This is a solid data analytic article by Lakshya Jain on Sabato's Crystal Ball. Jain uses Catalist data to simulate the partisan lean of recent midterm and presidential electorates if they had voted by demographic group as they did in 2020. This indicates some possible shifts in the traditional relative Republican lean of midterm electorates due to the shifting loyalties of the white college graduate demographic. Jain's key tables are below.
His overall conclusions:
— "Midterm electorates are typically whiter and more educated than presidential electorates.
— At one time, this sort of change from the presidential to the midterm electorate might have made midterm electorates worse for Democrats. But given changes in the electorate, this midterm turnout pattern may actually aid Democrats, or at least not hurt them as much as it once did.
— Minority turnout has fluctuated and is a wild card that plays a big role in determining baseline partisan leans and advantages — presidential-level turnout means Democrats enjoy the advantage, whereas dips favor Republicans.
— The outcome in key swing states whiter than the national average, such as Pennsylvania, Wisconsin, and New Hampshire, may be influenced heavily by educational turnout differential. In states with large nonwhite cores, such as North Carolina, Georgia, Arizona, and Nevada, minority turnout will play a more critical role."