I'd be interested in seeing the underlying data because there's a lot of noise you introduce by just equally weighing all colleges as data pts and then teasing

I'd be interested in seeing the underlying data because there's a lot of noise you introduce by just equally weighing all colleges as data pts and then teasing out a weak trend. You could boil down the entire ban and non-ban populations to a single number for each group which would probably tell you more than these graphs do. Better yet do it over time.