Two types of outlier

This morning TNS released a new poll showing figures of CON 28%, LAB 35%, LDEM 6%, UKIP 18%, GRN 7% (tabs). The seven point Labour lead is striking, and out of line from the broader trend. Usual caveats apply, but for once I haven’t seen many people over-react to it. Most have sensibly enough seen it as the just the other side of the coin to the ICM poll earlier this week showing a solid Tory lead – two outliers in opposite directions. However, it is worth looking at the different reasons why these two polls went against the trend.

On average ICM produce figures that are slightly more Con/less Lab than average, but only by a tiny bit, generally they show Labour -v- Tory leads that are much the same as those from other companies. The reason that their poll on Monday showed a four point Tory lead when other companies are showing the race almost neck-and-neck is almost certainly just random sample error. Samples vary from day-to-day, month-to-month and sometimes you get one that’s a bit Laboury or a bit Conservativety in ways that weighting does not correct. I expect ICM’s next poll will be in line with those from other companies.

TNS is a different sort of outlier. The seven point Labour lead may be different from that shown by other pollsters, but its actually in line with TNS’s previous polls. In their previous four polls they showed Labour leads of 6 points, 0 points, 7 points and 7 points. Over on my chart of house effects TNS’s polls on average show a Labour lead three points larger than other companies (and if anything the gap is growing!) This isn’t random sample error, this is a consistent methodological difference between TNS and other companies, and unless they switch methods I expect their future polls will continue to show bigger Labour leads than other polls.

Two outliers, but one is probably just a random blip from a pollster who normally shows the same as other companies, one looks like a typical poll from a pollster who regularly produces polls that show a bigger Labour lead than the pack.

311 Responses to “Two types of outlier”

1 4 5 6 7
  1. @ Laszlo,

    I certainly don’t think his thinking-on-his-feet skills are Miliband’s biggest problem- Hague is fantastic at this and regularly trounced Blair at That-Which-Must-Not-Be-Named, and we all know how well that worked out for him- but I do think he’s objectively quite bad at it. He’s fine as long as Cameron answers as he expects him to, but if Cameron throws him a curveball he can’t recalibrate his next question.

    As Energy Secretary he was perfectly competent at giving fluid answers to opposition questions so I don’t think he’s intrinsically incapable, but for whatever reason since he became Opposition Leader he’s lost the ability.

    You’re absolutely right, if he could articulate public discontent in a way that convinced people Labour had the solutions then his debating skills wouldn’t matter. But he hasn’t, so the election is still close and it would be quite helpful for him to have a 100% chance of winning the leadership debates instead of ~50%.

    @ Crossbat 11,

    I think the old fashioned electioneering techniques and virtues will hold sway in May.

    David Burrowes may be in trouble, then:

  2. To be fair to Burrowes, it’s easy to stray a little bit. I was canvassing the very southernmost tip of Crookes yesterday, where there are three roads which have the even numbers in Sheffield Hallam and the odds in Sheffield Central.

    Half a mile is pushing it though. Train your board runners. (Or however the Tories do it).

  3. @ Funtypippin

    Looking back at those Survation figures I’m curious about how they fit into Unicorn’s quadratic model.

    Perhaps the least said about that the better, but since you ask…

    Like @ ON’s rolling YouGov averages they don’t fit at all well. Remember that what I was trying to do was to describe the SNP VI trajectory since August. The quadratic curve did a good job of capturing the rapid rise and subsequent levelling of the graph and the next step was supposed to be an equally dramatic plunge. In contrast with the SNP/Lab margins, the SNP figures don’t go down, so the fit is just as bad as it has been for the YouGov figures.

    I think you can treat the quadratic description as a Monty Python parrot.

  4. The Daily Record have come up with the same seat tally I came up with.

    Forty-five per cent of Scots now plan to vote SNP on May 7, down one point from our last poll in January. Support for Labour rose two points to 28 per cent.

    The Tories rose one point to 15 per cent and the Lib Dems dropped two points to a humiliating five per cent.
    *If our results were repeated in the actual vote, the SNP would win 47 seats and Labour 10*
    The Lib Dems would lose all but one of their 11 Scots MPs and the Tories would keep their solitary Scottish MP.

  5. @Unicorn

    Are you predicting an insurgency of Nordic Conservatives (possibly demised) into Scotland ?

  6. @ Old Nat

    “As one North American leaves – another returns! Welcome back.
    Fabulous report from West Holywood, that makes Holyrood seem boringly banal.”

    Thanks man. Yeah, I can’t be here as much as I used to but glad to see how exciting this election is.

    I suppose it does make Holyrood look banal. I don’t know what is going on. But let’s just say this, I think there’d be quite an uproar if it was learned that David Cameron hired people he met off Grindr to high level staff positions.

    @ Phil Hanes

    “Current odds:
    UKIP 4/1 to win 50 or more seats (Hills).
    Greens 4/1 to win 2 or more seats (SkyBet), 16/1 to win 6 or more, and 2/1 to win none. They are expected to win 1.
    I think that’s quite telling, as 4/1 is long odds but not quite long enough to rule out any chance of winning without a huge change of circumstances. Greens are in contention in just two or three seats nationally, whereas UKIP are in contention in perhaps around 50 of them.”

    The election is in May right? I’d argue that’s a lifetime in politics.

  7. @Northumbrian Scot

    No idea. I don’t do UNS calculations. I’m sure a omni-horned horse, or a chap from Mexico might be able to help though.

  8. @ Statgeek

    Horse!? Hah!

  9. @TheOtherHoward

    “Have a good day all.”

    Thank you.

  10. @ JohnB160

    A Labour total of 311 and a SNP total of 43.

    SNP currently have 6, From whom would the other 37 come? Let’s assume 9 Lib Dem and 28 Labour. That leaves Labour with 13 in Scotland so….298 in E&W.. When you factor in the Scots losses that would be almost 70 gains.

    That is a lot.

    I think Ed would be very happy with 298 in the whole of GB
    And probably so would Dave

    At what point do the parties start saving funds for 2015’s second election?

    At what point do the parties start saving funds for 2015’s second election?

    When all except Lab & Con are ready for one unless Lab & Con can agree to vote NO together in a vote of confidence?

  12. @ Little Red Rock. Word of course is that the Conservatives have already started saving (or have enough money not to worry) so for Labour there is a big incentive to make the first result work if possible. Whether they are culturally attuned to the kind of collaborative working this is likely to require is another matter

  13. Latest Ashcroft constituency polling now up, though it’s a bit small – four UKIP target seats:

    UKIP not actually ahead in any of them – even Boston.

    I assume he’s saving something for his his next batch:

    My next round of marginals polling – among other things – will be launched at my polling presentation on 4 March. Why not book your ticket now?

    because this is fewer than we’d normally get from him.

  14. @ Chris in Cardiff

    Are you suggesting that their cultural attunement is sub optimal? That sounds very serious.


    But if neither has an overall majority how would Con or Lab contrive to lose a confidence vote and “break” the fixed term act unless they both want an early election?

  16. So Lord A has been looking at Con/Ukip marginals, concluding close but no cigar (but very close in Castle Point)

    I was a bit surprised to read:

    “UKIP appear to have the edge in these seats in the ground campaign. The battle seems most closely fought in South Basildon & East Thurrock, but the difference is most marked in Castle Point, where people were more than twice as likely to say they had received literature, letters, visits or phone calls from UKIP as from the Conservatives.”

  17. but I’m not bold enough to predict the actual numbers at the election, at least not yet.

    when? 8th May?

  18. UKIP have a lot of money, and only a handful of seats to target. The Tories have a lot of money, but seats where they got 44%+ of the vote in 2010 (as with all of these) are likely to be low down the spending list.

  19. Spearmint,
    You’re absolutely right, if he could articulate public discontent in a way that convinced people Labour had the solutions then his debating skills wouldn’t matter.

    These skills are intimately connected. It’s all about verbal skills, expression and rhetoric. debating and the ability to speak clearly and persuasively…communication skills…executives pay thousands for training in this….Mili’s communication skills are very weak.

  20. @Guymonde

    I suspect that’s more to do with the lack of a ground campaign from the Conservatives, which seems to be true nationally. See past discussion on the party being reliant on paid leafleting companies.

  21. I also note that UKIP voting intent actually drops for the “In your constituency” question on two out of four.

  22. Peter Crawford

    “but I’m not bold enough to predict the actual numbers at the election, at least not yet.
    when? 8th May?

    Very droll. I am happy to just say a small Tory majority. The same forecast as for the last four years. Nothing in the polling to date suggests anything else to me.

  23. ‘At what point do the parties start saving funds for 2015’s second election?’

    Don’t rule out that the post-Election Parliament will conclude ‘reluctantly’ that the Electorate will become the lucky source of all future Political Party funding.

  24. After a quick glance, it looks to only be the spiral of silence adjustment that is putting the Tories ahead in Castle Point, and even then by well less than what I presume would be the margin of error of +/- 3%. A definite too close to call I reckon.

    Shame he hasn’t released a Thurrock poll as there is the further factor of a (on paper) competitive Labour Party in that seat.

  25. Re: UKIP Targets

    Unless Nigel woos the public in the campaign or has an “I agree with Nigel” debate it seems unlikely that UKIP will improve significantly on present VI so a handful of seats at best.

    What was the latest poll in Thanet, and was that before Al Murray entered the race?

  26. Hawthorn…

    not sure what you mean by the “spiral of silence adjustment”, but if you’re saying that knowledge that it’s neck and neck may push things in ukip’s favour in these seats, I agree it’s very close indeed.

  27. The score so far in Clacton this year – two Tory leaflets; one UKIP.

  28. @JayBlanc

    “I suspect that’s more to do with the lack of a ground campaign from the Conservatives,”

    In most post-war elections when the eventual winner was more or less known months in advance of the election, then the strength or otherwise of the respective parties ground campaigns didn’t really matter. It provided mere window dressing to a decision already made. However, when matters are close, as they are for the election in May, then campaigning becomes more important, possibly crucial, in terms of getting your core support out and maybe influencing those still undecided.

    The Tories, once dominant in this area, are now at a disadvantage such is the decline in their membership and the number of activists available to them. Accordingly they are much more reliant on the “air war” where their vastly superior financial resources give them a clear advantage, but I think they suffer on the ground now, certainly in comparison to Labour, UKIP and the Lib Dems. This could be a problem for them when the foot-slogging part of the campaign gets under way.

    Up to 1992, I always felt that the Tories were better organised and more numerous than Labour when it came to ground campaigning, but that changed utterly in 1997 when the sense I had was that they had almost entirely melted away. They gave 2010 a better go, but I don’t think they’ve recovered from their mid 90s near death experience.

  29. Wow, those Ukip polls are quite bad for them. If they can’t even win Boston they’re unlikely to gain anything, except maybe South Thanet if Farage has enough star power. Possibly Thurrock, but I suspect Labour and Ukip will split the anti-Government vote and let the Tories hang on.

    @ Peter,

    These skills are intimately connected. It’s all about verbal skills, expression and rhetoric.

    No they’re not. Blair was an fantastic communicator who regularly lost debates; Foot was a fantastic debater who couldn’t even communicate with his own party, much less the country. Miliband happens to have neither skill, but they don’t necessarily co-occur.

  30. UKIP ahead in Boston and Castle point before reallocation of don’t knows.


    I have spent about one minute looking at Ashcroft’s table. If you do so too, all becomes clear.

  32. Actually, I don’t think this is too bad for UKIP unless they were genuinely expecting to get more than around 5 seats.

    A lot of their gains were always going to be close.

  33. @Little Red Rock

    Electoral Calculus is currently showing 69 Lab gains and 29 losses based on a vote share of 33% Lab and 32% Con. A two point Labour lead would increase the number of gains a bit and get you to something like the figure that Anthony’s advanced swingometer is producing. Not a prediction obviously but an illustration of how much DC still has to do.

  34. JOHNB160
    I agree that we are, at the moment, much to my surprise, a net gain for Labour of about forty seats, getting ‘the party’ close to 300, with the three SDLP -ers, rudely called by the ‘Shinners’ as the Stoop Down Law Party, taking the Labour whip.

    That should be enough to put Ed and Ed into numbers Ten and Eleven Downing Street at the head of a minority Labour Government.
    Then the ‘fun’ will start. Wilson called it a ‘racket’ in March 1974.

  35. As we’re discussing UKIP this morning I was wondering last night about UKIP’s prospects of getting MPs elected to the Scottish Parliament via the regional list.

    The remarkable thing about UKIP’s Scottish vote in the 2014 European Elections was how consistent it was at a regional level.

    They outpolled their national average in 3 of the 8 regions:
    Highlands & Islands 12% up
    South of Scotland 7% up
    Central Scotland 5% up

    They underperformed in 5 of the regions:
    North East Scotland 3% down
    West Scotland 3% down
    Mid Scotland & Fife 6% down
    Glasgow 9% down
    Lothian 17% down

    If we look at the threshold necessary for election on the list over the years it has varied with people elected on as little as 4.5% of the vote and not elected on as much as 6.7% of the vote.

    Generally though anyone polling 5.5% or more will be elected so I’ll use that as my cut off point.

    What this means is that assuming a similar UKIP vote distribution on 5% of the regional list vote UKIP are likely to get only 1 or 2 MSPs in Highlands & Islands and possibly either South or Central Scotland.

    On 6% of the regional list vote UKIP could get around 6 MSPs, missing out only in Lothian and Glasgow.

    On 7% of the regional list vote (which they polled in yesterday’s Survation poll) UKIP are likely to get a full set of 8 MSPs and could overtake Lib Dems to become the 5th party in the Scottish Parliament (still behind SNP, Lab, Con & Green).

    To get more than 8 MSPs they would need to poll 10% or more which seems unlikely at present.

    Will be interesting to see what the Scotland Votes model makes of UKIP when it is updated.

  36. Another Smithson observation on those UKIP seats:

    “Just 37.5% of Ukip voters in latest @LordAshcroft marginals polling voted LAB/CON/LD at GE2010. Looks like a big number of non voters”

    The headline, that they should hold these seats is encouraging for Tory supporters, but it should be sobering for those hoping to get more than a couple of percent VI from squeezing UKIP prior to the GE.

  37. The talk of Labour being outgunned by Conservatives in terms of campaign funds doesn’t seem to be supported by the latest figures for the last quarter of 2014.

    Labour just over £7m and Tories just over £8m. Not a dramatic difference.

  38. The “spiral of silence adjustment” is one name for it. A less charitable name, in the present context, would be the “same as last time adjustment”, because that is its practical effect.

    There are a couple of reasons why UKIPers might be a bit upset by its use in Conservative/UKIP contests:

    1. As implied in (my) title for it, it is designed to work against newly insurgent parties, so by design UKIP will get next to no reallocation from the “don’t knows”.

    2. If there really is a “spiral of silence” at work, can we buy the idea that Conservatives are the pariah party in Con/UKIP contests and that as a consequence their supporters might therefore be more reticent than UKIP supporters in disclosing their intentions? Might it if anything be the other way around, and that polling is more unkind to parties the closer they get to the far end of the right wing spectrum?

    In the context of this poll, it’s best to read it before the final adjustment is made, which would have put UKIP ahead in two out of four of the seats.

    I’ve commented before about the need to be a bit sceptical about aspects of Lord Ashcroft’s polling and I think this is another example where that warning should be born in mind. He is not a disinterested observer, but someone with an agenda.


    “I still think that Labour will win a majority, because I expect them to get about 34%, and the Tories to lose votes to UKIP during the campaign”

    If Westminster is to have any credibility then this outcome would not be welcomed. Fine if Labour (or the Tories) win most seats and run as a minority administration or shuffle along with some sort of c&s but a winner takes all on such a mediocre percentage in my view starts chucking up alarm bells over the legitimacy we are governed.

    34% surely can not impose its will on 66%?

  40. Northumbrian Scot

    Interesting analysis.

    Have you had a look at which parties would potentially lose list seats to UKIP under these scenarios?

    As we know, even when the Returning Officer does his sums right, who gets the last seat can be very close.

  41. @RichardW

    It’s not about funding. It’s about people. Funding gets glossy adverts, it does not produce motivated political activists. And the ground game needs those people for canvassing, locally tailored leafletting, turning up for events and what’s more to give a local-legitimacy to candidates.

    Those things may well only be shifting VI by a point and a half… But in this election that really matters. And it’s where the Conservatives are having campaign problems.

  42. “34% surely can not impose its will on 66%?”


    Lab 35.2% Con 32.4% LD 22.0%

  43. The pre-reallocation figures are interesting (ignoring under 5%):

    Boston and Skegness

    Con 36%

    Lab 17%

    UKIP 37%

    Castle Point

    Con 36%

    Lab 16%

    UKIP 38%

    Other 6%

    NE Cambridgeshire

    Con 45%

    Lab 17%

    Lib Dem 7%

    UKIP 26%

    South Basildon and East Thurrock

    Con 34%

    Lab 28%

    UKIP 30%

    I suspect the ‘Others’ on Castle Point are Canvey Island Independents. As these seem to have an agreement with UKIP this might provide them with extra voters (though the English Democrats are also standing, so some of it could be for them).

    Irritatingly the Ashcroft polls still don’t seem to have fixed the LTV bug, though I haven’t checked to see if it makes any difference. Unlike with the dodgy tables for the Leaders’ polls it’s also difficult to see if the reallocation has correctly.

    One strange thing is that in the movement from SVI to CVI, UKIP tend to lose as much as they gain – not what you’d normally expect in a small Party’s target seats. Presumably, unlike with the Lib Dems, many voters haven’t registered that they are in a UKIP target.

  44. @OldNat

    It’s very tricky to predict, as you say the final list seat is often very close.

    Looking at the 2011 numbers it looks like most of the time UKIP would be gaining at the expense of SNP, with a few Labour. If SNP pick up a few more West of Scotland constituencies than 2011 then the danger is more for Labour.

    They would also be a danger to the Conservatives in Regions where they have 2 list MSPs on 10-12% of the vote (Highlands & Islands and Lothian) and could make it more difficult for the Greens to get a second list MSP in Lothian and Glasgow.

  45. The publication of a mini-batch of Ashcroft constituency polls provides another opportunity to compare the on-going accuracy of the models that publish individual constituency projections.

    With data from just four constituencies, it is impossible to draw any robust conclusions. But taken together with earlier analyses it is possible that the new material could sharpen the focus a bit.

    As before, May2015 was markedly less accurate in its predictions than Electionforecast (EF) and ElectoralCalculus (EC). Ashcroft had Ukip as the runner-up to the Conservatives in all four constituencies. May2015 picked different runners-up in all four cases: an independent candidate (presumably Bob Spink – not yet a declared candidate) for Castle Point, and Labour in each of the other cases. As a result of this, May2015 provides no VI figures for Ukip in these seats and so it is impossible to calculate Euclidean Distance (ED) measures for them.

    Turning to a comparison between the EF and EC models, the average Euclidean Distance (target inaccuracy) measure showed that EF did better in these four constituencies (Mean ED = 5.8 for SVI and 7.3 for CVI) than was the case for EC (Mean ED = 12.9 for SVI and 12.1 for CVI). In the past, EC has occasionally matched EF for initial accuracy, but then fallen back because it is not so effective at incorporating the new constituency polling data into the database.

    Notwithstanding necessary caveats about the very small sample size, these data continue to support the view that the Electionforecast model provides the most accurate current description of VI profiles across constituencies. (Note that the benchmark for accuracy here is the generation of VI profiles that are very similar to those that emerge in Ashcroft constituency polling. If these polls are themselves flawed in any way, than all bets are off…)

  46. In fact looking at the Holyrood numbers Scottish Conservatives are really quite vulnerable and a relatively small drop in percentage support on the list could drop them from 15 seats to 10.

  47. @ Jayblanc

    “It’s not about funding. It’s about people.”

    I agree. My point was that the suggestion that Labour would be massively outspent by the Tories looks implausible (and I do wonder how much money can be usefully spent in the last couple of months).

  48. Allan Christie

    34% surely can not impose its will on 66%?

    I think you might be overegging it a bit there. It’snot really that far from 36% imposing on 64% (2010) or 35% imposing on 65% (2005) or even in principle 43% imposing itself on 57% (1997). None of those seemed to convince the public that FPTP should be scrapped in 2011. Why would it be different this time?

  49. You’d think at least Castle Point residents would realise they were in a UKIP target – I presume the actual polling was done after UKIP’s campaign launch?

    Surely this could swing voters to UKIP if they realise that it isn’t a wasted vote in these seats?

  50. Roger Mexico

    Have you contacted the good Lord directly about the weighting issue?

1 4 5 6 7