Today is Brexit Crossover Day. Supposedly enough old people have died since the referendum that blah blah. You can guess the rest. I have yet to read anything sensible about this.

Anyway, Kat, this is one for you. You probably have the data, you certainly have the expertise. If you have the time and the inclination are you able to repeat the analysis but with sensible assumptions?

Sure! If you still trust statistics
Possibly this already answers your question:
https://www.linkedin.com/feed/update/ur ... 2804326401

I've made that simulation after the Femi Oluwole, the leader of Our Choice, Our Future exclamation mark, almost stopped Brexit by saying:
"The under 55 population of the UK voted to Remain. So, in five years’ time by absolutely anyone’s maths, we have a population that voted to remain in the EU."

The simulation is rather conservative - since I myself voted Leave and was aware of the possible bias, I made some assumptions favourable for Remainers, I think: high turnout of the youth with the same high odds (otoh, even if that turnout was said to be unusually high, one might say that more of them would vote today...), and the sentiments don't change with age (while in reality people tend to become more conservative when they get older). My mortality forecasts (using my fancy-shmancy deep neural network model) turned out to be accurate and ONS released just one extra year of statistics compared to what I fed to that microsimulation, so I would say it's still up-to-date.

(I'm fighting with some scandalous text I got an opportunity to write for you long ago - I'm close to being almost happy with it )

I remember that now! I’d like to see it with sentiment changing with age. (It’s known colloquially as “growing up.”) E.g. keep % the same but adjust population. Could you do that?

If you want to bias things towards remainers, because of your own bias, you can always have the poor Leavers dying younger because they don’t have the same healthcare as rich remainers!

I'm back home at 9 pm at the earliest and will set up the simulations.
The turnout modelling seems to be the problematic too. I myself consider not voting in the second referendum - it's making fun of people and statistics.

I'm back home at 9 pm at the earliest and will set up the simulations.
The turnout modelling seems to be the problematic too. I myself consider not voting in the second referendum - it's making fun of people and statistics.

The result I am curious about in your model is the sensitivities: as sentiment swings between remain/leave for each individual age group,
which age group affects the total vote the most ? (don't know how many dynamic factors you have but I guess simple weighing by the size of each age group
will be the big determiner).

One aspect of the current dynamic is the pretty awful UK jobs headlines coming through at the moment (especially retail and auto)
and intuition suggests that weighs more heavily on the middle aged than youngsters or "no skin in the game" oldies.

The polls are as unreliable as ever (perhaps more so) but nonetheless a few turn to remain now.

Your crossover time is 2023 with no sentiment change to contrast with someone else's crossover of around now.
I am intrigued as to how sensitive is that to various sentiment shifts of key age groups?

Assuming the turnout and sentiments from 1/7/2016 (the micorosimulation has 3-month resolution, so this is the data which is the closest to the referendum) for the population demographics from 1/1/2019, I get 52% for Leave. (corrected, sorry - I previously copied the old result I rerun as a test)

It's a pity 6-year olds can't vote. All those stodgy 60+ers...

It's a pity 6-year olds can't vote. All those stodgy 60+ers...

The outcome could depend on what the different parties are "selling".

Assuming the turnout and sentiments from 1/7/2016 (which was the closest to the referendum date in the microsimulation) for the population demographics from 1/1/2019 (the micorosimulation has 3-month resolution, because of the women reproductive cycle / 3M postnatal infertility period), I get 50.34% for Leave.

tw, a quick answer for now, turnout is proportional to the group size, which is governed by the fertility and mortality trends. I calculated sensitivities for those variables as I used some deep NN modelling.  Apart from 40+ girls, who cannot decide whether they will start to mass-reproduce or join the trend of staying child-free, nothing interesting happens. Well, the mortality trends are rebounding, which is a big relief for insurers

The job headlines are in a way and to some degree taken into account by the migration trends (I assumed a "status-quo" scenario after the referendum - as if the Brexit didn't take place). Only British nationals can vote, though.

Interesting.  Thanks very much for looking at it. Thankfully the mortality and fertility data is not (yet) affected by losing access to vital medicines

Sir John Curtice was speaking on the radio today about recent polling results with an eye on a 2nd referendum.
The gist of what he seemed to be saying is that since 2016's 48/52 leave result, polls on recent averages have swung
to 55/45 remain. His explanation was based on a view that the non voters in 2016 are 2-1 in favour of remain.

That begs some interesting questions. Is there a breakeven turnout given the assumption of the 2-1 hypothesis? Is that split by age group?
Which other preference shifts in the data can move 48/52 against to 55/45 for?  I am trying to pull some demographic data to play around with those
scenarios.

Sorry, I've edited my previous post.

I don't know how they calculated the Brexit cross-over. Certainly not from these statistics: https://www.ons.gov.uk/peoplepopulation ... nagegroups

Re tw's questions, the sensitivity to turnout becomes interesting when you correlate it not just age, but also with political inclinations. There are many ways to model it, and it doesn't require microsimulations.

I calculated sensitivities and elasticities of major demographics to fertility and mortality forecasts, as well as time sensitivities of those two variables (I obtained their forecasts from deep NN models - for phun). Apart from girls after 40, who cannot decide whether they will start to mass-reproduce or join the trend of staying child-free, nothing interesting happens. Well, the mortality trends are rebounding (the behaviour is similar to Japan 1980' - see my old presentation, page 44), which is a big relief for insurers (serious people, not some City funds riff-raff, approved of the result). There are also many lot of interesting trends in mortality, which found sensible interpretations, but that's a different story (more interesting than Brexit!).

The job headlines were in a way taken into account in the migration trends (the post Brexit dates assume a "status-quo" scenario - as if the referendum and its consequences didn't impact the demographics).

@"Thankfully the mortality and fertility data is not (yet) affected by losing access to vital medicines"
I personally don't believe that drug shortages experienced by NHS have much to do with Brexit.

@ Leavers elasticity - when one can assume so little owing to the limited information, I think the best way is what politicians instinctively do: fight for the largest group of voters. (Only Tories don't get it and fawn a bunch of sick dimwits posing for "toffs" to tear apart animals, attack people and scrape penny selling ivory trash as "antique".)

It's a pity 6-year olds can't vote. All those stodgy 60+ers...

The outcome could depend on what the different parties are "selling".

I was killing at birth those children who wouldn't achieve the voting age on time, because they were irrelevant for Brexit.

I don't know how they calculated the Brexit cross-over. Certainly not from these statistics: https://www.ons.gov.uk/peoplepopulation ... nagegroups

they probably made it up because they want a second referendum

maybe they should hold a referendum on whether to hold a referendum

Assuming the turnout and sentiments from 1/7/2016 (the micorosimulation has 3-month resolution, so this is the data which is the closest to the referendum) for the population demographics from 1/1/2019, I get 52% for Leave. (corrected, sorry - I previously copied the old result I rerun as a test)
Brexitcrossover.png

The past decided about the future.