General Election '24

[lpm]

It's understandable there's problems with the post in this election. The post only started in 1660, needs a few more centuries to bed in.

[El Pollo Diablo]

I've kept track of all the final polls this week, and have run seat numbers for every single one through electoral calculus and FT.

Here's the current state of play whilst there's a lull in the poll reporting. I'll update again later on once the final ones are in.

Here are the final polling intention numbers:

7-day Average
Con: 20.9%
Lab: 39.6%
Lib: 11.5%
Grn: 6.0%
Ref: 15.7%

Range of Support in 7 days (Max/Min)

GE24FinalWeek1.png (52.71 KiB) Viewed 6627 times

Seat Forecasts
EC in general has much more buoyant forecasts for Labour, and much worse for the Tories. Its methodology appears to have shifted since the start of this final week to downplay Reform seats predictions. FT are a little more believable.

I'm pasting the seat forecasts as a picture because I just can't be arsed to do it any other way.

GE24 Seat ForecastsPNG.PNG (86.44 KiB) Viewed 6627 times

[bob sterman]

Ready for Tory pundits on election night to describe anything less than a 200 seat majority as a failure for Labour.

[jimbob]

Ready for Tory pundits on election night to describe anything less than a 200 seat majority as a failure for Labour.

And then they could keep Sunak in as leader as he's obviously done well


That would almost be worth it for the lols

[lpm]

I've kept track of all the final polls this week, and have run seat numbers for every single one through electoral calculus and FT.

Here's the current state of play whilst there's a lull in the poll reporting. I'll update again later on once the final ones are in.

It is a masterpiece.

My guess from where we are now is about 250 majority, 80 Con seats.

Quite hard for the Cons to get to 150, based on your range of outcomes. Even 100 is a stretch.

[lpm]

And then they could keep Sunak in as leader as he's obviously done well

That would almost be worth it for the lols

Most lols is Tories win only one seat, and it is Sunak's, so he has to stay as leader.

If they have a disaster, who controls the Conservative Party? They have little offices everywhere and Conservative clubs and intangible assets of databases. And a Young Conservatives apparatus with a valuable collection of Nazi songs.

Presumably it can never die and can only fade away?

[dyqik]

Ready for Tory pundits on election night to describe anything less than a 200 seat majority as a failure for Labour.

And then they could keep Sunak in as leader as he's obviously done well


That would almost be worth it for the lols

This expectation management would explain the stories today about them expecting Sunak to lose his seat.

[IvanV]

I've kept track of all the final polls this week, and have run seat numbers for every single one through electoral calculus and FT.

Here's the current state of play whilst there's a lull in the poll reporting. I'll update again later on once the final ones are in.

It is a masterpiece.

My guess from where we are now is about 250 majority, 80 Con seats.

Quite hard for the Cons to get to 150, based on your range of outcomes. Even 100 is a stretch.

I have so much experience of being disappointed on election nights. I prefer to set a more modest benchmark as to what is an expected outcome, in the hope of being pleasantly surprised.

EPD preferred the FT model. 5 of the last 8 polls predict 120+ seats for the Conservatives if the FT model is applied to them. A while ago I was thinking 120 seats might be my benchmark. The polls seem to have moved a little in the Tories' favour since then. In this zone the number of seats is very sensitive to vote share. So I think I might retreat to 150. Only 1 of the last 8 is predicting more than that.

[bob sterman]

Ready for Tory pundits on election night to describe anything less than a 200 seat majority as a failure for Labour.

And then they could keep Sunak in as leader as he's obviously done well


That would almost be worth it for the lols

This expectation management would explain the stories today about them expecting Sunak to lose his seat.

So "PM retains seat after tight recount" will be a stunning victory and a damning verdict on Labour.

[Gfamily]

Do we want a separate thread for the results?
I have a Google sheet ready for the results if people want to fill them in as they come in.

[lpm]

I have so much experience of being disappointed on election nights. I prefer to set a more modest benchmark as to what is an expected outcome, in the hope of being pleasantly surprised.

That is of course irrational. We all do it, probably because we have inbuilt superstitions about jinxing it. Or magical thinking about having an effect on external forces.

All of us are toning down the MRPs. Emotionally, sure. But intellectually, why?

Normally we should trust experts, particularly when they have skin in the game and have invested their reputation on their forecast. Us amateurs shouldn't really come along and rewrite their work. Need to trust them to develop their model, follow it through and not make adjustments because the output seems too good to be true.

[lpm]

FocalData - their final MRP

Probabilistic count:
LAB 444
CON 108
LD 57
SNP 15
RFM 2
PC 2
GRN 1

Leading Count:
LAB 447
CON 105
LD 58
SNP 16
RFM 2
PC 2
GRN 1

No idea why their two counts are very close while Survation's had big differences.

[lpm]

YouGov - their final MRP

Labour: 431
Con: 102
Lib Dem: 72
SNP: 18
Reform: 3
Plaid: 3
Green: 2

Their range:

Labour: 391-466
Conservative: 78-129
Lib Dem: 57-87
SNP: 8-34
Reform 0-14
Plaid: 1-4
Green: 1-4

[lpm]

I suppose there are just far too many close contests to be able to predict.

There are plenty of narrow switches. For example Poole:

Survation: Labour 31-28
YouGov: Conservative 33-31

89 seats within 5 points, according to YouGov.

[TopBadger]

Starmer not helping himself by reminding us of his reticence to rejoin?

https://www.theguardian.com/politics/ar ... ys-starmer

Of all the economic damage the Tories created, he's not willing to consider undoing the most damaging aspect... really?

[IvanV]

I have so much experience of being disappointed on election nights. I prefer to set a more modest benchmark as to what is an expected outcome, in the hope of being pleasantly surprised.

That is of course irrational. We all do it, probably because we have inbuilt superstitions about jinxing it. Or magical thinking about having an effect on external forces.

All of us are toning down the MRPs. Emotionally, sure. But intellectually, why?

Normally we should trust experts, particularly when they have skin in the game and have invested their reputation on their forecast. Us amateurs shouldn't really come along and rewrite their work. Need to trust them to develop their model, follow it through and not make adjustments because the output seems too good to be true.

I'm deliberately trying to set my benchmark as something that is near one end of the experts' distribution of plausible outcomes. And I think I have done that.

For example, Electoral Calculus gives us their 95% confidence limits. For their displayed prediction today, those 95% limits are 22 to 157 for the Tories. So in fact, even on those facts, I have done what I set out to do. When I previously said 120, that was also close to that end of EC's 95% confidence limits at that time. So I have, in a sense, trusted those experts, and set my benchmark consistently on what they said, according to my chosen criteria.

But at the same time, there are assumptions behind that 95% number. And those assumptions may be wrong. We are all tending to have our suspicions about MRP methods, because we know how they work and we know the world isn't quite like that. Pollsters have an unfortunate history in this country of seeing outcomes outside their prediction confidence limits, for reasons that they could not have known about in advance. There are still potentially such issues here, that we can't know about in advance. There is no objective probability distribution for such unknowables. So, knowing there is an "unpredictability factor", that's another reason for choosing a conservative benchmark outcome at the low end of the prediction range. If my benchmark is breached, probably something odd that we couldn't predict happened. But that has happened quite often. And this is a very odd election.

Psephologists in the US have done very well of late, at least in presidential elections. I think Nate Silver called every state correct in the last one. He was a bit lucky because some states were very close. But there were no large errors. And that's because that election is determined with state level votes, and opinion polls at state level opinion polls exist. But he can't do so well here, because we don't have sufficient data at constituency level, where these things are determined.

[noggins]

Starmer not helping himself by reminding us of his reticence to rejoin?

https://www.theguardian.com/politics/ar ... ys-starmer

Of all the economic damage the Tories created, he's not willing to consider undoing the most damaging aspect... really?

Nobody is going to vote tory because starmer isnt pro-eu enough.

So, which tranche of voters is more electorally significant?

a) A few pious rejoiner twits who might vote libdem in a tory/labour 2 horse race constituency

b) brexit voting tablod readers

[lpm]

You know what I'm looking forward to most on election day?

How the entirety of UK political twitter stops exchanging informed commentary with each other.

And instead reminds each other to take their ID and go and vote. Then in the second half of the day retweets their reminders to take ID. Then at the end of the day tweets reminders to vote.

[El Pollo Diablo]

Thirteen polls so far today. We've still got, I think, probably three polls to come (Ipsos, Norstat and Findoutnow). But I need to make dinner and then clear the garage so I'll do an update now and reredo it later on again if I have to.

Final polling intention numbers:

7-day Average
Con: 21.2%
Lab: 39.5%
Lib: 11.4%
Grn: 6.2%
Ref: 15.8%

Range of Support in 7 days

GE24FinalWeek1.png (56.01 KiB) Viewed 6434 times

Seat Forecasts
I've now added vote share to these tables so people can get a better sense of what they're looking at.

GE24 Seat ForecastsPNG.PNG (80.51 KiB) Viewed 6434 times

The MRPs are much more bullish than the raw voting numbers, which is perhaps to be expected - they'll capture the tactical voting better, probably. Note as well that More in Common and YouGov have, during this election, been more generous to the Conservatives, and the numbers there are still in laughter-land, if not perhaps quite as totes hilares as we all wanted.

The Lib Dems have suffered in these last few days, not sure why. The Greens have picked up some support too - we might perhaps be looking at either the youth factor, or ex-Tories deciding they can't stomach Labour and switching to a protest vote, is my guess.

My prediction from these is that the actual vote for Labour won't be as starkly different to the polls as we've seen in the past, as it seems already built in. I'm going for a one-point drop.

[DJL]

I have so much experience of being disappointed on election nights. I prefer to set a more modest benchmark as to what is an expected outcome, in the hope of being pleasantly surprised.

That is of course irrational. We all do it, probably because we have inbuilt superstitions about jinxing it. Or magical thinking about having an effect on external forces.

All of us are toning down the MRPs. Emotionally, sure. But intellectually, why?

Normally we should trust experts, particularly when they have skin in the game and have invested their reputation on their forecast. Us amateurs shouldn't really come along and rewrite their work. Need to trust them to develop their model, follow it through and not make adjustments because the output seems too good to be true.

You're completely correct. But it's in our nature. I'm feeling like I did before birthdays when I was a kid and had been looking forward to receiving something specific for weeks, knowing that my parents had gone to the shop, but fearing that they'd mistakenly bought the wrong thing. It happened to me on my tenth birthday and I've never quite got over it.

So yeah, I'm hoping for a 200 seat Labour majority but I'd settle for 100.

[dyqik]

My guess is that the Green vote will come in a bit low, possibly to the benefit of the Lib Dems. I'm really not sure about Reform <-> Tories in polls vs. votes.

I'd also guess at a bit more tactical voting than usual - that's likely when the election is more about getting rid of someone than who's replacing them.

[TopBadger]

You know what I'm looking forward to most on election day?

How the entirety of UK political twitter stops exchanging informed commentary with each other.

And instead reminds each other to take their ID and go and vote. Then in the second half of the day retweets their reminders to take ID. Then at the end of the day tweets reminders to vote.

Hopefully they also remind them what 14 years of Tory government has achieved: https://www.youtube.com/watch?v=cKAeO-5saqQ

And how to vote to kick them out: https://stopthetories.vote/

[monkey]

Starmer not helping himself by reminding us of his reticence to rejoin?

https://www.theguardian.com/politics/ar ... ys-starmer

Of all the economic damage the Tories created, he's not willing to consider undoing the most damaging aspect... really?

Nobody is going to vote tory because starmer isnt pro-eu enough.

So, which tranche of voters is more electorally significant?

a) A few pious rejoiner twits who might vote libdem in a tory/labour 2 horse race constituency

b) brexit voting tablod readers

It's about so much more than just keeping a subset of voters happy. If it were just that, they could've probably got away with rejoining being policy.

Rejoining or Being Norway would have meant a rerun of 2019, but with both big leaders being unpopular and boring. The right wing press would not have been as nice to them if Labour were to say "Yay! let's do another referendum!", or "The single market's good, yo. Let's be Norway."*. This tanks Starmers already limited popularity, but increases Sunak's, because he's now the Good Guy. It would have led to Reform not standing candidates in Tory/Labour marginals. The press might have even ignored Ed Davy Ed Davying. Labour would not be looking forward to a stonking great big majority, they'd end with a much smaller one with a split party. Meanwhile everyone would be arguing about Brexit while their world collapses around them (Yay Fun times are back again!).

Might not be the right thing to do, but it is the obvious choice for a Labour party that has decided that being in power is the most important thing.


*I could write a manifesto. "I should be PM, 'cos I got bear skills, innit."

[nekomatic]

I haven’t checked this statement, but:

IMG_5882.jpeg (588.45 KiB) Viewed 6305 times

IMG_5883.jpeg (287.01 KiB) Viewed 6305 times

[dyqik]

The way the MRPs work is that they randomly assign wins to close seats (<5% margin?). That means that they will disagree about which close seats are "wins". And some of the remaining 11 will only be there by coincidence (how many MRPs are there?)

However, the polling errors are likely highly correlated between the seats. On the downside, that means that all the close seats going Tory is a reasonably likely outcome. On the lolside, it means that all of the close seats, including some of these 11, going against the Tories is just about as likely.