alt Hacker NewsAgreed, totally a discussion worth having there. But you kinda point to my assumption on this: "like any large organization."
I'm doubtful that the organizations at the top end of the list are 4x more bureaucratically bloated than those at the bottom end of it.
I'm highly confident that much more sophisticated research has much higher indirect cost, because a defining characteristic of "sophisticated research" is that it entails exquisite facilities and equipment that cannot possibly be paid for under individual studies.
Another thing I'm confident of is that Harvard et al have much more talented negotiators than the smaller schools, and I'm sure that plays a role. I would be surprised if it explains the bulk of the discrepancy.
It really makes no sense for overhead to be calculated at the university level, but it sounds like that's how it is done?
Sophisticated research into particle physics, material science, and (for the last several years) AI does come with significant overhead costs for opex. Sophisticated research into most of computer science, mathematics, and other theoretical scientific disciplines does not, let alone humanities research.
I think a large portion of the difference in overhead rates is due to the last item you are confident of in your list (i.e. Duke can tell NIH "if you want our world class researchers to work on this problem, here are our rates", and some random school cannot).