United Kingdom General Elections: December 12th, 2019

[Filuwaúrdjan]

I think it's based on sizeable majorities insulating governments somewhat of which there's some political theory behind; 1987 helping Major in 1992, 2001 helping Labour in 2005 despite just a 3 point lead (and 2005 making 2010 harder for the Tories). This is the third (quick) election for the incumbent Tories and if Boris walks away with a majority of 100, it's probably not going to become a Labour majority of say 10 in one cycle.

There's a certain logic to the idea of mass-incumbency bonuses, sure, but I would say that it is more of an occasional tendency than a rule. Three figure majorities melted away at the first challenge in 1964 and 1970, for instance. A long time ago now, a literal lifetime away, yes, but the issue is the operation of this most obviously idiotic of electoral systems rather than direct comparison. And of course one only needs to look at what happened in Scotland in 2015 to see what can happen when the electorate has decisively changed its mind these days: if things turn, they turn. Party affinity and party loyalty at present are also so extremely low now that I wouldn't even be particularly surprised if a genuinely new party were to do randomly very well out of nowhere at some point.

[Filuwaúrdjan]

So far of the weekend glut we have had...

Panelbase: Con 42, Lab 32, LDem 14, BP 3, Greens 3. A change from a 13pt lead to a 10pt one on the week.
YouGov: Con 42, Lab 30, LDem 16, Greens 4, BP 3. No change on a poll they did midweek, but a change from an 18pt lead to a 12pt one on their last weekend poll.
Opinium: Con 47, Lab 28, LDem 12, BP 3. A change from a 16pt lead to a 19pt one on the week.

BMG: Con 41, Lab 28, LDem 18, Greens 5, BP 3. A change from an 8pt lead to 13pt point one on the week, but last week's BMG poll made no adjustment for the Brexit Party standing down in Conservative-held constituencies.

Deltapoll: Con 43, Lab 30, LDem 16, BP 3, Others . A change from a 15pt lead to a 13pt one on the week.

[Filuwaúrdjan]

So far of the weekend glut we have had...

Panelbase: Con 42, Lab 32, LDem 14, BP 3, Greens 3. A change from a 13pt lead to a 10pt one on the week.
YouGov: Con 42, Lab 30, LDem 16, Greens 4, BP 3. No change on a poll they did midweek, but a change from an 18pt lead to a 12pt one on their last weekend poll.
Opinium: Con 47, Lab 28, LDem 12, BP 3. A change from a 16pt lead to a 19pt one on the week.

BMG: Con 41, Lab 28, LDem 18, Greens 5, BP 3. A change from an 8pt lead to 13pt point one on the week, but last week's BMG poll made no adjustment for the Brexit Party standing down in Conservative-held constituencies.

Deltapoll: Con 43, Lab 30, LDem 16, BP 3, Others . A change from a 15pt lead to a 13pt one on the week.

ComRes: Con 42, Lab 32, LDem 15, BP 5, SNP 4, Greens 2. A change from an 8pt lead to a 10pt one on the week, but a change from an 11pt one to a 10pt one compared to a poll out midweek.

[Filuwaúrdjan]

Every election is different, but 2017 comparisons are often being distorted by false memory. A lot of people seem to recall that as an election in which the government started far, far ahead and in which its lead cracked and shrunk due to a sustained Labour surge that, bit by bit, took things to just below the wire on polling day. This is not what happened. What actually happened, is that both parties made substantial gains during the early weeks of the campaign as the third parties collapsed, but the Conservative lead remained stubbornly very high. Then, the Conservative manifesto was unveiled and it was a total disaster. Their lead immediately slumped, more than halving almost overnight in many polls. After that, things bounced around a bit during the last few weeks - there was some distinct movement towards Labour picked up after Corbyn's unexpectedly strong response to the terrorist atrocities, for instance, but that didn't all seem to last - and by polling day the general consensus was of a solid but not massive Conservative lead, with more firms giving them leads over 10pts than suggesting that it might be fairly close. This is why the exit poll was such a great shock to everyone.

The curious part is that we still don't know why the polls were so badly wrong in the end. We know what went wrong in 2015: a very tight race was expected, and so the firms all herded, afraid to put out anything that did not fit with that. But 2017 is a mystery: it is not as simple as 'higher than expected youth turnout' even if that must have been a factor. Conclusions to this? None, exactly. Other than to caution against drawing any conclusions from false memory.

[Filuwaúrdjan]

Survation's poll for Monday morning... Con 41, Lab 30, LDem 15, BP 5, Greens 3. Labour up two, Tories down one, LibDems up one.

[Filuwaúrdjan]

At the last election about 18% of votes were cast via postal ballot. Broadly speaking, postal voters skew older and more partisan. Political parties like their most reliable supporters to register for postal votes, because it means there's no chance that they'll accidentally miss casting a vote that way, though this is more a benefit in local elections than General Elections. Rates are particularly high in places where the Blair government had experiments for a time with postal voting only in local elections - Newcastle, for instance - but once those exceptions are ignored, rates are pretty similar across the country.

[Filuwaúrdjan]

One thing I noticed about that poll is the regional crosstabs show this vote in the South East
Conservative: 44%
Labour: 36%
Lib Dem: 15%

Seems like Labour is way to high there

ICM's 'South East' region includes London.

[Filuwaúrdjan]

What about Swinson - has Neil interviewed her already or did I just imagine it?

No, that inevitable fiasco (I mean they're all going to be trainwrecks, the only question would be the size...) is yet to come. She had a bad ordinary interview with him about a month ago, apparently, which would explain your confusion.

[Filuwaúrdjan]

Can we please not have partisan slanging matches in this thread?

[Filuwaúrdjan]

Turns out that the BBC have yet to schedule a feature-length Neil interview with Johnson. It appears that everyone else signed up and were allotted slots on the assumption that everyone was one board. I don't like Corbyn and I don't like feeling bad for him,* but this is really serious, a potential breach of the BBC's commitment to balance and political impartiality. Negotiations are supposedly ongoing, but that is just not good enough: they now have to set a date and empty-chair Johnson if he doesn't show up. This is very disturbing.

_{*I also think Neil's style of interview does not do our democracy any favours.}

[Filuwaúrdjan]

Another ComRes: Con 41, Lab 34, LDem 13, BP 5, Others 7

_{Fieldwork 25th and 26th.}

[Filuwaúrdjan]

Remember that all this is is a demographic modelling tool that projects a large national survey down to constituency level. It relies heavily on a series of assumptions, and if any are even a little bit out then the projections will be wildly off. It's clever, but no more inherently accurate than doing a quick swing calculation and checking out the constituency pendulum.

Or, rather... everyone remembers that the model called Canterbury correctly last time. Everyone forgets its projected figures for Finchley & Golders Green...

[Filuwaúrdjan]

They are all clearly putting a lot of weight on projected referendum results for each constituency, I think more so than last time. This may well be right, it may well not be.

[Filuwaúrdjan]

I see that my constituency is in the news! I do hope it is for something charming and/or edifying and not...

Oh dear.

[Filuwaúrdjan]

Is anyone keeping a running total of candidates suspended by their parties for antisemitism? Anyway, for the first time it has happened to a candidate with a realistic chance of actually winning, the SNP candidate at Kirkaldy & Cowdenbeath.

[Filuwaúrdjan]

Campaigning has been suspended because of the terrorist attack.

Also, the first of the weekend poll glut has arrived. Panelbase: Con 42, Lab 34, LDem 13, BP 4, Greens 3

^{Plus two to Labour, plus one for the Greens and Brexit Party, minus one for the LibDems. Implied swing of 3pts to Con since 2017. Fieldwork on the 27th and 28th.}

[Filuwaúrdjan]

Hm. This one shows a different story.
https://mobile.twitter.com/ElectionMapsUK/status/1200852564457205760

ComRes put out two polls a week at present rather than one - those figures are compared with the midweek poll. Compared with their last weekend poll there's no shift in the lead. In general they have been (a bit bizarrely given the reputation they used to have) one of the more stable pollsters this time round.

[Filuwaúrdjan]

The biggest difference between ComRes and BMG is probably the polling dates.

Yes, the fact that the main commissioners of polling during an election are the Sunday papers is... problematic. Note that Panelbase polls at the same time as the Sunday glut but releases on Saturday. Doesn't mean their findings are older.

[Filuwaúrdjan]

I don't think that does show them going different ways. That would be a 5pt swing towards the Tories nationwide and a 2.5pt swing in London.

And, in any case, they are by different polling firms...

[Filuwaúrdjan]

Is this a last minute Labour surge or not?

Well it isn't last minute yet - over a week to go still, and that's quite a length of time in a British campaign, as hard to comprehend as that often seems to be for Americans. There has been some clear movement towards Labour over the past couple of weeks, and for the most part the implied swing in the polls is also somewhat lower as well. Whether you can call that a 'surge' or not depends on how dramatic you like your language.

There was, incidentally, no last minute polling surge for Labour in 2017. A couple of polls hinted in what turned out to be the correct direction, but they were in a minority (a small minority) and were not believed. For the most part the polls were just wrong.

[Filuwaúrdjan]

It isn't entirely clear (and I suppose now never will be) quite what went wrong in 2017, so adjusting for whatever that was isn't really possible.* Adjustments, though, are made all the time, usually to fit in with whatever is presumed to be the reality at the present moment. The British polling industry is really not very good and when it gets things about right it is mostly by accident.

_{*Although ICM no longer make the structural pro-Conservative adjustment that caused them so much embarrassment that year.}

[Filuwaúrdjan]

I'm really not convinced that fussing excessively over subsamples of subsamples is a particularly good use of everyone's time.

[Filuwaúrdjan]

Supposedly this is the product of showing people a "full choice" of options.

But in many seats "others" are not standing at all, and in many places where they do they will poll derisory votes. It genuinely makes little sense at all.....

Excellent example of why it's best to view the British polling industry with general contempt and to take its findings (any of them) with a degree of caution. Making a major methodological change with a week to go... Christ.

[Filuwaúrdjan]

I'm not sure what Dennis Skinner is actually like as a constituency MP in all fairness.

Old fashioned. Not ineffectual (he's done quite a bit of good for it), but not the the sort to open every fête, kiss every baby and to write angry letters about every pothole and dog turd. Though the big mystery there is whether the 2017 figures represent the actual starting point or not - the situation there last time (with the constituency being intensively and intensely targeted on the one hand and with the CLP turning out not to have canvassing data etc. on the other) was rather unusual.

[Filuwaúrdjan]

Tiny sample size is never great news, but the doubling of the Plaid percentage is a big flashing light there. Not that that particular constituency isn't vulnerable, particularly with Lucas standing down.