## YouGov/Sun – CON 32, LAB 42, LD 8, UKIP 12

The Sun have tweeted out tonight’s YouGov poll – topline figures are CON 32%, LAB 42%, LD 8%, UKIP 12%. This is somewhat at odds with the Labour leads of two, three and three points so far this week, though for what it’s worth all four polls would be within the normal margin of error of CON 34, LAB 39, LD 9, UKIP 12, the average of this week’s figures.

When an unusual poll comes along I personally rather discount it – more often than not it’ll just be a blip. When the same happens two days in row it gets my attention, but I wouldn’t conclude anything. When you get three in a row I normally take it seriously, it looks as though something is afoot.

But it can still just be random chance. Right now we don’t really know what the position is. It could be that tonight’s poll is an outlier and other polls will continue to show lower leads. Alternatively it could be that actually nothing’s changed and its all just been random variation around the six point lead we’ve had for months. As ever, time will tell.

chrislane1945Thanks Anthony.

Makes tomorrow’s A Level Politics lesson a bit of fun.

ColinTen in a row-minimum.

mrnamelessPete B,

I know AW doesn’t like us carrying over discussions, so I’ve left a response on the old thread.

As for this poll, yes it’s a weird one. Probably an outlier (like the 2) but time will tell.

Worth noting that these polls average out to about a 6 point lead, i.e. exactly where we’ve been since September.

rosieanddaisieI think it is fair to conclude that this is a pukka pi…. I mean poll – and the previous ones were, wot we experts call, wrong.

mrnamelessI’m sure he’ll spin it as ‘volatile polls for Labour mean voters aren’t convinced’ or something else that’ll get him an earful from his mum.

chrislane1945Agree R and D and Roger H.

R HuckleCould it be the party political broadcasts that have caused this sudden change in Labours lead. We will see over the course of the next week or so. I doubt it, but from what I saw of Labours PPB, it came across quite well, as it was upbeat.

Is there any polling analysis as to whether PPB’s have any effect on polling ?

rosieanddaisieMore importantly, Arsenal’s attempts to sign up the entire German national team are on hold at the mo.

Pete BMr N (and AW)

I hadn’t realised that all these are within the MOE. Still, it’s unusual for swings from one extreme to the other. As Anthony says, it could well just end up back at the 6% lead that’s been the norm.

One point worth noting is that it seems to be the Lab vote that is most volatile. The others seem to be very stable.

howardLabour at 42 looks a bit high.

Anthony WellsR Huckle –

I don’t think there is any evidence of them having any impact at all, nor would I expect them to.

carfrewLet’s suppose that the chance of Margin of Error showing a Tory improvement in the range of two-to-three points is roughly 20%.

The chance of getting three such leads in a row would then be 0.2×0.2×0.2 = 0.008 = 0.8%.

Doesn’t sound much, but if you have say 300 polls in a year… it’s gonna happen from time-to-time.

rosieanddaisieIf “stuff” has any effect and if Rennard DID have an effect on LD support then it was a long time showing.

‘S’funny how sometimes correlation seems fairly self-evident but at other times one has no idea.

I still feel that all serious evidence points to the Tories having as close to no chance at all, barring really weird things happening, of achieving an overall majority and are very unlikely to have enough mps/votes [between both parties] to hope for another coalition.

I don’t think it matters in the least how good a story the LDs think they can tell or how brilliantly they tell it: the reality is that nobody will be listening.

carfrewwell, shouldn’t say “leads”, but improvements…

mrnameless“the Tories having as close to no chance at all, barring really weird things happening”

Never forget that if there’s one thing we can be sure of in this world, it’s that weird things have a propensity to happen!

sen5cOne thing that the poll isn’t is an outlier and nor were any of the earlier polls this week.

I always prefer to look at the individual party shares.4 of them are common in the last few weeks.Only Labour’s 42% is relatively unusual – the last in November I think. Yesterday’s Con figure was a fairly rare high too.

Consecutive figures are deceptive even though attention-getting.What I’ll be looking for will be regularly occurring 42 or 36 ( or higher) over the next few weeks which would indicate improving support for the party concerned with established economic reasons why that could be so.

I don’t expect to see anything of the sort though (in the absence of more ‘events’) The ‘true’ shares are close to Anthony’s average and likely to stay there until the Budget.

rosieanddaisiemr n

Well, life is weird, yes.

But I is talking about really, really weird.

Mars invades and they’re all bleedin’ Tories for example.

[They probably would be, the buggers.]

richIf Labour get 42% at the next election I’ll eat my hat. In fact I’ll eat a hat shop.

carfrewSo every hundred polls = 0.8% x 100 = 80% chance of it happening…

or roughly every four months, there’s an 80% chance of folk having a needless heart attack over the polls…

Colin DavisI haven’t checked today, but I think a Labour 39 would have brought their monthly average splat onto 38. So it’s 38 and a teensy bit over for the month, exactly where it’s been all along. Conservatives are on just over 32 for the month, exactly where etc. etc. UKIP on just under 12.25, LDs almost splat on 10.

Moments of this and that (Balls sounding like he’d be running his own austerity policy; Tory press tell the world about the economic miracle) and the Don’t Knows on the fringe waver and the weather seems to change, but in any kind of underlying way it doesn’t.

What do the Tories have to do to get the UKIP squad back on side? Changing a Don’t Know’s mindset is hard when (almost by definition) he hasn’t got one. Osborne is surely planning to become Loadsamoney Osborne pdq.

Chris MartinThe averages are what count. Anything else is augury.

Colin DavisI’ve just posted to say nothing changes. I’ve mentioned all parties and made no comments on any of their policies – and still it’s ‘moderation’. So let’s see. Nothing changes. Will it go through?

Guymonde“If Labour get 42% at the next election I’ll eat my hat. In fact I’ll eat a hat shop.”

Let me know before you eat it. I’ve been looking for a hat shop for months (not to eat it)

Seriously – about events – this has not been a very auspicious day for them what’s in power what with Zummerzett and Commonsishambles. Too early for these to affect polls I suppose.

rogerhPerhaps it’s due to the popularity of the 50% tax rate!

Paul Bristol@ ALLAN CHRISTIE

“It’s getting warmer….

They’re behind you….

Oh no they’re not..

Oh yes they are…”

———-

Oh no they’re not…LOL

I knew this one would come back to haunt you. ;-)

Is this the first poll where all the field work was done after the announced 50p tax rate return to be reinstated by Ed Balls?

Very interesting if it is.

Chris Martin@Carfrew – “So every hundred polls = 0.8% x 100 = 80% chance of it happening…”

You can see for yourself that that is incorrect by imagining what would happen, using your maths, every 200 polls = 0.8% x 200 =160% chance of it happening! Even to non-mathematicians that ought to be obviously impossible, thus disproving your methodology.

Instead, you ought to have calculated 1 – ( 1 – 0.8% ) ^ 100 = 55% of it happening. Incidentally, 1 – ( 1 – 0.8% ) ^ 200 = 80% of it happening, a neat coincidence.

howardYou have to await tomorrow’s demolition of the sample by Statsgeek and Roger Mexico, spiced with a churn analysis by Spearmint before you can claim anything here.

rosieanddaisieIt always amuses me, the silence from Sine in Nomine, when there’s a favourable poll for Labour.

[TIC/Irony]

Chris MartinAnthony, may I ask if your “three consecutive polls = serious” approach takes into account the likely dependence of residuals in a time-series such as this one?

rosieanddaisie@ ALLAN CHRISTIE

“It’s getting warmer….

They’re behind you….”

They are indeed…. quite a long way behind in fact Allan.

Still, neh’mind ay?

rosieanddaisieMaybe the LOC should have a concerted chant of

“You’re not singing, you’re not singing

You’re not singing anymore.

You’re not si-inging any more.”

After four please…………………

Four.

Chris MartinWhen are the tables likely to become available, does anyone know?

Bill PatrickLooks like reversion to the mean to me, which shouldn’t be too surprising, since there was no reason for a sudden shift in VI, and no consistent story told by the various polls during the brief snap out of the polldrums.

Of course, if double-digit leads become common again, THAT would be a very interesting shift, and demand some sort of explanation.

carfrew@Chris Martin

Yeah well, I haven’t done stats since Uni, and that was Bose-Einstein stats and stuff. That said, one of my New Year’s resolutions was – inspired by this site – to get a book on Stats and dust off some maths brain cells…

As you can see, I haven’t opened it yet. I got waylaid by cameras and stuff…

On the other hand, my approach was close enough to make the point and a lot easier; I look forward to you explaining yours!! Especially the time series thing…

Chris Martin@Rich – If Labour get “within the margin of error” of 42% at the next election, I won’t be greatly surprised.

Bill PatrickI enjoy it when no-one sings nad we have non-partisan discussion of polls.

rosieanddaisie“I enjoy it when no-one sings nad we have non-partisan discussion of polls.”

I’m the same with footy and goals Bill.

Anarchists Unite@R & D

“Mars invades and they’re all bleedin’ Tories for example.”

But would they be able to get past immigration?

rosieanddaisieA U

Have you seen their ray guns?

Paul BristolI’ll eat my hat ……if this poll is widely reported, let alone make headlines ,,,,,,

Ah well, back to sixes and sevens tomorrow, no doubt.

@ROSIEANDDAISIE

“sing when your winning…you only sing when your winning………………..”

Chris Martin@Carfrew – I wasn’t telling you off, just sharing the accurate figures!

The calculation you should have done involves the following reasoning:

A given event has probability p. We are interested in whether that event occurs at least once – not every time, not precisely once, but at least once. If you think about it, “This event happens at least once” is the same as saying 1 – “This event never happens”. That is, either the event happens once in the series or it never happens – there are no other options.

The probability of the event never happening is ( 1 – p ) ^ n, where n is the number of instances in the series. Again, that’s because at any given instance, the event either happens or it doesn’t happen – there are no other options, so all that’s left over once you discount p is 1 – p. The ^ n is hopefully obvious because it’s just multiplying the probability by itself every time there’s a new instance.

We can combine these things together now! The probability that the event never happens in the series is ( 1 – p ) ^ n. The probability that the event happens at least once in the series is 1 – “never” = 1 – ( ( 1 – p ) ^ n ). QED! :)

rosieanddaisiePB

We is.

Chris Martin@Carfrew – Now for the time series.

When you are trying to draw a line through some dots on a scatterplot (which is sort of what the polling average does), the dots are all at various distances from the line.

Sometimes, in time series, because the direction of travel of the line (left to right, i.e. time) is not random (it’s directional, because time is directional), those distances between the dots and the line are not random either, but are directional, like time.

That is, sometimes in a time series the dots skew below or above the line more “clumpily” than you would expect on a graph that wasn’t a time series.

I wondered whether Anthony was accounting for that extra clumpiness when he said that he takes three consecutive polls seriously. The subtext was that I wondered whether he should be even more cautious about clumps of data points than he is already.

rosieanddaisiechris martin

“The calculation you should have done involves the following reasoning:

A given event has probability p….. ”

Let me stop you there – I have no idea what yer on about.

Pete BR&D

I suppose that’s understandable – dogs don’t go to school.

Chris Martin@R&D – “Let me stop you there – I have no idea what yer on about.”

Hello,

I often feel the same when reading your comments about football and other such non-polling, non-political subjects. It is certainly always surprising to find you discussing such an array of non-polling topics on a blog purportedly about polling.

If you have any questions that could help you to engage better with the polling and numbers, please let me know and I (and others!) will surely try to answer.

Kind wishes,

Chris Martin

R HuckleBad council by election result tonight for the Tories and good for the Lib Dems. If the Tories lose this many votes to UKIP in a general election, they will be in trouble.

Chadsmead (Lichfield) Result:

LibDem – 36.0% (+4.8) – 206

LAB – 27.4% (-1.6) – 157

UKIP – 18.8% (+18.8) – 108

CON – 17.8% (-22.0) – 102

Phil HainesLight at the end of a week’s polling tunnel.

rosieanddaisiechris m

“it’s directional, because time is directional”

http://www.youtube.com/watch?v=NvpbW7JRu0Q

I think you must be studying with Miss Cunk.

Chris Martin@R Huckle – Yes, Mike Smithson points out that “the Tories have lost every principal authority seat they were defending in January”.

What interests me about the result is the suggestion that Lab-to-Lib tactical voting is alive and well.