[jltsiren]

Specialization is the greatest source of efficiency.

Almost every organization has more people in support roles than in line positions. If not directly, then indirectly. If the "overhead" is not high enough, the people who are supposed to do line work have to waste their time as incompetent janitors and secretaries. That's a very common situation in universities, which employ many administrators for regulatory reasons but often don't have the money to hire enough support personnel.

If you want to pay grad students $100k, then the grant must be $165k. Or maybe $180k to allow the grad student to focus more on their research. Maybe the sum can be a bit less, if you manage to reduce the regulations and reporting requirements. But then you have to accept that a larger fraction of grant recipients will misuse the money or spend it on research many taxpayers would consider frivolous.

[logisticseh]

In CS this is just false. Nearly all of the CS research we fund could be done in inexpensive office space with no administrators. The one exception is DARPA-funded robotics research, and even then it's only a subset.

In wet lab disciplines the story is more complicated. But there are many possible configurations that are more efficient than the current university-as-gatekeeper system.

[jltsiren]

You can do many kinds of research with minimal overhead, as long as everyone is self-employed and your funding comes with minimal restrictions and reporting requirements.

I did that for a couple of years as a CS postdoc on a private grant. I believe my overhead rate was 15-20%, which was mostly health and other insurances. To reach that, I did accounting and taxes on my own. I also had a loose affiliation with a university, so I didn't have to pay for office space or computing.

If you start using paid services, hire employees, and accept more restrictive grants, your overhead rate can easily climb above 50%.