I didn't set out to write this thread but got provoked by a statistical comment and it developed into a substantive political point that I warmed to. Since that developed in the replies I'm making a thread of it here. TL;DR Religion matters but not too much, and that's good. 1/x
Partisanship regressed on detailed religious affiliation. Mainline protestant is the base category. Negative coefficients are more Republican, positive are more Democratic than mainline. @MULawPoll surveys, 2013-2019. Party ID is scored -2 R -1 LeanR 0 Ind 1 LeanD 2 Dem 2/x 
To which @BruneElections replied "R^2 is tiny though. Although these are all significant they aren't close to everything." That's true but nothing puts me in a fighting mood faster than a debate over R^2. So I replied, spitefully, "Then by all means ignore the analysis." 3/x
Had we stopped there, this would be a typical Twitter exchange that lowers the value of human capital in the universe. But my thanks to Ryan for actually making me think about his comment, beyond initial churlishness. So I developed a more civil and, I think, useful point. 4/x
Here is my point. There are times to assess model fit and even pay some attention to R^2. But what would it mean to find a large R^2 in this case? 5/x
A large R^2 would mean a nation so polarized by religion that it explains most of partisanship. But are there no Democratic Baptists? No Republican Jews? Of course there are both because religion DOES NOT completely, even largely, explain party. 6/x
This is mostly a statistical objection on my part. R^2 is not a good basis for evaluating this relationship. But the substantive implications are substantial. We are not in a religious war, even if there are substantial, non-trivial, important differences on average by faith 7/x
Thank whatever higher power that the R^2 is not higher. And celebrate that this analysis shows there is enormous variation in political views within each faith. Something to be grateful for in this season of religious celebrations. 8/x
To close, this is an example of how the statistical estimates shows us that faith groups do differ significantly from one another, consistent with obvious reality. But the weakness of the fit, the R^2, shows us something even more substantively important about our society. 9/x
Our divisions are real, but they are far far less than hyperbolic rhetoric would claim. Our rhetoric paints "enemies" when the "enemy" actually shares our pews. Prays with us in synagogue or mosque. Coaches our daughter's soccer team. 10/x
This is why I love political methodology and statistics. The stat is not just "variance explained" but a normatively meaningful insight into our current situation. Real divisions on average are still far short of perfect polarization. Thank God. 11/11
