Does Research "Lose Money?" Depends on What Is Measured

"We lose money on research." Some version of that line shows up in nearly every conversation about university finance, from a president's budget forum to a congressional hearing on indirect costs (also called Facilities and Administration or F&A). It is a true statement when accounting for these costs one way and misleading by another. The trouble is that the way institutional finance gets talked about, and the way it gets disclosed, only ever surfaces the first way, so most people never learn there is a second.

On a campus three groups interact with this claim, usually separately, and each sees a different segment of the same picture. Leadership sees the fixed-cost base it must fund and calls research activity a “loss leader.” Faculty see their own labs and research portfolios, where the indirect costs taken look far larger than what their work costs to run, and read the rate as grossly exaggerated. Research administration sits closest to the money but is trained to manage and apply the rate, not to ask where it goes. This is less disagreement than a shared blind spot: no one is working from the same understanding, and nothing in how the finances are presented would show them the difference.

That shared understanding is the point. Not to determine who is right or which costing method we should apply, but to put the same two numbers in front of everyone, so the entire research community can ask better and maybe more honest questions. As the political scientist Deborah Stone argues, numbers are never neutral, and those that get measured and repeated are the ones useful to the people invoking them. For an institution making the case that research is underfunded or can’t distribute more indirect cost recovery to departments, one of these two numbers is far more useful than the other, which mostly explains why you may have only been provided with one message.

The claim, and the number under it

Research as a loss leader is a claim about the indirect costs against which a grant's direct costs are assessed. When an institution says it loses money on research, it means its negotiated indirect cost recovery does not cover the fully loaded indirect costs it calculates and assigns to research: the buildings, the administration, the debt.

The institution's indirect costs are pooled and then divided by the direct-cost base of research to set the rate. That rate is an average, and it appears as the research rate in a Negotiated Indirect Cost Rate Agreement (NICRA), an accounting instrument for allocating a share of shared costs to each grant. It charges each grant a proportion of the building debt and of central administration costs the grant did not create and that would not disappear if the grant vanished.

That observation leads to a different way of thinking about institutional cost. If every sponsored project disappeared tomorrow, how much of the indirect cost pool would disappear with it? Some certainly would. But libraries, human resources, police, central administration, utilities, debt service, and many compliance functions would remain, at least in the near term. That distinction matters, because most of that pool stays whether research grows or shrinks. The rate treats it as a cost of research. Much of it is a cost of running a university. So, "we lose thirty cents on the dollar" is an average cost statement. It is each grant being charged for a portion of bills the institution already had, capturing the portion attributed to use by research activity.

The claim runs entirely on that number. There is a second number that is rarely mentioned: marginal cost, which is the additional cost that one more grant creates. A little more electricity from running a lab longer, the hours the grants office spends setting up and administering the award, some wear on shared equipment. Real expenses, but small next to the building, the maintenance, and the people already on payroll to staff them. Whether a grant is worth taking operates on this number, not the average cost. Does the money it brings in cover the extra cost it creates, with something left toward the fixed bills that are due no matter what? For nearly every grant, the answer is yes, by a wide margin.

Three viewpoints, three numbers

Leadership reasons from the average cost, because that is the number it is responsible for. It must fund the entire fixed-cost base, and indirect costs are what is supposed to cover it. From that vantage the loss is real, and objectively so: recovery falls short of fully loaded cost. What the average hides, or does not reveal, is that a large part of that base is expansion the institution chose, charged back to grants as though research caused it when it was a growth strategy instead.

Faculty reasons from the margin, from what they can see: their lab space, grants, and people. From that vantage the indirect cost sweep looks enormous next to what their work costs to run, so "cut the rate, put more into direct costs" feels obvious. They are correct that their marginal footprint is far below the recovered rate. What they miss is the shared fixed costs they use and never see, the compliance apparatus, the core services, the building, the debt, which is not obvious in their bundle of projects but no less representative of real expenses.

Research administration sits closest to the money and applies the rate in day-to-day work. RAs build budgets, process awards, and confirm the collected rate is correct. But most of that work is administering to a number, not interrogating it. The rate is a given, the sweep happens above the unit, and nothing in the departmental job teaches where the recovery goes or how the average was calculated and negotiated. So, the group with the best line of sight to the whole mechanism usually looks at one piece of it to perform their job.

One could observe the three talk past each other in real time over the 15% indirect cost cap federal agencies announced in February 2025. Faculty argued the rate was too high and money should move to their direct cost budgets, reasoning from their own footprint. Administrators and leadership defended the rate, reasoning from the average cost base, with RAs understanding that research costs money even if they couldn't tell you exactly how those rates are constructed and negotiated. Both were talking about the same indirect cost rates and understanding it differently, one as a marginal cost, the other as a fully loaded cost average, and neither of them named that the disagreement was about which number counts. It looked like an argument about whether indirect cost rates had drifted too high. It was really an argument between two numbers, with neither side naming that.

Revealed preference

If a significant portion of individual grants lost money, universities would not compete aggressively for grant funding. Instead, they hire faculty to win grants, build labs to house them, and treat research volume as a measure of health and prestige. That is not the behavior of an enterprise losing money on grant funded awards. Either universities are acting against their own interest at enormous scale, or the loss is not what it appears. The behavior points to the second.

Let’s look at an example using a hotel. A room sits empty, but the building has debt service obligations whether it is occupied or not. A guest offers $60 against a fully allocated cost of $150 per night, which on paper looks like a $90 loss. But the only new cost of filling the room is cleaning it, say $20. Renting at $60 leaves $40 toward a mortgage that was due anyway. Turning the guest away to avoid the "$90 loss" would be the actual mistake, because the $90 was mostly debt service payment, which is due at a specific rate regardless of occupancy.

The grant is the room. The building represents the payment on the debt that a university also may have. The recovered indirect cost is the $60 accepted room rate. The cleaning is the marginal cost. On the fully allocated books the grant looks like a loss. On the margin it is helping to pay the debt on the building where work takes place. Both numbers are true, but they answer fundamentally different questions.

This is why "we lose money on research" and "we keep building more research" are both true and not in conflict, though they seem like it. On the fully allocated books each grant carries a share of the building debt and shows a loss. On the margin each grant helps pay the debt incurred. And as the institution adds grants, that same fixed cost base spreads across more of them, so the loss per grant on the average cost books shrinks. Nothing got cheaper. A fixed number was divided by a larger one.

There's a fair objection here: maybe universities expand research for prestige, rankings, and philanthropy, not because it pays. That may be true. But it cuts the same way. If research delivers those returns, then the size of the unrecovered indirect cost tells you almost nothing about whether the research was worth doing. Similar to the hotel…it might keep taking that $60 offer for reasons beyond the $60, repeat business, reputation, a full lobby.

This is the same average-versus-marginal gap that lets an airline sell its last seat cheap and pushes a software company to chase scale. Every high-fixed-cost business lives on its marginal cost, because that is where the decisions get made. A business knows its own because it uses it, to set a price or to decide how much to produce. A university does neither. It cannot set its F&A rate, which the government negotiates and caps, and it does not choose how many grants it wins. So the one number that normally drives both decisions has no decision to drive, and it goes unmeasured. Universities produce the same gap as every capital-heavy business and then, alone among them, never look at it.

Watch the pool get spread

A simple version makes the innerworkings visible. Say the fixed-cost base is one million dollars. It does not grow when the institution adds grants, so the allocated cost per grant falls purely because the denominator grows:

Now set recovery beside it. Say each grant recovers $300,000 in indirect cost and creates $80,000 in real marginal cost, so each one contributes about $220,000 toward that fixed million. With one grant, the institution is deep underwater, and this is the year research loses money is correct on all counts. Around five grants it clears the million and breaks even. At ten, it is more than a million to the good. At twenty, a comfortable margin.

Same institution, same grant terms. It moves from losing money on every grant to generating a surplus purely by adding grants until the fixed-cost base is spread thin enough to clear. That is the engine, and it is why institutions expand even while presidents and finance officers call research a loss.

There is one catch. The average cost per grant keeps falling only while the building has room. The moment the institution runs out of space and must build, the fixed-cost base jumps from one million to something larger, financed with debt, and the break-even count resets higher, which can push the next negotiated indirect cost rate upward. The institution rides the cheap part of the curve until capacity forces a new fixed obligation, and then it needs even more grants to clear the new, bigger base.

It’s worth noting that interest on construction debt is an allowable indirect cost (§ 2 CFR 200.449), and a 1982 revision to Circular A-21 first made it so. Robert M. Rosenzweig, president of the Association of American Universities from 1983 to 1993, argued that the 1982 revision to OMB Circular A-21 allowing interest on construction debt as an indirect cost may have contributed to the campus building boom that followed.[1] So, indirect cost is, by design, partly a mechanism to recover the debt service on research buildings. Build with debt, recover against that debt through the facilities component of every grant, and that recovery is what keeps the debt serviceable. The reimbursement described as cost recovery is, in one of its parts, a way to service building debt, and institutions have stacked fixed obligations on top of it.

Why nobody talks about the marginal number

If marginal cost is the number that tells you whether a grant is worth taking, why does no one measure or report it?

Firstly, the federal reimbursement system was never built to ask whether a grant was worth taking. That is the institution's own call, which is why an institution will accept a foundation grant at a much lower indirect rate: marginal cost and capturing direct costs almost always make it worth it. The federal system asks a narrower question: what is the government's fair share of the shared costs this institution already has (with the fair share being an average by construction)? The rate is the indirect pool divided by the research base. Nothing in it asks what one grant costs at the margin, because that is the institution's investment question, not the government's reimbursement question. The government is splitting a bill, not appraising a deal.

Second, average cost is auditable and marginal cost is not. Hand an auditor a pool and a base and they can verify the calculation. Marginal cost is not a tidy number. What does one more grant cost? It depends on how full the building already is, what gets shared, the timing, and how you would even carve out the share of an electricity bill for a single project. These questions are answered by modeling judgment calls built from a long list of arguable inputs, and a compliance system will never anchor on a figure that cannot be made objective.

Finally, nobody with the power to require it benefits from having it. The government pays a negotiated rate, so it has no use for the figure. The institution's entire underfunded position runs on the average cost, so it has every reason not to surface the marginal one. The only people who would want it are outside analysts, and they cannot get the internal data to model it. It is an orphan metric, wanted by no one who could compel it.

Two numbers, better decisions

None of this means an institution should go measure marginal cost. On any single grant it would be a lot of work for a soft number, and nobody should be ranking awards by it in terms of relative value. Its value is not as a metric to compute, but as a way to think about what is missing from the conversation.

Where it earns its place is capacity. The decision to grow, a building, a center, a cluster of hires, is a marginal decision wearing an average disguise. What that new space will actually cost, and what new research it will have to attract to carry its debt, is a marginal question, and it is usually asked, if at all, after the commitment is made rather than before. Bringing that question into the room before the bonds are signed is the one place the second number changes an outcome.

The rest is simpler and smaller: everyone on campus a little more conversant in how the money works. Faculty who understand the rate is an average of a fixed-cost base, not a measure of what their own work costs, argue about something real instead of "the rate is a rip-off." Leadership that names when it is speaking in fully loaded terms lets the room hear the claim accurately, not so it can be attacked, but so an average and a margin are not mistaken for the same thing. And research administration, closest to the machinery, can put the marginal question on the table before a commitment is made. When a director wants to add a faculty line, what it adds in staffing, space, and support usually goes unasked, and the answer is often not a net gain for a unit already stretched thin. None of that requires a new number. It requires a shared vocabulary for the two we already have.

This is not an attempt to prove that research pays for itself because marginal costs exist. It is an attempt to separate ideas that get collapsed into one word: subsidy. When an institution says it subsidizes research, that single number is holding several different things at once. Some of it is unrecovered allocated cost, the share of a building or an office the rate assigns to research even though it would exist anyway. Some of it is the gap left by caps and exclusions. And some of it is deliberate investment, space built and debt taken on as a bet on future growth. Those are not the same, and the indirect cost rate cannot tell you how much of each you are looking at.

That is why the two numbers matter, and why it matters that all three groups can understand them. The average tells you what research was assigned. The margin tells you what it costs. Most of what gets called subsidy lives in the space between, and right now that space is described almost entirely by the people with a reason to describe it one way. The point is not to replace their number with a better one. It is to make the second number sayable, so that faculty, administrators, and leadership are at least arguing about the same thing. A conversation where everyone can name what they mean by "subsidy" is a smaller ask than reform and a larger one than a talking point. It is the difference between a claim that can be examined and one that only must be believed.

[1] Robert M. Rosenzweig, argued in The Politics of Indirect Costs (Council on Governmental Relations, 1998) that the 1982 revision to OMB Circular A-21 allowing construction-debt interest as an indirect cost may have contributed to the campus building boom that followed; discussed in Azoulay, Gross, and Sampat (2025). https://www.nber.org/papers/w33627