Cook Political Report (CPR) and NBC News have put together a nice interactive that's worth taking a look at. The basic idea of the interactive is that the set the baseline turnout and (two party) voter support figures at the levels they estimate for the 2016 election and then provide sliders where you can move these turnout and party support levels up or down for the 2020 election and see what effect these changes have (the changes are apparently applied uniformly across states). They also provide a baseline scenario where turnout and party support levels remain constant and only the eligible distribution by demographic group changes in 2020.
This is actually quite similar to analyses done by the States of Change project, most recently in our 2018 report on America's Electoral Future: Demographic Shifts and the Future of the Trump Coalition. And we actually have a new report coming out in October, America’s Electoral Future: The Coming Generational Transformation, that incorporates generational change into our analysis, which the the CPR/NBC interactive does not touch.
There are significant similarities between the methodology we use and the estimation methods apparently used in this CPR/NBC analysis. But I think our approach and the underlying data we use is a bit more detailed--and our results, where comparable, seem a bit different. For instance, CPR/NBC note the significance of the decline of the white noncollege eligible voter population and rise of more liberal populations--fair enough--and claim this is enough, holding turnout/support from 2016 constant, to hand Trump a loss in 2020, flipping FL, MI, PA and WI. In our new report, as you will see when it comes out, our results are not as optimistic for the Democrats in this kind of baseline scenario (though once generation is incorporated, the picture changes quite a bit).
I'm not sure what drives this difference in the CPR/NBC results, but I do note that they use the exit poll data as an important data source for their estimation procedures, which I am dubious of. Perhaps this is why the Democrats' white noncollege deficit seems too high in their data, which could have an effect on their results. However, that's national data and they don't show their state level estimates so it's pretty hard to tell what's really going on here.
At any rate, an interesting entry in 2020 election data sweepstakes. Play around with it and see what you think, but keep in mind that their baseline is perhaps a bit too pro-Democratic. And look for our new report, coming out next month on the 19th!