[dual_basis]

Sure, but that's the sort of data we have to deal with in this arena. Who's going to run a large randomized trial where students are purposely assigned to different classes (keeping in mind that schedule conflicts already add additional constraints to this which may bias these assignments) and then, furthermore, have the instruction fixed apart from how easy the assignments are? Is it even fair to the students to knowingly assign students to relatively poorer teaching? Clearly it happens all the time, every department has that professor who is known for being a bad teacher yet they still have to assign classes.

In speaking with other grad students this seemed to be a well-known phenomenon, to the point that most other grad students intentionally didn't put much time into their teaching and basked in the positive reviews as a result. It was suggested many times to me that I was spending too much time thinking about my teaching. In my case, the lax teaching was not intentional, I simply was overcommitted that semester and had less time to prepare.