OK so inspired by @lxeagle17's recent demographic regressions, here's the @bot_2024 model run using 2020 demographics and partisanship by demographic, but using midterm turnout numbers 
As you can see, 2018 turnout is very favorable for democrats, while 2014 turnout is about neutral to 2020. Strangely enough, 2010 turnout does well for democrats, presumably because of suburban/exurban surge vs wwc
Looking at presidential years, we can see that our model believes 2012 and 2016 actually had more favorable turnout profiles than 2020 when considering 2020 partisanship, by about 1 point in both cases
A few caveats.
1) These are all using 2020 partisanship. Partisanship might change drastically in the 2022 midterms, and persuasion is a much bigger effect than a lot of people like to think. A swing of R+7 would not be unprecedented, and would wipe away even 2018 turnout
1) These are all using 2020 partisanship. Partisanship might change drastically in the 2022 midterms, and persuasion is a much bigger effect than a lot of people like to think. A swing of R+7 would not be unprecedented, and would wipe away even 2018 turnout
2) Our model is fit to only 6 elections, 2010-2020, only modeling the presidential race in presidential years and house races in house years. It as such might have somewhat overfit demographics and thus the demographics it has learned might be a bit different year to year.
But with those caveats in mind, I think its clear our model belives that the "democrats do badly in midterms due to turnout" effect that a lot of people seem to imagine is an iron law of politics is a myth. With modern coalitions, this effect probably works in reverse.
This thread was inspired by this post in particular, which provides a similar kind of evidence, there focusing only on white voters.
https://twitter.com/lxeagle17/status/1406004601958981635?s=20
• • •
Missing some Tweet in this thread? You can try to
force a refresh
