Jonn Elledge: Okay, so this thing where whatever the polling industry does, it seems, they never quite get Labour's voteshare right: Is it possible there's a sort of Heisenberg's Uncertainty Principle thing? The very act of measuring Labour's vote share moves the polls?

Okay, so this thing where whatever the polling industry does, it seems, they never quite get Labour's voteshare right: Is it possible there's a sort of Heisenberg's Uncertainty Principle thing? The very act of measuring Labour's vote share moves the polls?

If Labour are doing really well (1997, 2024), fewer people vote Labour, because a lot of them are doing it to get the Tories out. If Labour are doing badly (2010, 2017), they get sympathy votes drifting back.

I also feel like maybe voters on the broad left are just more used to this kind of context-dependent, multiple-options calculus at this point, so they're a lot more... fluid?

In 2017 in particular, I think there were alot of nose-holding remain votes lent to Labour.

This is technically less of a quantum mechanic thing, and more of a graph modelling problem with a cycle IMO

Polls influence votes, which influence polls, which influence votes etc

When it comes to modelling, cyclic graphs are a pain, because often you lose guarantees of a convergent solution

We love DAGs (Directed Acyclic Graphs) for this reason

this may be above my paygrade I'm already baffled

Haha honestly the maths is worth it 😅 I would love to see an issue of your newsletter going into the weeds of polling models

Honestly this is as good an explanation as I've seen

I think so, yes. If the party's comfortably winning, it's easy to indulge your preference for the Tooting Popular Front as a protest against Labour policy on bin collections. If it's way behind, the prospect of unfettered Tory power probably focuses a few minds.

Post