Real Poll Averages: UK General Election 2024
Excluding "Don't Know" is bad science - and lying about your data
If you didn’t know, the UK General Election has been scheduled for July 4.
Since poll aggregators lie in their published averages, and say the “poll average” contains 0 undecideds, I’ll be publishing the “real” poll averages here.
There is no gold standard for how to take a poll average - you’ll note I included a few different methods. But the “lying standard” that says there are 0 undecided is not an acceptable substitute.
From Election Polling UK
This is wrong.
Pollsters in the UK and non-US countries usually report their data with “undecideds/don’t know excluded.” Which is to say, they guess about what undecideds will do - and that guess is largely proportional to what “decideds” say they will do.
Those guesses, which are the shittiest possible forecasts, which they have the audacity (or, more likely, ignorance) to call “polls” continue to fail miserably…but they don’t care.
Not even a week after almost all of those pollcasts said Labour would win the London Mayoral contest by 20+ (in which they only won by 11) they were right back on their GE reports, with nonsense such as “Labour up by 30.”
It's baffling how these people are allowed to stay in business, and without extreme public backlash. This isn't the best we can do.
I’ll publish my forecast soon, which I will also update periodically. But I’ll tell my bettor (“punter”) friends who can do so legally: if you can find a sucker taker to give you even money or better on Labour winning by 20 or less - an outcome that would be largely unexpected given the “poll” data, remain calm, tell them “I couldn’t possibly wager more than a few thousand quid on that” and pray they try to talk you into betting more.
It’s possible Labour wins by more than 20, sure, even by more than 25. But if I were a betting man, I’d take the “under” for whatever amount my opposition was good for - and depending on our relationship, wouldn’t make them hate me.
UK GE 2024 “Real” Poll Average
This poll average does not mean “Labour up by 20” because that tells you nothing about what undecideds will eventually do.
To be extremely simplistic (because even experts don't understand this basic math): when “undecided” is as high as it is here, estimating how they will decide will make or break your forecast.
Assuming they will split evenly (as US analysts do) or proportionally to the decideds (as non-US analysts do) is terrible math, bad science, and why their forecasts will continue to fail.
It’s important to properly understand poll data: as a “base of support” before undecideds decide - and even before people change their mind.