Thread by @NateSilver538

1. Since I'm up late, an *extremely* nerdy thread...

The aggregate popular vote for the U.S. House—which is basically what the generic ballot is trying to measure—is a useful benchmark for the national environment in Congressional races, but it suffers from a couple of flaws.

2. One is that whichever party did better in the prior election has more incumbents, and therefore overperforms in the popular vote relative to if everyone was starting from scratch. It can be useful to separate out the incumbency advantage from the partisan environment.

3. Another is that the popular vote is sensitive to how many districts a party manages to nominate a candidate in. If you fail to nominate someone in a few dozen districts, that can you a couple of points in the popular vote, even if the districts wouldn't have been competitive.

4. Anyway, I developed a measure that assesses the partisan environment based on the *median* district (which de-emphasizes what happens in non-competitive districts) and adjusts for incumbency. Here's how it compares to the popular vote.

5. One takeaway is that the House popular vote somewhat understates D performance in elections since 1994 because R's have almost always had more incumbents. 2012 is a good example; the GOP had *lots* of incumbents after 2010, but Dems did very well in competitive races.

6. Another interesting finding is that my alternative measure of the partisan environment matches the generic ballot a lot better than the actual popular vote does. This implies the generic ballot has issues with accounting for (i) incumbency and (ii) uncontested races.

7. Put another way, the generic ballot is probably better thought of as a (useful) indicator of the partisan environment, especially in competitive districts, rather than as a literal prediction of the popular vote.

OK, hope that was dorky enough. Good night.