To
minimize bias from student sorting by instructors, I will use a course - set fixed effect
model that compares between students who take exactly the same set of courses during their first semester of college enrollment; I will further augment the
model by combining it with an instrumental variable approach which exploits term - by - term fluctuations in faculty composition in each department, therefore controlling for both between - and within - course sorting.
The availability of test scores in multiple subjects for each student permits us to estimate a
model with student fixed effects, which helps
minimize any
bias associated with the non-random distribution of teachers and students among classrooms within schools.