If you’re a scientist, and you apply for federal research funding, you’ll ask for a specific dollar amount. Let’s say you’re asking for a million-dollar grant. Your grant covers the direct costs, things like the salaries of the researchers that you’re paying.

If you get that grant, your university might get an extra $500,000. That money is called “indirect costs,” but think of it as overhead: that money goes to lab space, to shared equipment, and so on.

This is the system we’ve used to fund American research infrastructure for more than 60 years. But earlier this year, the Trump administration proposed capping these payments at just 15% of direct costs, way lower than current indirect cost rates. There are legal questions about whether the admin can do that. But if it does, it would force universities to fundamentally rethink how they do science.

The indirect costs system is pretty opaque from the outside. Is the admin right to try and slash these indirect costs? Where does all that money go? And if we want to change how we fund research overhead, what are the alternatives? How do you design a research system to incentivize the research you actually wanna see in the world?

I’m joined today by Pierre Azoulay from MIT Sloan and Dan Gross from Duke’s Fuqua School of Business. Together with Bhaven Sampat at Johns Hopkins, they conducted the first comprehensive empirical study of how indirect costs actually work. Earlier this year, I worked with them to write up that study as a more accessible policy brief for IFP.

They’ve assembled data on over 350 research institutions, and they found some striking results. While negotiated rates often exceed 50-60%, universities actually receive much less, due to built-in caps and exclusions.

Moreover, the institutions that would be hit hardest by proposed cuts are those whose research most often leads to new drugs and commercial breakthroughs.

Thanks to Katerina Barton, Harry Fletcher-Wood, and Inder Lohla for their help with this episode, and to Beez for her help on the charts.

Let’s say I’m a researcher at a university and I apply for a federal grant. I’m looking at cancer cells in mice. It will cost me $1 million to do that research — to pay grad students, to buy mice and test tubes.

I apply for a grant from the National Institutes of Health, or NIH. Where do indirect costs come in?

Dan Gross: Research generally incurs two categories of costs, much as business operations do.

* Direct or variable costs are typically project-specific; they include salaries and consumable supplies.

* Indirect or fixed costs are not as easily assigned to any particular project. [They include] things like lab space, data and computing resources, biosecurity, keeping the lights on and the buildings cooled and heated — even complying with the regulatory requirements the federal government imposes on researchers. They are the overhead costs of doing research.

Pierre Azoulay: You will use those grad students, mice, and test tubes, the direct costs. But you’re also using the lab space. You may be using a shared facility where the mice are kept and fed. Pieces of large equipment are shared by many other people to conduct experiments. So those are fixed costs from the standpoint of your research project.

Dan: Indirect Cost Recovery (ICR) is how the federal government has been paying for the fixed cost of research for the past 60 years. This has been done by paying universities institution-specific fixed percentages on top of the direct cost of the research. That’s the indirect cost rate. That rate is negotiated by institutions, typically every two to four years, supported by several hundred pages of documentation around its incurred costs over the recent funding cycle.

The idea is to compensate federally funded researchers for the investments, infrastructure, and overhead expenses related to the research they perform for the government. Without that funding, universities would have to pay those costs out of pocket and, frankly, many would not be interested or able to do the science the government is funding them to do.

Imagine I’m doing my mouse cancer science at MIT, Pierre’s parent institution. Some time in the last four years, MIT had this negotiation with the National Institutes of Health to figure out what the MIT reimbursable rate is.

But as a researcher, I don’t have to worry about what indirect costs are reimbursable. I’m all mouse research, all day.

Dan: These rates are as much of a mystery to the researchers as it is to the public. When I was junior faculty, I applied for an external grant from the National Science Foundation (NSF) — you can look up awards folks have won in the award search portal.

It doesn’t break down indirect and direct cost shares of each grant. You see the total and say, “Wow, this person got $300,000.” Then you go to write your own grant and realize you can only budget about 60% of what you thought, because the rest goes to overhead. It comes as a bit of a shock the first time you apply for grant funding.

What goes into the overhead rates? Most researchers and institutions don’t have clear visibility into that. The process is so complicated that it’s hard even for those who are experts to keep track of all the pieces.

Pierre: As an individual researcher applying for a project, you think about the direct costs of your research projects. You’re not thinking about the indirect rate. When the research administration of your institution sends the application, it’s going to apply the right rates.

So I’ve got this $1 million experiment I want to run on mouse cancer. If I get the grant, the total is $1.5 million. The university takes that .5 million for the indirect costs: the building, the massive microscope we bought last year, and a tiny bit for the janitor. Then I get my $1 million. Is that right?

Dan: Duke University has a 61% indirect cost rate. If I propose a grant to the NSF for $100,000 of direct costs — it might be for data, OpenAI API credits, research staff salaries — I would need to budget an extra $61,000 on top for ICR, bringing the total grant to $161,000.

My impression is that most federal support for research happens through project-specific grants. It’s not these massive institutional block grants. Is that right?

Pierre: By and large, there aren’t infrastructure grants in the science funding system. There are other things, such as center grants that fund groups of investigators. Sometimes those can get pretty large — the NIH grant for a major cancer center like Dana-Farber could be tens of millions of dollars per year.

Dan: In the past, US science funding agencies did provide more funding for infrastructure and the instrumentation that you need to perform research through block grants. In the 1960s, the NSF and the Department of Defense were kicking up major programs to establish new data collection efforts — observatories, radio astronomy, or the Deep Sea Drilling project the NSF ran, collecting core samples from the ocean floor around the world.

The Defense Advanced Research Projects Agency (DARPA) — back then the Advanced Research Projects Agency (ARPA) — was investing in nuclear test detection to monitor adherence to nuclear test ban treaties. Some of these were satellite observation methods for atmospheric testing. Some were seismic measurement methods for underground testing. ARPA supported the installation of a network of seismic monitors around the world. Those monitors are responsible for validating tectonic plate theory. Over the next decade, their readings mapped the tectonic plates of the earth. That large-scale investment in research infrastructure is not as common in the US research policy enterprise today.

That’s fascinating. I learned last year how modern that validation of tectonic plate theory was. Until well into my grandparents’ lifetime, we didn’t know if tectonic plates existed.</