United Kingdom General Elections: December 12th, 2019

[Oryxslayer]

Anyway, The Guardian failed to produce nice charts like last time, so here's the three London Constituency Polls I mentioned. Link to the relevant Guardian piece.


Hendon (Barnet):

Con: 51% (+3)
Lab: 33% (-13)
Lib: 12% (+8)
Grn: 1% (-)

If only Lab/Lib had a change of winning the seat:

Con: 55% / 53%
Lab: 41% / 7%
Lib: 4% / 39%
Grn: 0% / 1%

Cities of London and Westminster:

Con: 39% (-8)
Lib: 33% (+22)
Lab: 26% (-12)
Grn: 1% (-1)

If only Lab/Lib had a change of winning the seat:

Con: 49% / 42%
Lab: 44% / 5%
Lib: 7% / 51%
Grn: 0% / 1%

Chelsea & Fulham:

Con: 48% (-5)
Lib: 25% (+14)
Lab: 24% (-9)

If only Lab/Lib had a change of winning the seat:

Con: 57% / 49%
Lib: 7% / 43%
Lab: 35% / 8%

Obvious disclaimer about constituency polls is obvious. Changes are  with the 2017 results.

[Oryxslayer]

The Lib Dems have had a real problem of strong support not being translated into seats under FPTP. It's no surprise that they're big electoral reform advocates.

Also seems that they underperform in every campaign recently (2010 is more of a mixed bag though where they overperformed expectations at the start of the campaign while underperforming end of campaign expectations)

LDs have kinda a weird situation. They always underperform their projected vote, because their voters are on average more likely to be white-collar, educated, politically-attuned, and be at least stable in their situation. This leads to a voter base that if presented with a seat where it is obvious the LDs stand no chance, the LD voter will be more likely to cast a red or blue vote depending on his/her opinions. On the other hand, the LDs will outperform the number of seats they should be getting for said percentage. This is because the LD strategy is to narrow in on targets with a greater propensity to flip orange. This often makes LD swing impossible to calculate since they could potentially have a 'latent' voter base in a seat that will be activated by campaign resources. A LD incumbent in this regard is a powerful resource. I personally have the LDs a lot higher in my current 'excel prediction' than most models, because of these historical trends. Mostly this is because Wealthy West London seems poised to be 'activated' and go into strategic voting mode.

[Oryxslayer]

Everyone has a different 'story' from 2017, so paralleling back will likely result in comparisons which should not be. For example, my personal experience of 2017 was a Tory majority prediction that gradually shrank to a very slim one by election day. This is because my family has worked for YouGov's nonpolitical polling division before, and when the other teams put out that model we knew it was the real deal.

In the end, the polling industry is just trying to fight today's battles, but their knowledge, like us, is faulty. Pollsters always need to get it right, so corrections are always made with the knowledge of history. Said corrections often fail because yesterday is not today. In 2010 it was the LDs getting overshot in seats, in 2015 it was the Torys getting underpolled, in 2016 it was Leave, in 2017 it was Labour. Corrections for past mistakes are usually overcorrections. In the US, this leads to House effects swinging with each election, because there are only two camps to poll. In multiparty systems there are multiple pillars and communities making everything that much more uncertain. What weights are applied and how heavy they need to be for accuracy is a question that keeps pollsters up at night.

[Oryxslayer]

Anyway, the Tory Manifesto dropped. Like I thought, it's relatively benign as far as right-wing party platforms go. I'm sure the resident Labourite's are going to disagree, but that's my take. Fairly center-right unlike May's red-meat platform for the base. No cuts, some spending (far below labour), and the Key plank of Brexit. They clearly are afraid of what happened in 2017 and wish to capitalize on Corbyn's 'radical' platform. However, it's relative moderation is skewed by the entire thing being held up with the key proposal of "Get Brexit Done." So moderate but harsh on Brexit, just what BoJo wants to win Leave seats.

BBC Analysis.

Regarding Scotland, the Manifesto warns of the  "coalition of chaos" and recommits the party to Unionism.


I guess once the document has settled in with the rest of them we will see if something found it's way through that can become 2019s 'dementia tax.'

[Oryxslayer]

I think it is right to say that the worlds where the Lib-Dems make huge gains in the 'slice' of West London and the world where the Lib-Dems remain in the teens are mutually exclusive. I have already mentioned before how the Lib-Dems seat totals are hard to predict because of 'activation' and how if their voters are not activated by a targeted Lib-Dem campaign then they scatter. With this in mind, I m going to say that right now my personal model/prediction has the Lib-Dems in the mid-30s, which is far higher than anyone else. This however is all because of that wealthy slice  of London I keep going on about. Outside of London the Lib-Dems aren't doing that much better than the usual models, and all their gains are coming from the usual suspects like Winchester, Colchester, Sheff-Hallam, and St. Albans. Perhaps it's closer to their 'high-end' outside of London, but this is more to square with what is going on inside the city. Currently I am saying that the Lib-Dems walk away with more than 10 seats from London, mostly from the Tories but some from Labour. Together, the Lib-Dems and Labour push the conservatives to their 'bleed-over' seats along the edges of the city that hold more similarities to the SE/E than the city.

Now, there is a reason why I always perk up at the mention of this 'slice,' why I never shirk from discussions concerning it's vote, why I keep posting the constituency polls from the region, and why I have it all going Orange. In my eyes, wealthy West London is going to not only determine the Lib-Dems future as a party, it's the key to this entire election. So lets tackle each of those:

Lib Dems Future: Since the demise of the coalition the Lib-Dems have lacked any seats they can truly call their 'base.' Old strongholds in the celtic fringe isn't coming back in the numbers it did before. The Lib-Dems, both locally and nationally have therefore been a party of locals and targeted issues. This isn't sustainable, especially if polarizing culture wars issues are to dominate the entire island. Leafy West London offers the Lib-Dems a launchpad towards a future base and a future niche. The area is too wealthy for Corbyn's populism, but too Remain for BoJo's Leave. This all seems like a natural fit for the Lib-Dems, and it can grow into a base if the Tories keep taking this nationalistic approach to politics in the future.

Key to the Election: Right now, constituency polls from the region are exactly what the Lib-Dems are hoping for. Throwing aside the uncertain topline, we see that a majority of voters, nearly all Llib-Dems and Labourites, along with some Tories, are primed to vote tactically. Fairly understandable considering the demographics of the region match with those demographics more attuned to the political winds. This means that once the Lib-Dems seriously get in there we may see voters get activated and seats start flipping. If you think West London is going to have a lot of Orange, than Boris's path gets that much harder. He loses 5-8 seats right there. If Corbyn ends up turning this thing around (a harder question to answer) than BoJo cannot afford to just write these off. If he is on track for a small majority, which is more likely right now, than losing these seats mean replacements need to be found in Wales and the N/NE. If you think the Lib-Dems have the ability to swipe a bunch of seats off the board for both parties than the other parties will need to work harder to get to a dominant position. Therefore, a vote in the 'Slice'  is likely to be the most powerful vote in the country, ahead of the Scottish seats and the universal-swing seats between Labour and the Conservatives.

[Oryxslayer]

New welsh barometer poll from YouGov/ITV.

This is a gain of nine for Labour since early November, a gain of four for the Tories, down one for PC, down three for Lib Dems, down two for Greens, and down seven for Brexit. last time Labour led the barometer by 1 overall.

Now time for the downside for Labour: wales is one of the places where the Tories are far more vote efficient at the outset. Losing 11 points from 2017 will disproportionately effect those seats outside of the southern valley, so the combined opposition probably gains more than universal swing suggests. The welsh barometer from the same time period (two weeks out) of the 2017 campaign is presented below, subsequent barometers had Labour gain a tiny bit more:



The barometer was very accurate in 2017.

[Oryxslayer]

https://twitter.com/PaulBrandITV/status/1198894041510948865

FTR, that fact that we are arguing about Labour and Lib-Dem proposals probably means that 2017 won't be repeated as far as 'dementia tax' and 'police cuts' are concerned. This unfortunately means the Tories likely win this media cycle, since May set the bar incredibly low manifesto-wise.

[Oryxslayer]

I saw the hype rise up on twitter. Everyone needs to watch Neil eviscerate Sturgeon over the entire interview. Unfortunately, the SNP will survive because they got partisan loyalists these days who will stand firm with their identity of Scottish nationalism.

[Oryxslayer]

YouGov's releasing a similar model to their 2017 one supposedly on Wednesday. Last time their model came out a week out from the poll, this time  it will be two weeks. Therefore, there will likely be less certainty around the results and we should expect some change similar to a 538 live model, whereas the YouGov model from last time always had a 4% Tory lead.

Now, the model wasn't as perfect as hyped to be, but it was good. I have their last editions data downloaded for an GIS analysis when this 2019 version drops. One thing that does seem to be a consistency in their previous model though is that they missed the LD (and some unionist in Scotland) tactical voting effect that we have discussed at length here.

[Oryxslayer]

I have a lot to say on Antisemitism, Islamophobia, and other scapegoating within various European parties, but since nobody has the audacity to deny that this is an issue for Corybn and Labour then I think I will hold my tongue.

Anyway,

 https://twitter.com/EuropeElects/status/1199302074330206209

On one hand, it's a big Labour gain. On the other  hand, said gains just bring the poll into line with the rest of the pack, so it may just be herding or weighting for Brexit voters.

[Oryxslayer]

YouGov poll remains with the  pack.

https://twitter.com/ElectionMapsUK/status/1199372537836253184

[Oryxslayer]

Some questions about postal voting. About how much of the electorate is expected to vote by mail, and are ballots already being mailed in?

As long as royal mail isn't on strike (there were worries when the election was called) postal voting will likely be up this cycle. Short daylight hours and cold weather are  going to dissuade more potential voters than usual from going to the polls in a traditional fashion. The campaign for those voters who cast ballots early ends when that ballot is in the mail or handed in at your count, so you have less time to campaign for them. Despite popular belief, it isn't just those who have made up their mind voting early. The deadline expired an hour ago for applying for a postal ballot, I'm not sure if they put out numbers. However, if that number is sufficiently large, Labour may have less time to win over voters than it appears.

[Oryxslayer]

So Corbyn had a horrible interview with Neil, though it wasn't as well roasted as Sturgeons. I can't wait for the Boris interview, since it appears Neil will bat three for three.

[Oryxslayer]

Okay, so since YouGov should be releasing a 2019 version of their very accurate 2017 model, I figured I will do a breakdown of their previous model. I know it’s late and most actual residents will be asleep, but I need to get this out in case YouGov’s model is released while I am asleep. There will be quite a lot of people, very likely some who are prominent, who will claim this is the word of god, and that we already know the end result. It is therefore prudent for me to look at what parts of the model we should trust, and what parts we should treat with as much skepticism as other election models.

Firstly, YouGov did not include Northern Ireland in their model last time around. If they do, it will hurt their model since Northern Irish politics do not necessarily conform to swing. Sectarian politics mean there is little swing between blocks, and it is all personalistic turnout. If they do add in a projection for Northern Ireland, we should therefore treat it with appropriate skepticism. From here forward, we are only going to concern ourselves with the 632 seats in Britain.

Including the Speaker, YouGov’s topline numbers last time around were 42% Conservative and 38% Labour, with 304 seats for team blue and 269 for Corbyn. If we only focus on this topline, then the model is exceedingly accurate, within the appropriate Margin of Error. They actually undershot Labour but 2% in the popular vote, but undershot the Conservatives in seats. Most of the errors between the big two can be explained by Scotland, which I will get to in a moment. If there was a house effect in the 2017 numbers, if projected Labour to be more vote efficient in England than it actually was. The topline is trustworthy, no matter how much it ends up differing or conforming to our herding expectations.

Going into the weeds a bit, YouGov got many of the most shocking calls of the night right in their seat-by-seat projections: Darlington staying red, Canterbury flipping to Labour, and Kensington throwing off the Tories. However, they also missed a few Labour-Tory calls, which can be seen below. The way how YouGov’s model worked on a seat by seat basis is they gave each party a ‘high’ and a ‘low’ forming an MOE in each seat. Seats were not really outside of their MOE, even when they were miscalls for the party. Therefore, we can conclude that the ‘secret sauce’ used between Labour and the Conservatives is going to have the final 2019 results somewhere within their MOE.


Seats that the  YouGov model projected incorrectly, colored for incorrect winner

It’s here where the model gets a bit less trustworthy. Focusing on English and welsh third parties as a whole the model fails to capture the level their vote was concentrated. This is understandable even when your model has a ton of respondents and data behind it. There are less PC and LIB voters numerically. It’s clear that those parties powered by tactical voting, localist campaigns, and concentrated voter activation are hard to model. The model gave the Lib-Dems 8 seats (one of which is Ceredigion so it’s also wrong), 2 seats for the PCs, and had Claire Wright win East Devon. For example, most models will decry the fact that Finchley and Golders Green can go Lib-Dem after returning less than 10% for team orange in 2017, but I don’t think anyone will be surprised if the seat has a 40% swing in two weeks.


Scottish Seats that the  YouGov model projected incorrectly, colored for incorrect winner

Finally, there is the peculiarity of Scotland. While labour on its own seems like the most underpolled, getting projected at one seat but winning seven, there is a general trend of the unionists overall getting polled at less seat than they should. On some level, this is similar to the Liberal Democrats getting undercounted when it comes to seats – tactical voting isn’t captured by models. Concentrated unionist strength flipped more seats than even weighted swing projected. For two cycles now Scotland has had a political trajectory unique to herself, and we should therefore not count out for tactical voting between the four in some fashion.

In Conclusion, respect the Red-Blue topline, be rationally skeptical of the minors and Scottish seats.

[Oryxslayer]

Hard Data on Mistaken Seats

Median Projected/Actual

(Many other actual margins were not near the projected average, but said errors did not result in seat flips)

Bolton West: 44.7% Labour, 43.9% Conservative / 47.9% Conservative, 46.1% Labour

Broxtowe: 44.8% Labour, 44% Conservative / 46.8% Conservative 45.3% Labour

Calder Valley: 46.2% Labour, 42.5% Conservative / 46.1% Conservative, 45.1% Labour

Carlisle: 47.1% Labour, 44.7% Conservative / 49.9% Conservative, 43.8% Labour

Copeland: 47.6% Labour, 42.7% Conservative / 49.1% Conservative, 43.1% Labour

Corby: 45.9% Labour, 44.1% Conservative / 49.2% Conservative, 44.7% Labour

Crewe and Nantwich: 44.8% Conservative, 44.5% Labour / 47.1% Labour, 47% Conservative

Finchley and Golders Green: 48.1% Labour, 37.5% Conservative / 47% Conservative, 43.8% Labour

Halifax: 45.3% Conservative, 44.6% Labour / 52.8%  Labour, 41.7% Conservative

Hastings & Rye: 48.3% Labour, 39.9% Conservative / 46.9% Conservative, 46.2% Labour

Hendon: 46.3% Labour, 40% Conservative / 48% Conservative, 46% Labour

North East Derbyshire: 45.6% Labour, 42.7% Conservative / 49.2% Conservative, 43.5% Labour

Peterborough: 46.6% Conservative, 45.6% Labour / 48.1% Labour, 46.6% Conservative

Portsmouth South: 35.9% Conservative, 35.4% Labour, 20.8% Lib-Dem / 41% Labour, 37.5% Conservative, 17.3% Lib-Dem

Pudsey: 47.4% Labour, 45.4% Conservative / 47.4% Conservative, 46.7% Labour

Preseli Pembrokeshire: 43.4% Labour, 39.8% Conservative / 43.4% Conservative, 42.6% Labour

Southampton Itchen: 46.5% Labour, 43.9% Conservative / 46.5% Conservative 46.5% Labour.

South Swindon: 48.8% Labour, 39% Conservative / 48.4% Conservative, 43.6% Labour

Stockton South: 45% Conservative, 44.4% Labour / 48.5% Labour, 46.9% Conservative

Thurrock: 45.4% Labour, 37% Conservative, 15.1% UKIP / 39.5% Conservative, 38.8% Labour, 20.1% UKIP

Walsall North: 44.5% Labour, 43.1% Conservative / 49.6% Conservative, 42.8% Labour

Third Party Seats

Arfon: 40.9% Labour, 36.8% Plaid, 20.7% Conservative / 40.8% Plaid, 40.5% Labour, 16.4% Conservative

Carshalton and Wallington: 40.8% Conservative, 36.2% Lib-Dem, 19.6% Labour / 41% Lib-Dem, 38.3% Conservative, 18.4% Labour

Ceredigion: 32.9% Lib-Dem, 22.1% Labour, 19.7% Plaid, 18.4% Conservative / 29.2% Plaid, 29% Lib-Dem, 20.2% Labour, 18.4% Conservative

Eastbourne: 43.5% Conservative, 41.2% Lib-Dem, 13.8% Labour / 46.9% Lib-Dem, 44.1% Conservative, 8.1% Labour

East Devon: 48.4% Indie, 34% Conservative / 48.5% Conservative, 35.2% Indie

North Norflok: 41.7% Conservative, 40.9% Lib-Dem, 17.4% Labour / 48.4% Lib-Dem, 41.7% Conservative, 9.9% Labour

Oxford West and Abingdon: 41.2% Conservative, 39.5% Lib-Dem, 16.7% Labour / 43.8% Lib-Dem, 42.4% Conservative, 12.6% Labour


Scottish Seats

(Scottish Politics is too hectic for percentages like the above seats)

Aberdeen South: SNP / Conservative

Angus: SNP / Conservative

Ayr, Carrick and Cumnock: SNP / Conservative

Banff and Buchan: SNP / Conservative

Caithness, Sutherland and Easter Ross: SNP / Lib-Dem

Coatbridge, Chryston and Bellshill: SNP / Labour

East Lothian: SNP / Labour

Glasgow North East: SNP / Labour

Gordon: SNP / Conservative

Kirkcaldy and Cowdenbeath: SNP / Labour

Midlothian: SNP / Labour

Perth and North Perthshire: Conservative / SNP

Rutherglen and Hamilton West: SNP / Labour

Stirling: SNP / Conservative

[Oryxslayer]

https://twitter.com/lewis_goodall/status/1199381314639675398

Fascinating thread on Tory prospects in the Home Counties. I'm not sure I think there's any prospects for more than one or two seat flips, but it certainly backs up what cp has been saying about his own seat on this forum. Perhaps we're in for an election where the Tories win a bunch of Labour heartlands in the North but lose Esher? Probably not, but interesting nonetheless.

Heres my list of Realistically Possible (aka anything less than 99% chance of Con victory) Tory losses in the SW right now:

Wokingham, Reading West, Guildford, E&W, Wycombe, and Isle of Wight, Lewes, Hastings & Rye, Southampton Itchen, Winchester, Crawley, Spelthorne, Witney, Wantage, and the Milton Keynes.

That isn't much in one of the largest 'regions' of the UK. The problem for the opposition is that the Tories have large majorities, and if the Lib-Dems don't pour resources into every seat they will not see voters get activated. The majorities needed to be seriously reduced in a previous election, or the Lib-Dems needed to catch fire and be able to run a national not a targeted campaign. At the start of this campaign there was a serious chance of the SW holding a bunch of Tory loses, (I once mentioned the data was there for May to have a portillo moment if invested against) but the parties don't seem to want to serious campaign for Tory Remainers in the SW, outside of targeted strongholds.

[Oryxslayer]

Yes, if you wish to open a discussion on Corbyns 'leaks' (I mean they were just sitting on Reddit) it's the perfect topic for the international discussion board. We're here to discuss their impact electorally. We'll have to just wait and see for the long term impact, but the fact the documents were provided to the press as well may lead to "Corbyn said this but it actually says that." Corbyn after all has an interest in making this potential gun smoke as much as possible.

[Oryxslayer]

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.}

I agree that Andrew Neil trashing each leader in turn does no favours for democracy or journalism, but Sturgeon and Corbyn have did their turn, taken the heat including setting up narratives for the press and now it appears Boris and Swinson might not do it. As you say, that's going to have big consequences for the BBC if they hadn't actually got agreement for all four.

It's all very helpful for the new ONLINE tranche of Boris supporters who probably will love the fact he's dodged a bullet and 'bested' the BBC.

I mean, it didn't have to be be Boris. If any leader hadn't signed up to an interview ahead of time and watched Neil destroy everyone  in front of him, you would be running for the hills. The loss is inevitable, you would want to minimize it if possible. This is one of the very few times where pulling out is the  smart play, it's less cowardice rather than self-preservation. I'm sure the BBC can find something that isn't nice for BoJo or Swinson to fill their timeslots if they duck.

[Oryxslayer]

So this YouGov model is being released in 80 minutes or so... given we've seen the Best For Britain model as well today which, to put it politely, is a complete crock of sh!t, is there much we should be looking at with this? It'll obviously be better than that and it was relatively accurate last time iirc, but I'm not sure anyone should be drawing the conclusions from it that they inevitably will.

My post last page was literally about what conclusions can be drawn. The thing about MRP models is that they are far easier to screw up than regular polls, because a bad weight on one group can upset everything. Therefore, we should probably only trust (somewhat) the initial MRP poll and give all these  new models that are try out the hot new thing appropriate skepticism.

[Oryxslayer]

Remember what I said about the Scottish seats and Third parties  people. -_- However, that Blue-Red number is believable and likely.

[Oryxslayer]

https://yougov.co.uk/uk-general-election-2019/ official site.

[Oryxslayer]

Easily visible Map format. Remember that third parties and tactical voters like the Unionists or Lib-Dems got underpolled last time.

[Oryxslayer]

Easily visible Map format. Remember that third parties and tactical voters like the Unionists or Lib-Dems got underpolled last time.

Great maps thanks! Just to let you know Brecon & Radnorshire should be Tory, the Lib Dems are quite a way behind in this model.

That was a value call. Last time, they seriously missed the other By-Election flip in Copeland, saying that it would return rather nicely to it's former owner because the general data is based  off of the  last election. So, if there was a color for 'weird' that would be it.

[Oryxslayer]

What is that one area on the southeast tip of England that is labour?

Canterbury. University, remain, the whole shebang. Very different from rest of Kent, which is more  like the East of England politically. Labour picked it up last time slimly, after a near 100 years of tories, a length that was extended by LD/Lab splits in the town center. Frankly, I'm surprised Stroud isn't a Lab hold, since it was  in a slightly similar situation last cycle.

[Oryxslayer]

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...

Something to keep in mind also: this poll is a snapshot of time and things may move quickly over the next week. If there's a swing back to Labour (or away from Labour), it may not be uniform. Relying on MRP to think about the election is seductive for political geography enthusiasts, we all want to see detailed poll results, but you can use common sense and imagine similar predictions less the absurd outliers that no one should anticipate, even if they excite us on some level. Calm down, use your head folks!

The fact YouGov released this model 1 week earlier than the comparative point in 2017 though suggests that they will be seriously updating it a la 538 in the coming weeks. So if there is change, then we will see  it.