Why federal R&D delivers high social returns

 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.

Dan: Santi, when were you born?

1997.

Dan: So I’m a good decade older than you — I was born in 1985. When we were learning tectonic plate theory in the 1990s, it seemed like something everybody had always known. It turns out that it had only been known for maybe 25 years.

So there’s this idea of federal funding for science as these massive pieces of infrastructure, like the Hubble Telescope. But although projects like that do happen, the median dollar the Feds spend on science today is for an individual grant, not installing seismic monitors all over the globe.

Dan: You applied for a grant to fund a specific project, whose contours you’ve outlined in advance, and we provided the funding to execute that project.

Pierre: You want to do some observations at the observatory in Chile, and you are going to need to buy a plane ticket — not first class, not business class, very much economy.

Let’s move to current events. In February of this year, the NIH announced it was capping indirect cost reimbursement at 15% on all grants.

What’s the administration’s argument here?

Pierre: The argument is there are cases where foundations only charge 15% overhead rate on grants — and universities acquiesce to such low rates — and the federal government is entitled to some sort of “most-favored nation” clause where no one pays less in overhead than they pay. That’s the argument in this half-a-page notice. It’s not much more elaborate than that.

The idea is, the Gates Foundation says, “We will give you a grant to do health research and we’re only going to pay 15% indirect costs.” Some universities say, “Thank you. We’ll do that.” So clearly the universities don’t need the extra indirect cost reimbursement?

Pierre: I think so.

Dan: Whether you can extrapolate from that to federal research funding is a different question, let alone if federal research was funding less research and including even less overhead. Would foundations make up some of the difference, or even continue funding as much research, if the resources provided by the federal government were lower? Those are open questions. Foundations complement federal funding, as opposed to substitute for it, and may be less interested in funding research if it’s less productive.

What are some reasons that argument might be misguided?

Pierre: First, universities don’t always say, “Yes” [to a researcher wishing to accept a grant]. At MIT, getting a grant means getting special authorization from the provost. That special authorization is not always forthcoming. The provost has a special fund, presumably funded out of the endowment, that under certain conditions they will dip into to make up for the missing overhead.

So you’ve got some research that, for whatever reason, the federal government won’t fund, and the Gates Foundation is only willing to fund it at this low rate, and the university has budgeted a little bit extra for those grants that it still wants.

Pierre: That’s my understanding. I know that if you’re going to get a grant, you’re going to have to sit in many meetings and cajole any number of administrators, and you don’t always get your way.

Second, it’s not an apples-to-apples comparison [between federal and foundation grants] because there are ways to budget an item as a direct cost in a foundation grant that the government would consider an indirect cost. So you might budget some fractional access to a facility…

Like the mouse microscope I have to use?

Pierre: Yes, or some sort of Cryo-EM machine. You end up getting more overhead through the back door.

The more fundamental way in which that approach is misguided is that the government wants its infrastructure — that it has contributed to through [past] indirect costs — to be leveraged by other funders. It’s already there, it’s been paid for, it’s sitting idle, and we can get more bang for our buck if we get those additional funders to piggyback on that investment.

Dan: That [other funders] might not be interested in funding otherwise.

Why wouldn’t they be interested in funding it otherwise? What shouldn’t the federal government say, “We’re going to pay less. If it’s important research, somebody else will pay for it.”

Dan: We’re talking about an economies-of-scale problem. These are fixed costs. The more they’re utilized, the more the costs get spread over individual research projects.

For the past several decades, the federal government has funded an order of magnitude more university research than private firms or foundations. If you look at NSF survey data, 55% of university R&D is federally funded; 6% is funded by foundations. That is an order of magnitude difference. The federal government has the scale to support and extract value for whatever its goals are for American science.

We haven’t even started to get into the administrative costs of research. That is part of the public and political discomfort with indirect-cost recovery. The idea that this is money that’s going to fund university bloat.

I should lay my cards on the table here for readers. There are a ton of problems with the American scientific enterprise as it currently exists.

But when you look at studies from a wide range of folks, it’s obvious that R&D in American universities is hugely valuable. Federal R&D dollars more than pay for themselves. I want to leave room for all critiques of the scientific ecosystem, of the universities, of individual research ideas. But at this 30,000-foot level, federal R&D dollars are well spent.

Dan: The evidence may suggest that, but that’s not where the political and public dialogue around science policy is. Again, I’m going to bring in a long arc here. In the 1950s and 1960s, it was, “We’re in a race with the Soviet Union. If we want to win this race, we’re going to have to take some risky bets.” And the US did. It was more flexible with its investments in university and industrial science, especially related to defense aims.

But over time, with the waning of these political pressures and with new budgetary pressures, the tenor shifted from, “Let’s take chances” to “Let’s make science and other parts of government more accountable.” The undercurrent of Indirect Cost Recovery policy debates has more of this accountability framing.

This comes up in this comparison to foundation rates: “Is the government overpaying?” Clearly universities are willing to accept less from foundations. It comes up in this perception that ICR is funding administrative growth that may not be productive or socially efficient. Accountability seems to be a priority in the current day.

Where are we right now [August 2025] on that 15% cap on indirect costs?

Dan: Recent changes first kicked off on February 7th, when NIH posted its supplemental guidance, that introduced a policy that the direct cost rates that it paid on its grants would be 15% to institutions of higher education. That policy was then adopted by the NSF, the DOD, and the Department of Energy. All of these have gotten held up in court by litigation from universities. Things are stuck in legal limbo. Congress has presented its point of view that, “At least for now, I’d like to keep things as they are.” But this has been an object of controversy long before the current administration even took office in January. I don’t think it’s going away.

Pierre: If I had to guess, the proposal as it first took shape is not what is going to end up being adopted. But the idea that overhead rates are an object of controversy — are too high, and need to be reformed — is going to stay relevant.

Dan: Partly that’s because it’s a complicated issue. Partly there’s not a real benchmark of what an appropriate Indirect Cost Recovery policy should be. Any way you try to fund the cost of research, you’re going to run into trade-offs. Those are complicated.

ICR does draw criticism. People think it’s bloated or lacks transparency. We would agree some of these critiques are well-founded. Yet it’s also important to remember that ICR pays for facilities and administration. It doesn’t just fund administrative costs, which is what people usually associate it with. The share of ICR that goes to administrative costs is legally capped at 26% of direct costs. That cap has been in place since 1991. Many universities have been at that cap for many years — you can see this in public records. So the idea that indirect costs are going up over time, and that that’s because of bloat at US universities, has to be incorrect, because the administrative rate has been capped for three decades.

Many of those costs are incurred in service of complying with regulations that govern research, including the cost of administering ICR to begin with. Compiling great proposals every two to four years and a new round of negotiations — all of that takes resources. Those are among the things that indirect cost funding reimburses.

Even then, universities appear to under-recover their true indirect costs of federally-sponsored research. We have examples from specific universities which have reported detailed numbers. That under-recovery means less incentive to invest in infrastructure, less capacity for innovation, fewer clinical trials. So there’s a case to be made that indirect cost funding is too low.

Pierre: The bottom line is we don’t know if there is under- or over-recovery of indirect costs. There’s an incentive for university administrators to claim there’s under-recovery. So I take that with a huge grain of salt.

Dan: It’s ambiguous what a best policy would look like, but this is all to say that, first, public understanding of this complex issue is sometimes a bit murky. Second, a path forward has to embrace the trade-offs that any particular approach to ICR presents.

From reading your paper, I got a much better sense that a ton of the administrative bloat of the modern university is responding to federal regulations on research. The average researcher reports spending almost half of their time on paperwork. Some of that is a consequence of the research or grant process; some is regulatory compliance.

The other thing, which I want to hear more on, is that research tools seem to be becoming more expensive and complex. So the microscope I’m using today is an order of magnitude more expensive than the microscope I was using in 1950. And you’ve got to recoup those costs somehow.

Pierre: Everything costs more than it used to. Research is subject to Baumol’s cost disease. There are areas where there’s been productivity gains — software has had an impact.

The stakes are high because, if we get this wrong, we’re telling researchers that they should bias the type of research they’re going to pursue and training that they’re going to undergo, with an eye to what is cheaper. If we reduce the overhead rate, we should expect research that has less fixed cost and more variable costs to gain in favor — and research that is more scale-intensive to lose favor. There’s no reason for a benevolent social planner to find that a good development. The government should be neutral with respect to the cost structure of research activities. We don’t know in advance what’s going to be more productive.

Wouldn’t a critic respond, “We’re going to fund a little bit of indirect costs, but we’re not going to subsidize stuff that takes huge amounts of overhead. If universities want to build that fancy new telescope because it’s valuable, they’ll do it.” Why is that wrong when it comes to science funding?

Pierre: There’s a grain of truth to it.

Dan: With what resources though? Who’s incentivized to invest in this infrastructure? There’s not a paid market for science. Universities can generate some licensing fees from patents that result from science. But those are meager revenue streams, realistically. There are reasons to believe that commercial firms are under-incentivized to invest in basic scientific research. Prior to 1940, the scientific enterprise was dramatically smaller because there wasn’t funding the way that there is today. The exigencies of war drew the federal government into funding research in order to win. Then it was productive enough that folks decided we should keep doing it. History and economic logic tells us that you’re not going to see as much science — especially in these fixed-cost heavy endeavors — when those resources aren’t provided by the public.

Pierre: My one possible answer to the question is, “The endowment is going to pay for it.” MIT has an endowment, but many other universities do not. What does that mean for them? The administration also wants to tax the heck out of the endowment.

This is a good opportunity to look at the empirical work you guys did in this great paper. As far as I can tell, this was one of the first real looks at what indirect costs rates look like in real life. What did you guys find?

Dan: Two decades ago, Pierre and Bhaven began collecting information on universities’ historical indirect cost rates. This is a resource that was quietly sitting on the shelf waiting for its day. That day came this past February. Bhaven and Pierre collected information on negotiated ICR rates for the past 60 years. During this project, we also collected the most recent versions of those agreements from university websites to bring the numbers up to the current day.

We pulled together data for around 350 universities and other research institutions. Together, they account for around 85% of all NIH research funding over the last 20 years.We looked at their:

* Negotiated indirect cost rates, from institutional indirect cost agreements with the government, and their;

* Effective rates [how much they actually get when you look at grant payments], using NIH grant funding data.

Negotiated cost rates have gone up. That has led to concerns that the overhead cost of research is going up — these claims that it’s funding administrative bloat. But our most important finding is that there’s a large gap between the sticker rates — the negotiated ICR rates that are visible to the public, and get floated on Twitter as examples of university exorbitance — and the rates that universities are paid in practice, at least on NIH grants; we think it’s likely the case for NSF and other agency grants too.

An institution’s effective ICR funding rates are much, much lower than their negotiated rates and they haven’t changed much for 40 years. If you look at NIH’s annual budget, the share of grant funding that goes to indirect costs has been roughly constant at 27-28% for a long time. That implies an effective rate of around 40% over direct costs. Even though many institutions have negotiated rates of 50-70%, they usually receive 30-50%.

The difference between those negotiated rates and the effective rates seems to be due to limits and exceptions built into NIH grant rules. Those rules exclude some grants, such as training grants, from full indirect cost funding. They also exclude some direct costs from the figure used to calculate ICR rates. The implication is that institutions receive ICR payments based on a smaller portion of their incurred direct costs than typically assumed. As the negotiated direct cost falls, you see a university being paid a higher indirect cost rate off a smaller — modified — direct cost base, to recover the same amount of overhead.

Is it that the federal government is saying for more parts of the grant, “We’re not going to reimburse that as an indirect cost.”?

Dan: This is where we shift a little bit from assessment to speculation. What’s excluded from total direct costs? One thing is researcher salaries above a certain level.

What is that level? Can you give me a dollar amount?

Dan: It’s a $225,700 annual salary. There aren’t enough people being paid that on these grants for that to explain the difference, especially when you consider that research salaries are being paid to postdocs and grad students.

You’re looking around the scientists in your institution and thinking, “That’s not where the money is”?

Dan: It’s not, even if you consider Principal Investigators. If you consider postdocs and grad students, it certainly isn’t.

Dan: My best hunch is that research projects have become more capital-intensive, and only a certain level of expenditure on equipment can be included in the modified total direct cost base. I don’t have smoking gun evidence, it’s my intuition.

In the paper, there’s this fascinating chart where you show the institutions that would get hit hardest by a 15% cap tend to be those that do the most valuable medical research. Explain that on this framework. Is it that doing high-quality medical research is capital-intensive?

Pierre: We look at all the private-sector patents that build on NIH research. The more a university stands to lose under the administration policy, the more it has contributed over the past 25 years — in research the private sector found relevant in terms of pharmaceutical patents.

This is counterintuitive if your whole model of funding for science is, “Let’s cut subsidies for the stuff the private sector doesn’t care about — all this big equipment.” When you cut those subsidies, what suffers most is the stuff that the private sector likes.

Pierre: To me it makes perfect sense. This is the stuff that the private sector would not be willing to invest in on its own. But that research, having come into being, is now a very valuable input into activities that profit-minded investors find interesting and worth taking a risk on.

This is the argument for the government to fund basic research?

Pierre: That argument has been made at the macro-level forever, but the bibliometric revolution of the past 15 years allows you to look at this at the nano-level. Recently I’ve been able to look at the history of Ozempic. The main patent cites zero publicly-funded research, but it cites a bunch of patents, including patents taken up by academics. Those cite the foundational research performed by Joel Habener and his team at Massachusetts General Hospital in the early 1980s that elucidated the role of GLP-1 as a potential target. This grant was first awarded to Habener in 1979, was renewed every four or five years, and finally died in 2008, when he moved on to other things. Those chains are complex, but we can now validate the macro picture at this more granular level.

Dan: I do want to add one qualification which also suggests some directions for the future. There are things we still can’t see — despite Pierre’s zeal. Our projections of the consequence of a 15% rate cap are still pretty coarse. We don’t know what research might not take place. We don’t know what indirect cost categories are exposed, or how universities would reallocate. All those things are going to be difficult to project without a proper experiment.

One thing that I would’ve loved to have more visibility into is, “What is the structure of indirect costs at universities across the country? What share of paid indirect costs are going to administrative expenses? What direct cost categories are being excluded?” We would need a more transparency into the system to know the answers.

Does that information have to be proprietary? It’s part of negotiations with the federal government about how much the taxpayer will pay for overhead on these grants. Which piece is so special that it can’t be shared?

Pierre: You are talking to the wrong people here because we’re meta-scientists, so our answer is none of it should be private.

Dan: But now you have to ask the university lawyers.

What would the case from the universities be? “We can’t tell the public what we spend subsidy on”?

Pierre: My sense is that there are institutions of academia that strike most lay people as completely bizarre.

Hard to explain without context?

Pierre: People haven’t thought about it. They will find it so bizarre that they will typically jump from the odd aspect to, “That must be corruption.” University administrators are hugely attuned to that. So the natural defensive approach is to shroud it in secrecy. This way we don’t see how the sausage is made.

Dan: Transparency can be a blessing and a curse. More information supports more considered decision-making. It also opens the door to misrepresentation by critics who have their own agendas. Pierre’s right: there are some practices that to the public might look unusual — or might be familiar, but one might say, “How is that useful expense?” Even a simple thing like having an administrator who manages a faculty’s calendar might seem excessive. Many people manage their own calendars. At the same time, when you think about how someone’s time is best used, given their expertise, and heavy investment in specialized human capital, are emails, calendaring, and note-taking the right things for scientists [to be doing]? Scientists spend a large chunk of their time now administering grants. Does it make sense to outsource that and preserve the scientist’s time for more science?

When you put forward data that shows some share of federal research funding is going to fund administrative costs, at first glance it might look wasteful, yet it might still be productive. But I would be able to make a more considered judgment on a path forward if I had access to more facts, including what indirect costs look like under the hood.

One last question: in a world where you guys have the ear of the Senate, political leadership at the NIH, and maybe the universities, what would you be pushing for on indirect costs?

Pierre: I’ve come to think that this indirect cost rate is a second-best institution: terrible and yet superior to many of the alternatives. My favorite alternative would be one where there would be a flat rate applied to direct costs. That would be the average effective rate currently observed — on the order of 40%.

You’re swapping out this complicated system to — in the end — reimburse universities the same 40%.

Pierre: We know there are fixed costs. Those fixed costs need to be paid. We could have an elaborate bureaucratic apparatus to try to get it exactly right, but it’s mission impossible. So why don’t we give up on that and set a rate that’s unlikely to lead to large errors in under- or over-recovery. I’m not particularly attached to 40%. But the 15% that was contemplated seems absurdly low.

Dan: In the work we’ve done, we do lay out different approaches. The 15% rate wouldn’t fully cut out the negotiation process: to receive that, you have to document your overhead costs and demonstrate that they reached that level. In any case, it’s simplifying. It forces more cost-sharing and maybe more judicious investments by universities. But it’s also so low that it’s likely to make a significant amount of high-value, life-improving research economically unattractive.

The current system is complicated and burdensome. It might encourage investment in less productive things, particularly because universities can get it paid back through future ICR. At the same time, it provides pretty good incentives to take on expensive, high-value research on behalf of the public.

I would land on one of two alternatives. One of those is close to what Pierre said, with fixed rates, but varied by institution types: one for universities, one for medical schools, one for independent research institutions — because we do see some variation in their cost structures. We might set those rates around their historical average effective rates, since those haven’t changed for quite a long time. If you set different rates for different categories of institution, the more finely you slice the pie, the closer you end up to the current system. So that’s why I said maybe, at a very high level, four categories.

The other I could imagine is to shift more of these costs “above the line” — to adapt the system to enable more of these indirect costs to be budgeted as direct costs in grants. This isn’t always easy, but presumably some things we currently call indirect costs could be accounted for in a direct cost manner. Foundations do it a bit more than the federal government does, so that could be another path forward.

There’s no silver bullet. Our goal was to try to bring some understanding to this long-running policy debate over how to fund the indirect cost of research and what appropriate rates should be. It’s been a recurring question for several decades and now is in the hot seat again. Hopefully through this work, we’ve been able to help push that dialogue along.

This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.statecraft.pub