Thought it would be useful to do a quick explainer of how our voting intent methodology works and how it’s changed. There’s a full article here from the time (https://t.co/aaqBZHkVMp) but to explain it simply…

— Adam Drummond (@AGKD123) May 29, 2024

The problem is that polls are really bad at discerning who’s going to vote and who isn’t and most of the major polling errors you’ve seen (2015 especially) have been down to this problem.

— Adam Drummond (@AGKD123) May 29, 2024

For every other question in our poll we’re using “population weight” but for headline voting intention we use a “voters weight” based on targets from prior waves of the British Election Study face to face survey.

— Adam Drummond (@AGKD123) May 29, 2024

So if a proportion of a party’s 2019 voters say that they’ll vote but tick “don’t know”, those people are removed… pic.twitter.com/mVexkGCrf9

— Adam Drummond (@AGKD123) May 29, 2024

For a lot of 2022 (pre-Truss, then it all went nuts) other companies were showing bigger Labour leads because most of the movement was 2019Con’s moving to “don’t know” and then being removed from the final result.

— Adam Drummond (@AGKD123) May 29, 2024

Once a sizeable number of 2019Con’s started voting for other parties, then we started to see bigger Labour leads. Here's 2022: pic.twitter.com/XBnqUz50ZW

— Adam Drummond (@AGKD123) May 29, 2024