Thread by @gtryan

Ok, as promised I dug into the numbers a little more. Particularly around how people voted in the 2020 primary. There are better ways to estimate overall partisanship, but I've stuck with this for now.

First, I want to let folks revisit the breakdown of D and R primary voters who have voted so far, vs the breakdown through the first week of early voting in the general.

Dem "lead" is 154k vs 181k in the general on lower overall turnout. That's about D +16% vs D +17.4%.

This also lines up with what Nate Cohn has been posting about, which looks at this a lot more thoroughly than I think I am. Dems are doing well, but not necessarily as well as they were at this point in the general.

https://twitter.com/Nate_Cohn/status/1339941872542883842

What I want to talk about is Simpson's Paradox, which is my favorite paradox. And yes, I understand how it sounds saying I have a favorite paradox.

en.wikipedia.org/wiki/Simpson%2…

Basically a population of X can be made up of 2 kinds of things: A and B.

The averages of A and B can both go in one direction, but the overall average of X goes in the other because the portions of A and B change as well.

In this case, the overall Democratic performance appears to be a little down vs the general. But mail votes and in-person votes are actually both doing BETTER for Democrats than they did in the general.

Here's a table to demonstrate this a little more clearly.

D's are doing 0.1% to 0.3% better overall by each voting method, but because the percentage of in-person votes is so much higher overall it's bringing down the overall average.

Also, "no vote" numbers are pretty similar.

And that's Simpson's Paradox. I love it, and I'm not afraid to admit it.

One thing to point out is that while mail votes are down by 25% vs value in the general, applications are only down by 9%. So we may continue to see that gap close as ballots are returned and processed.