General Election 2024

[wordsfromlee]

I heard Sunak is also going to bring back pounds, shilling and pence as well as rickets. 
 

I found this little thread interesting too
 

 

[Devon Malcolm]

This National Service news is great for me personally as I can use it to threaten my 17yo next time he leaves breadcrumbs all over the chopping board.

[SuperBacon]

Dad's of the year clearly mate.

[Mr_Danger]

I’ve spent their entire lives forbidding them to go in to the armed forces, how am I gonna start using that to threaten them now? I’d be the Keir Starmer of dads.

[JNLister]

Only a few weeks till I write all about what happens with a hung parliament and it proves a ludicrously pointless exercise, but in the meantime, let’s talk opinion polls, how they work, and how to read them. The big thing since the last couple of elections is that we have two types of poll.

TL;DR: Don't pay much attention to individual polls, particularly exciting or surprising ones. Look at the averages and the trends. MRP polls can suggest how vote share might turn into seats but they aren't magic.

 

1)      Traditional opinion poll

 

How it works:

You ask a representative sample of around 1,000-2,000 people which party they are going to vote for. Simple as that, almost. The trick is how you make it representative.

Hard as it is to comprehend, 1,000 people is enough to get a good picture of the entire population. The best analogy is that it’s like tasting a spoonful of a saucepan full of soup after adding salt. As long as you’ve thoroughly stirred it, a spoonful is enough to know if it’s too salty or not salty enough. If you just chuck the salt in, you could drink a ladleful off the top and still not know.

With a poll, you just need to make sure your sample is representative. For example, you want it to be roughly half men and half women. Depending how deep you want to go, you’ll want to get the right proportion of age groups, ethnic background, class/occupation, education level and so on.

One way to do this is to simply carry on surveying people until you have the right number. Eg if you want a sample of 1,000 with a 50-50 split, you keep on asking people until you’ve got 500 women or men, then only ask the other group until you’ve got 500 of each. This can take longer and cost more.

The other way is weighting. Very simple example: you ask 1,000 people and 600 are men and 400 women. You then count each man as (500/600) 0.83 people and each woman as (500/400) 1.25 people when counting how many are voting for each group.

Reading the results:

Never get caught up too much in one poll. The way statistics works means that 19 times out of 20, a sample of 1,000 will produce a result that’s within three percentage points of the real figure (eg a poll result of 32 means the actual number is between 29 and 35). This is the “statistical margin of error.” The 20^th time is known as an outlier. (This term is often misused to mean “poll result I don’t like the look of.”)

Don’t pay too much attention to the figure in brackets that shows the change since last time that company did a poll. Firstly, you need to check when that last poll was. Secondly, if it’s a change up or down of 3% or less, there’s no way to know if the movement is real or its just the natural variation that comes with sampling.

Don’t pay much attention at all to anything other than the main voting figure. Numbers for how “18-30s” or “people in Scotland” aren’t useful when taken from a normal national opinion poll because the sample size for this specific is much smaller and not reliable representative.

Don’t assume that the figures in the last week of the campaign must be the most reliable because it’s nearly election day. You sometimes get a problem of herding” where polling companies worry about being wrong when everyone else it right, so they don’t publish a poll that gives a noticeable different result, meaning most published polls tend to have similar figures and appear to show more certainty.

The best advice is to look at poll averages. These vary a little in precise detail, but the general rule is to include the most recent poll from reputable companies, filter out anything older than a certain date (eg 30 days) and give more weight to more recent ones. The averages should give you the best picture of the general levels of the different parties and the trends over time. The FT has a good one behind a paywall while the best free one is probably:

https://news.sky.com/story/sky-news-poll-tracker-how-are-political-parties-performing-as-an-election-looms-12903488

 

Why do the results vary between different polling companies?

Part of it is just the fluke of the natural variation within the margin of error, but the two main variations are how you handle the problems of whether people will vote and what will happen with people who say don’t know. With the former problem you can:

• Assume everyone who responds will vote.
• Ask people how likely they are to vote and only count them if they are definite.
• Ask people how likely that are to vote (eg out of 10) and weight their response appropriately (eg if they say 7/10 you count them as 0.7 of a person).
• Weight people’s response based on historic patterns (eg you might weight a younger person’s vote less because you assume they are less likely to vote.)

 

With the problem of don’t knows you can:

• Ignore those people altogether in your results.
• Use follow-up questions (eg “if you had to vote” or “if you do vote” who would they pick)
• Make an assumption about who they’ll vote for based on, eg their age and gender, or who they say they voted for at the last election.

There’s also some variation between how polling companies ask about the parties, eg whether you include Reform/Greens in the list or you just say “Labour/Conservative/Lib Dem/Someone else) and then ask for the specific party if they say “Someone else”. It’s possible this affects the results.

Take all these variables, plus the different approaches to weighting, and you will naturally get variations in the results. The problem is you don’t know which approaches are most accurate until the election, and the best approach last time might not be the best one this time round (eg education level is now much more linked to voting choices than social class.)

 

2)      MRP

 

How it works:

Multilevel regression and post-stratification is its fancy name, and the big difference is that it doesn’t try to figure out the national vote share, but rather the actual winners in individual seats and thus the overall number of MPs for each party.

The first step is polling a much larger number of people, usually in the tens of thousands. This isn’t primarily about trying to get a bigger sample but about being able to get enough people to cover a wide range of demographic and other factors.

That’s because the MRP approach is all about building up patterns. For example, you figure out how much effect being female or being over 65 or having voted Leave in the EU referendum has on your planned party choice. You then take these factors, turn them into maths, and apply them to individual seats.

This can tackle a couple of problems with traditional polling. First, it might give you a better idea of how the national vote share translates to seats (eg whether the same number of votes nationally means Labour win lots of seats narrowly or wins some by massive majorities but fewer overall.)

Second it can uncover patterns you might not expect. For example, an MRP poll in 2017 forecast the hung parliament when that wasn’t expected. Part of that was because it forecast seats like Canterbury would (seemingly implausibly) go Labour because it turned out Corbyn’s platform went down well in places with lots of graduates.

How to read them/Why the results vary

MRP is just an approach. It’s not inherently magical. Whether it works well depends on the quality of your sample, the factors you choose to include, and how you do the bit I casually described as “turn them into maths.”

That means you can have good MRP polls and bad MRP polls. Best bet is to look for patterns in MRP projections. If there’s a consistent picture, they may be an insight into how the number of seats each party gets might differ from what you’d expect from their overall vote share.

[Ronnie]

Sunak's announcement of enforced volunteering for 18-year-olds has, perhaps predictably, not gone down well with his party's younger acolytes. The chairman of Birmingham Young Conservatives has just announced his resignation:

Fear not those of you worried that his interest in politics is over before it's begun: he's transferred to the Liberal Democrats.

[JNLister]

Following up on the polls post, a useful post by polling expert Mark Pack on whether the polls will be "wrong" this time (as in, be so off they indicate the wrong result).

https://theweekinpolls.substack.com/p/will-the-polls-get-the-general-election

TL;DR: there's no evidence any of the reasons polls have been wrong before apply this time. If they are really badly wrong, it's for a reason nobody can currently imagine.

[CharlesTuckerTheThird]

I want to try and remain optimistic, however this will be the third (fourth?) General Election since the Tories took power and we, as a nation, seem to either suffer from collective amnesia, collective masochism, collective stupidity, collective apathy or a combination thereof. I genuinely believe Rishi Sunak could announce a "Punch Every Toddler In The Face" policy tomorrow and there'd be more than a few people who'd vote Tory just to "oWn tEh LiBz" who'd be opposed to the idea.

[SuperBacon]

28 minutes ago, CharlesTuckerTheThird said:

"Punch Every Toddler In The Face"

If we're talking about other peoples kids then...

I'm very much enjoying this clusterfuck from them.

And how would an adult be responsible for the actions of another adult then? Clowns.

[wordsfromlee]

I’ve already seen kids discussing how they’re planning on getting out of their “Nashy Servs”

[SuperBacon]

2 minutes ago, wordsfromlee said:

I’ve already seen kids discussing how they’re planning on getting out of their “Nashy Servs”

Well they best hope the Tories don't win the genny lec then.

[Carbomb]

2 hours ago, Ronnie said:

Sunak's announcement of enforced volunteering for 18-year-olds has, perhaps predictably, not gone down well with his party's younger acolytes. The chairman of Birmingham Young Conservatives has just announced his resignation:

Fear not those of you worried that his interest in politics is over before it's begun: he's transferred to the Liberal Democrats.

God, what a little turd. Nothing but contempt for any younger person who continues to support the Tories at this point, let alone those who campaign for them.

That this was a straw too far means fuck-all; the thing that concerns me about his statement is the "rushed" reintroduction of national service. It comes across strongly, especially in the context of the gist of the letter, that he doesn't have a problem with the policy, but how it's been announced or planned.

[Thunderplex]

Don’t know if this has already been posted and really can’t be arsed looking, but it made me laugh.
 

[d-d-d-dAz]

It's probably worth saying, that a lot of European countries either have National Service or are looking to bring it in. Macron brought it into France a few years back.

So, the policy itself isn't necessarily far off crankery... it's just that, as ever with Sunak, the delivery was so poor as to be laughable, and conflating it with the security threats facing the country (just because he thinks there is fertile ground in attacking Starmer on it) made it one of the worst rolled out policies in recent memory.

I think in one of the Scandinavian countries, you're national service runs alongside your A Level studies (or equivalent activity in those years), for a set number of hours a week and you can choose to serve in the emergency services, military or in a variety of charities/third sector bodies. In return, teenagers get 'social credits' - or some such jargony word for 'voucher' - which can be used on further education or on tax rebates from work.

That's not a policy that spooks the horses, necessarily, and probably has some social benefits; but the Tories desire to appeal to the Reform lot made the whole thing an absolute shit soup.

My favourite development of the last few days is crafting an entire policy around catering to the 55+ crowd, but then deciding to attack Starmer for his age and call him 'Sir Sleepy'. Yeah, that'll endear the over 60's, telling them they're essentially redundant as people. Great stuff, lads.

Edited May 27 by d-d-d-dAz

[Keith Houchen]

10 minutes ago, d-d-d-dAz said:

I think in one of the Scandinavian countries, you're national service runs alongside your A Level studies (or equivalent activity in those years), for a set number of hours a week and you can choose to serve in the emergency services, military or in a variety of charities/third sector bodies. In return, teenagers get 'social credits' - or some such jargony word for 'voucher' - which can be used on further education or on tax rebates from work.

Here is a thread (sorry Chest) on Denmark, which has national service but, surprise surprise, has other things too.