Bryan Caplan's book[1] has a chapter where he specifically aims to control for this, and to separately calculate the ROI of different levels of education for "E

Bryan Caplan's book[1] has a chapter where he specifically aims to control for this, and to separately calculate the ROI of different levels of education for "Excellent", "Good", "Fair", and "Poor" students. The main result was that college is a good deal for Excellent/Good students, a terrible choice for Poor students, and only makes sense for Fair students in special cases (e.g. they got a scholarship, or plan to study engineering). The bad ROI for Fair and Poor students is mostly driven by their lower probability of actually graduating and getting the degree. I highly recommend the book if you're interested. [1] https://press.princeton.edu/titles/11225.html https://press.princeton.edu/titles/11225.html