,
7 tweets,
3 min read
Read on Twitter
This graphic from @ForecasterEnten's piece is a really good example of how important it is to be true to the data. He/they should have (A) shown both dimensions of DW-NOMINATE, (B) shown the full scale of the data, and (C) picked all of the comparable candidates. 
When you do that, the differences between the groups become much more muddy. I also can't help but think that the rescaling of the liberal range of DW-NOMINATE (just half of the scale) to 0-100 was just done to exacerbate the points they chose before they analyzed the numbers.
Tangentially: Means are important, but in political science we are often more concerned with distributions. That's why we have T-tests, which he could have run to see that these differences aren't significant.
Takes like these have been discussed before f.e.: legacy.voteview.com/pdf/nominate_a…
Even if you use Nokken-Poole scores, the differences between the 2020 nominees and 2008 nominees are not statistically significant (esp. once you add in Gabbard, Delaney, Biden, Brown, O'Rourke, etc).
It is true that Dems have drifted left in recent years. It is true that the Dem nominees _look_ more liberal qualitatively. Especially on a few issues. BUT It is not sure that the 2020 nominees are wildly more liberal. I think people get this impression because of the GND 1/2
... but that impression is not holistic. Anyways, I don't think these articles are very important to the avg voter or helpful to democracy. It's not abt Dems moving left vs 2008, it's about being representative and having a broad, contrasting message. That's more imp than DW-Nom.
