Thread by @lxeagle17

I've reconstructed the 2014 and 2018 electorates by demographic, thanks to @DKElections and @Catalist_US data and a lot of math

-Midterms are whiter, more educated (~2pt Dem boost with whites on education-based turnout differential)
-Minority turnout is a crucial wildcard

[1/]

Midterms are generally whiter + more educated; whites are ~2% more favorable to Dems on educational splits alone. This was more pronounced in 2018 than it was in 2014. If R voters are increasingly tied to Trump ballot presence, it could complicate things for the GOP. [2/]

I calculate the electorate demographic composition for 2014/2018 myself and project 2020 support by demographic onto each electorate to get an idea of what its partisanship would be now. 2012/16/20 demographic composition & 2020 2-way support by demographic are from Catalist [3/]

Minority turnout will probably end up making or breaking Democrats in 2022 if polarization continues and 2020 support rates remain; 2014 saw minority turnout absolutely crater from 2012 levels, while 2018 saw it stay at roughly presidential levels. [4/]

The nation is continuing to get more non-white, so that helps Dems, but they absolutely *need* to keep minority turnout up. I do not believe Dems need to compromise their white support to drive up minority turnout; that said, they'll want to also boost Hispanic margins [5/]

This makes me pretty confident that Democrats have a really good chance at keeping the Senate. Wisconsin and Pennsylvania are very white (WI is 81% white alone, PA 76%), and Democrats have a very good chance at picking up those seats. House will be tougher [6/]

Caveats:

-These are not exact estimates; these are only solid ballpark ones based on the available data
-This does not account for partisan turnout gap within demographics -- e.g. Hispanic Ds could turn out more/less than Hispanic Rs.
-Vote switching probably matters most

[7/]

Special thanks to the folks who helped me on this, whether via sanity-check, debugging, or through helping me with the math

@davidshor, @EScrimshaw, @PoliticalKiwi, @baseballot, @notkavi, and @Thorongil16

cc @NateSilver538, @Nate_Cohn, @gelliottmorris, @Redistrict, @JMilesColeman -- if anyone sees this, I'd love to hear your thoughts about the findings in this thread and how you think the 2022 midterm electorate might be impacted (or not) by any possible congressional laws

Lastly, the demographic data was fetched and scraped from the US Census Bureau, the state house electoral data was from @DKElections, and the 2020 2-way partisanship and the 2012/16/20 electorate demographic compositions are from @Catalist_US catalist.us/wh-national/.

This took a *lot* of work and theory, so if you liked it, please give it an RT or let me know of your thoughts/feedback! Open to hearing all types of feedback; I just ask that you please stay respectful, because I did spend a lot of time on this.