UK General Election, June 8th 2017

[Filuwaúrdjan]

An important thing to remember: even if the polls are not only overestimating Labour but are doing so significantly (neither fact is certain) then Labour is still looking at a significantly better result than seemed likely when the election was called. Anyway, I think the best thing for those on the Left to assume is that the election will be a worse loss than 2015, though this is for reasons of psychological preparation/emotional management than the product of rational analysis.

Still, there's a definite feel of UNSKEWED POLLS!!! to a lot of recent analysis; rather too blatant attempts to find reasons to dismiss findings that appall. Opinion polling in this land is very bad, but the one thing it can pick up fairly well is momentum. As David Butler has pointed out this election has seen the biggest ever in-campaign Lab/Con polling shift; there is simply no way that this has been caused purely by more young people telling pollsters that they will vote.

[YL]

here in Plymouth Moor View, a seat the tories gained from labour in 2015 with a majority of 1,026 (2.4%)
both tory and labour activists predict the tories will win by 8-10% majority

this poll doesn't reflect the reality in the marginals

Moor View is "Lean Conservative" in the YouGov model.  It has Sutton & Devonport as a Labour gain, though.

(I don't know that much about how Plymouth's demographics vary across the city; how different are the two constituencies?)

[Phony Moderate]

As David Butler has pointed out this election has seen the biggest ever in-campaign Lab/Con polling shift; there is simply no way that this has been caused purely by more young people telling pollsters that they will vote.

He was on Newsnight the other day; seemed in pretty good nick.

[YL]

Some more of that YouGov model's output:

Potential Lab gains from Con (just a selection, I haven't found them all):
Bedford
Ipswich
Hastings & Rye
Brighton Kemptown
Croydon Central
Plymouth Sutton & Devonport
Stroud
Bristol North West
Cardiff North
Lincoln
Keighley
Battersea

Potential Con gains from Lab:
Middlesbrough South & East Cleveland
Stoke-on-Trent North

They have Independent Claire Wright ahead in East Devon.  As for the Lib Dems, it has them losing Southport to the Tories and in trouble in Carshalton & Wallington, losing Leeds North West to Labour and in trouble in Sheffield Hallam, but gaining Twickenham, Eastbourne and Kingston & Surbiton.

Pinches of salt required, I suspect.

[vileplume]

Sigh this is a load of drivel. They have Labour getting 30% in Bridgwater and West Somerset for example. For a bit of context they didn't even get that in 1997 even when the boundaries were better for them (boundary changes in 2010 took in the monolithically Tory Exmoor).

[Phony Moderate]

Sigh this is a load of drivel. They have Labour getting 30% in Bridgwater and West Somerset for example. For a bit of context they didn't even get that in 1997 even when the boundaries were better for them (boundary changes in 2010 took in the monolithically Tory Exmoor).

Labour might get some surprisingly high percentages in some Southern Tory seats but yeah that seems high.

[Tintrlvr]

An important thing to remember: even if the polls are not only overestimating Labour but are doing so significantly (neither fact is certain) then Labour is still looking at a significantly better result than seemed likely when the election was called. Anyway, I think the best thing for those on the Left to assume is that the election will be a worse loss than 2015, though this is for reasons of psychological preparation/emotional management than the product of rational analysis.

Still, there's a definite feel of UNSKEWED POLLS!!! to a lot of recent analysis; rather too blatant attempts to find reasons to dismiss findings that appall. Opinion polling in this land is very bad, but the one thing it can pick up fairly well is momentum. As David Butler has pointed out this election has seen the biggest ever in-campaign Lab/Con polling shift; there is simply no way that this has been caused purely by more young people telling pollsters that they will vote.

Not totally sure this is true. Cleggmentum was only 7 years ago.

[Blair]

According to the Times Red Box podcast that has some former Tory staffer on it May only ever expected to win 60 odd seats; but that goes against the stupid briefing that was done at the start about Watson+Skinner etc losing their seats, and Labour getting wiped out.

It does seem now; at least from polling/various scatters of news that A.) The Tories have ran an awful campaign B.) The Public don't like May as much as they dd C.) Labour have a relatively winning message. I honestly believe that if someone without JC's kooky views on the IRA that this could genuinely work.

However part of me still feels someone at CCHQ is laughing looking at their internal polling; like in 2015

[vileplume]

The more I look the more it becomes apparent that they've just 'plucked numbers out of thin air'. As an other example they have the Tories up 10% to 60% in Sittingbourne and Sheppy yet unchanged in Dartford which has similar demographics, similar Brexit vote and only a slightly smaller 2015 UKIP vote in 2015.

Even if you accept the 41-38 split (which I think is very wrong anyway) their constituency numbers would still be way off.

Also their geography is in places as horrendously bad as their constituency numbers. Since when was Stroud south of Bristol?

[Shadows]

British people overwhelmingly support Corbyn in all wars related to foreign policy (except for Falklands). Iraq war has a disastrous rating with Libya & Afghanistan close by


In a speech today Jeremy Corbyn will address "the connections between wars our government has supported or fought in other countries and terrorism here at home".

Do you believe that wars that the UK has supported or fought are or are not in part responsible for terror attacks against the UK?

Wars the UK has supported or fought ARE responsible, at least in part, for terror attacks against the UK -  53%

Wars the UK has supported or fought ARE NOT responsible for terror attacks against the UK - 24%

Don't know - 23%

Voters say by more than 2/1 margin that British wars are responsible for terror attacks

[Shadows]

UK election: Theresa May sidesteps question on whether she will resign if Conservatives lose seats


Theresa May has sidestepped a question over whether she will resign if her decision to take the country to the polls sees the Conservatives lose seats in the Commons. “There is only one poll that matters and that is the poll that takes place on 8 June, and when it comes to that poll people have a very clear choice. “That choice is about who is going to be Prime Minister and it’s about who is going to lead the UK in those Brexit negotiations, who has the plan to do that, the determination to get the best deal, who has the strong and stable leadership to do that.

http://www.independent.co.uk/news/uk/politics/uk-election-theresa-may-resign-conservative-leader-tories-lose-seats-commons-majority-hung-a7765256.html


Jeremy Corbyn on Wenger, grassroots football and Labour's Istanbul moment


Are you ‘Wenger in’ or ‘Wenger out’?

Wenger in. I think you have to add up his achievements: the league wins, the Cup wins, the 20 years in the Champions League. I think he is the guy who has brought the club forward to a world status and we should recognise that.

Top four for next season?

Well, of course Arsenal will be there. I think it’s gonna be Arsenal, Liverpool, Spurs, Chelsea.

https://www.theguardian.com/football/copa90/2017/may/31/jeremy-corbyn-arsenal-league-arsene-wenger-election-labour

Corbyn sucks if he is still supporting Wenger - Bad decision!

[YL]

The more I look the more it becomes apparent that they've just 'plucked numbers out of thin air'. As an other example they have the Tories up 10% to 60% in Sittingbourne and Sheppy yet unchanged in Dartford which has similar demographics, similar Brexit vote and only a slightly smaller 2015 UKIP vote in 2015.

Even if you accept the 41-38 split (which I think is very wrong anyway) their constituency numbers would still be way off.

It's a statistical model and has a fairly large margin of error in each constituency, which it acknowledges.  Possibly that margin of error is too high for it to be useful (e.g. for my own constituency its range for the Labour vote is 27% to 45%) but it might explain many of your criticisms.  With 632 constituencies you're almost certain to get some rogue numbers somewhere anyway.

I think it has a definite Labour lean, but apart from that I'll wait until 9 June to judge it.

[Filuwaúrdjan]

He was on Newsnight the other day; seemed in pretty good nick.

And is now active on twitter!

[Filuwaúrdjan]

The thing about statistical seat model projects of that sort is that they're basically junk even if they end up being about right. I wouldn't pay them - whichever seat out this year we're discussing - much heed.

[Filuwaúrdjan]

Foot represented the old Devonport division (parts of which are in both present constituency) until he lost in 1959, Owen gained Sutton in 1966 and switched to Devonport in 1974 - which had not been won once since Foot's defeat, even in 1966 - when Plymouth gained a seat and neither the new Drake nor the redrawn Sutton looked particularly friendly. Plymouth has a very odd electoral history, but then that's often the way with naval towns. Presently Plymouth is split between three seats; Sutton & Devonport which covers most of its core, Moor View which covers the stunningly unremarkable northern suburbs, and S.W. Devon which covers the very middle class eastern suburbs (Plymstock etc) and a great swathe of commuter village hinterland.

[Shadows]

Jeremy Corbyn appeared on The One Show where he handed the presenters a pot of jam.


Besides offering jam, Corbyn also gave presenter Alex Jones a wink and spoke about his time working with polio victims in Jamaica. It marked a contrast with Theresa May and her husband Philip, who spoke about household chores and rubbish collection. The couple claimed there were ‘girl’ and ‘boy’ jobs in their household. The leader of the opposition shared pictures of his childhood as well as a famous image of him being arrested opposing apartheid.

http://metro.co.uk/2017/05/30/jeremy-corbyn-was-on-the-one-show-and-he-gave-them-a-pot-of-jam-6673430/


He was arrested for protesting apartheid - Wow!

[_]

Latest Tory poll leads:

YouGov: 3pt
Survation: 6pt
Kantar: 10pt
FT poll of polls: 9pt
ICM: 12pt
ComRes: 12pt
Panelbase: 15pt

Let it be noted that Kantar had the Libdems at 11%, and Survation had youth turnout at 82%, while ICM had it at 50%.

[Filuwaúrdjan]

Not totally sure this is true. Cleggmentum was only 7 years ago.

But there's no doubt that there really was a Centre surge during the middle of that campaign; the polling firms, however, were not prepared for this and so their models seem to have blown up...

And that's the thing this time: there's no doubt that Labour have made up significant ground during the campaign. How much? No one actually knows.

[Filuwaúrdjan]

Let's turn this into swings...

YouGov: 3pt
Survation: 6pt
Kantar: 10pt
FT poll of polls: 9pt
ICM: 12pt
ComRes: 12pt
Panelbase: 15pt

Panelbase: -4.0
ComRes: -2.5
ICM: -2.5
Kantar: -1.5
SurveyMonkey: -0.5
YouGov: 0.0
Survation: +0.5
Orb: +0.5
YouGov: +1.5

[Shadows]

BTW Yougov has a sample size of 50,000 vs 1K, 2K, 3K in other polls. So due to the huge sample size, it technically should be much MUCH more accurate than all others although constituency wise prediction is literally impossible ! I find the constituency prediction process pretty meaningless & useless, but the 50K sample gives a much better idea of overall picture !

Doug Rivers, YouGov's chief scientist, sets out how YouGov's 2017 General Election model works


Every day YouGov interviews approximately 7,000 panellists about their voting intentions in the 2017 General Election. Over the course of a week, data are collected from around 50,000 panellists. While this is a much larger sample than our usual polls, the samples in each of the 650 Parliamentary constituencies are too small (on average, only 75 voters per constituency per week) to produce reliable estimates.

Using data from the UK Office of National Statistics, the British Election Study, and past election results, YouGov has estimated the number of each type of voter in each constituency. It is important to understand the limitations of the model results. First, they are estimates of current voting intentions, not a forecast of how people will vote on 8 June. Second, the samples in each constituency are too small to be reliable by themselves and are subject to more than just sampling error.

https://yougov.co.uk/news/2017/05/31/how-yougov-model-2017-general-election-works/

[Filuwaúrdjan]

No evidence that giant polls are any better tbh.

[Filuwaúrdjan]

Regarding the Survation poll of Jewish voters... polls of specific minorities are problematic. Particularly when the minority is diverse, not very concentrated and quite small. I suspect the sort of figures Survation showed in 2015 and this survey are about right for people who are at least technically Orthodox, subscribe to the JC and are at least vaguely involved in community organisations (which is hardly a coincidence as unless their methodology has changed then that's basically the only people they're polling anyway). Not uninteresting,* but, as I said, obviously problematic as one of the main political dividing lines in British Jewry has always been Orthodox/Secular (note for Americans: in Britain the former are the majority, not a relatively small minority). Some attempts by other firms have been made in the past for a much broader sample, but I'd guess that'd either be crazy expensive or heading deep into voodoo territory. There have been similar issues with e.g. surveys of Sikhs etc. Not that there's much doubt that Corbyn-as-Leader is not great news for Labour candidates in certain constituencies.

_{*But in the same way that e.g. specific surveys of regular Anglican churchgoers - which are sometimes wheeled out during elections as well - are.}

[Shadows]

No evidence that giant polls are any better tbh.

True, it need not be. But 1/2K polls are usually meaningless. Most of this small sample size polls would even pass the 95% CL. With giant polls, you have a much larger sample size to analyze geographical, age wise pattern & so on. That is impossible in a smaller poll.

The last survation poll had a 50 odd sample size, that is meaningless & you have literally no idea from that ! For country wise issues, having 8-10K etc is pretty decent, perhaps 50K isn't needed (unless you want individual seat wise data which is hard even with 50K)!

[TheDeadFlagBlues]

Regarding the Survation poll of Jewish voters... polls of specific minorities are problematic. Particularly when the minority is diverse, not very concentrated and quite small. I suspect the sort of figures Survation showed in 2015 and this survey are about right for people who are at least technically Orthodox, subscribe to the JC and are at least vaguely involved in community organisations (which is hardly a coincidence as unless their methodology has changed then that's basically the only people they're polling anyway). Not uninteresting,* but, as I said, obviously problematic as one of the main political dividing lines in British Jewry has always been Orthodox/Secular (note for Americans: in Britain the former are the majority, not a relatively small minority). Some attempts by other firms have been made in the past for a much broader sample, but I'd guess that'd either be crazy expensive or heading deep into voodoo territory. There have been similar issues with e.g. surveys of Sikhs etc. Not that there's much doubt that Corbyn-as-Leader is not great news for Labour candidates in certain constituencies.

_{*But in the same way that e.g. specific surveys of regular Anglican churchgoers - which are sometimes wheeled out during elections as well - are.}

Interestingly, the JC released a survey of Jewish voters in 2010 that showed Labour with 30% of the Jewish vote. Presumably, this was a much stronger and robust survey that took more secular voters into account. It's unfortunate that JC dropped the ball here because it would have been very interesting to have some vague idea of how Corbyn has been received by Jewish Labour voters. Is he driving them away from the party or are they willing to hold their nose?

I guess we'll never really know because secular Jews aren't concentrated in particular neighborhoods in the manner that orthodox Jews are.

[bore]

No evidence that giant polls are any better tbh.

True, it need not be. But 1/2K polls are usually meaningless. Most of this small sample size polls would even pass the 95% CL. With giant polls, you have a much larger sample size to analyze geographical, age wise pattern & so on. That is impossible in a smaller poll.

The last survation poll had a 50 odd sample size, that is meaningless & you have literally no idea from that ! For country wise issues, having 8-10K etc is pretty decent, perhaps 50K isn't needed (unless you want individual seat wise data which is hard even with 50K)!

Apart from the fact that crosstabs aren't worth much for small polls, this is pretty much all wrong.

A 1000 sample poll has a margin of error of about +/- 3%, assuming that the sample is representative, which is more than enough to predict the results of the general election, roughly speaking. Because the decrease in the margin of error is logarithmic, there is not that much advantage to be had from having a sample much larger than this, unless you poll the whole country.

This is only true, however, if the sample is representative in the first place, which is the hard part of polling, and is just as much as an issue for 50,000 sample polls as it is for 1,000 polls. (Especially for yougov, as their samples are self selecting)