag line “Help for families on the menu. We look at a busy and memorable year for Sheffield Hallam MP Nick Clegg”

    It’s obviously one of those syrupy free magazines that we get every few weeks. The ones written for middle class arrivistes in the one affluent part of South Yorkshire, full of articles about the latest in zero-carbon rattan furniture, reeling folk in to the real content – the adverts.

    Anyway, this one has a saccharine piece about Clegg. Then it goes into a list of adoring comments by locals about why Clegg is their man. And then a list of LD achievements in Govt. And I swear, it was only by that time that the penny dropped and I checked the font size 2 text at the bottom of the front page “Published and promoted on behalf of the Liberal Democrats”.

    They are a brazen lot. A couple of years ago, we got an election pamphlet with a big red banner and LABOUR prominent in red on the front page. Obviously a Labour publication. Except that it then listed a series of failings of Labour. It was, of course, another bit of LD mendacity.

    I assume Halamshire LIFE means that the LDs are really cracking it in Hallam. I do hope so. They deserve every tremor of their nethers.

    And here’s the funny thing. My house is 200 yards on the wrong side of the constituency boundary.

  31. I think I might know almost exactly where you live, Lefty. Don’t worry, I won’t pop round.

    Other Lib Dem tactics recently have included that “Are Labour Worth Voting For?” leaflet which contained a bunch of anti-austerity stuff and called Ed Miliband “More Tory Than the Tories”. With the Lib Dem imprint in one point font of course.

    Oliver Coppard’s stuff isn’t incredibly red-rose and Neo Sans, but it does at least include a reasonably sized imprint on the back and a Labour logo in the corner.

    The high LD contact rate is down almost entirely to their massive leaflet delivery in my view – they’ve got the money here to do it. Coppard’s strategy is mostly doorstep and phone canvassing based, which is much more effective, if slower.

    What confuses me is the crappy Focus leaflets that are bad riso prints with typos and awful bar charts. I can only conclude they’re intended to look down-homey and grassroots because I know they’ve got the cash for proper glossy stuff.

  32. @oldnat

    Good luck for the op. We will all be thinking of you.

    Being a bit bleak for the moment, leave instructions with somebody telling them how to post here in the event of something bad happening, so we don’t lose touch. Both ukpr and pb have a cohort of elderly and/or poor health individuals who post below the line, and I have noticed a pattern whereby older posters drop out, for reasons not known. If something goes wrong, we would want to know.

  33. OldNat

    Best wishes from the girls [and from me] for a successful op. and a full recovery.


  34. OldNat

    Best wishes from the girls [and from me] for a successful op. and a full recovery.


  35. Oh, and best of luck for the op this week, Oldnat. Hoping for a swift recovery!

  36. ON

    All the best mucker. Get yourself back on here sharpish.

  37. Latest from Scotland (basically no change, but some pretty charts):


    @ Oldnat

    Get well soon!

  38. MrN

    Go on. Have a guess.

  39. Well you’re not a student so I’ll assume you don’t live in Crookesmoor. The other bit that people often assume is in Hallam which isn’t is Hunter’s Bar.

  40. Hope it goes well OldNat.

  41. Either that or you live on Fulwood Road just inside the Broomhill boundaries.

  42. Mr N

    Surprising as it may seem to students (especially at 4am last Wednesday when some were playing football with beer bottles on our road ) a lot of real people live in Crookesmoor.

  43. On behalf of da yoof I can only apologise. I’ve spent a fair bit of my time up there recently (looking for houses actually) and not ALL of them seem to be inhabited by half a dozen drunken idiots.

  44. MrN

    It’s only a minority whose pockets to pay for ale are deeper than their bellies for holding it. Every community has kids like this. Although in the mining village in which I grew up, any kids behaving like that would have only done it once.

  45. I find second years are the worst for it – they’ve spent their first years in the student villages insulated from normal society so haven’t quite learned yet. The third years and Masters students are mostly too stressed out to drink.

    Students are also a massive pain to canvass, since they move every year.

  46. OLD NAT
    I am sure that a fellow of your vigour, will be gracing this board very very soon. Very best wishes.


  47. Old Nat, hope the op is 100% successful and you are soon back posting.

    Best wishes, Peter

  48. @The Pups,

    I was in the Golden Lion in Barny on Thursday afternoon. Have to say that the staff were very friendly and also provided nibbles for our walking group. Kept my eyes open for the pair of you leading some old geezer around but didn’t see you.

  49. @ R&D

    Nice to see you posting. I know you’re not at 100% peak fitness so good to see you are fine. Obviously people can have reasons for not posting but it is nice if they pop in from time to time.

  50. @Old Nat

    Hoping all goes well for you. It would be a shame for all of us, as well as for you, for you to be out of action for long!

1 6 7 8 9 10 11