I imagine they've all changed models since then. They always fight the last war
I think up to 2017 pollsters had models assuming young people were less likely to vote than they said they were (see Lib Dem 2010 results) - as that had always proven to be the case in the past but that didn't work in 2017. I haven't read anything this time about age being a factor in turnout models.
It might be nicer to cross-reference it by what probability these people had said they would vote with in 2017. So if someone said 0 out of 10 then didn't vote, and is saying 10 out of 10 this time, they should be taken more seriously than someone who said 10 out of 10 last time but didn't vote. But then there is the problem with continually asking the same panel political questions - they start to become unrepresentative because you keep making them think about politics so it's not necessarily good to keep digging up the same respondents.
I think if this is really going to be the Brexit election that 2017 ended up not being, then arguably referendum participation should count too. Thinking about it more, I'd like to also look at whether these people voted in 2015 too.
ICM method seems too harsh - I don't think someone who gives themselves an 8 out of 10 chance of voting but didn't vote last time is only half as likely as someone who gives themselves an 8 out of 10 probability but did vote. On the other hand, I think the others are probably wrong to assume they are both equally likely to vote.
Last edited by LektorAJ; 11-10-2019 at 12:41 PM.