Rule of three for federal financing

For emergency preparedness, people often recommend prioritizing by a rule of threes. You can survive for:

• 3 minutes without air
• 3 hours without shelter (in a harsh environment)
• 3 days without water
• 3 weeks without food

So in a difficult situation, you work down the list for priorities – first make sure to get shelter, then worry about water. The “3” is not really a good point estimate, of course, but the big thing is that it converts a scary situation into steps one needs to take.

There is a crisis in US science funding. In biology, according to GrantWitness, as of now the number of awards (543) are half what they used to be, and the total amount of biology funding ($562M) is 2/3 of normal¹. That there is still this much reflects a lot of advocacy and hard work by many people, inside and outside of NSF. But going forward, NSF is only planning to give out 250-300 awards totalling $280M in its core biology program (though there are other sources of funds for biology). And for next year, NSF is requesting $3.9B (plus $900M for a ship), down from its current Congressionally-allocated budget of $8.8B. Congress might again allocate more than requested, though even so, spending within NSF might not continue as before. For example, though the overall budget for NSF this year is similar to last year, and not that much less than the previous year, the number of grants in social, behavioral, and economic sciences has gone from 716 in 2024 down to 476 in 2025 and now just 91 at this point in 2026; obligated money has gone from $219M to $95M in that time. The same could happen for biology.

I suspect that it’s worth having a rule of thumb for such an eventuality. I think a “rule of three” could also be useful when prioritizing science funding. One main mission for NSF is maintaining US science capacity; part of that is keeping people from being forced out of science. A quick rule of thumb based on how quickly people will be pushed out if they don’t get funding:

• 3 months: a graduating PhD student, current postdoc, or soft money staff member who can’t find a position
• 3 years: a pre-tenure faculty member who can’t get funding for their research once their startup ends
• 3 decades: a tenured faculty member without grants

The rule of three for funding is probably even less accurate than for emergencies (“But I got a grant only 19 years ago – sure there are some undergrads older than my last grant, but it’s unfair to say all”). But for an objective of maintaining the scientific workforce, it’s a decent rule of thumb. If we think of “where will people in science get money for food,” for undergrads it’s not from NSF (there are exceptions, like REUs, but while important for training and inspiration it’s not the case that a large proportion of biology undergrads are directly supported by NSF). For grad students, support comes from serving as teaching assistants, research assistantships, and dedicated funds like the NSF Graduate Research Fellowship. NSF support is critical there in creating opportunities for training the next generation of great scientists, and I was glad to see that the GRFP program is continuing. When those students graduate, for those who want to take an academic route, one of the only options is grant-supported postdocs; in basic science, NSF is a major contributor (it probably funds the vast majority of such postdocs, but I don’t have those data). Then, for faculty starting their labs in basic science, NSF is again one of the major funders. Though it shouldn’t be (see my rant on this here), getting external funding is seen as a requirement for tenure in many places, and the reality is that a lot of work does require funding to be feasible. People can get by for a few years with startup funds, but by the time they’re coming up for tenure they might be rejected if NSF hasn’t come through. Grants do tremendously help tenured faculty (I’ve benefited greatly from them, and largely used them to support excellent postdocs) but while not getting grants hurts the science they can do, it doesn’t lead directly to “not having a job” in the same way it does for grad students looking for a postdoc or assistant professors failing to get tenure.

Note that the “maintaining the scientific workforce” criterion is not the only metric that should be used (or which is used). “Supporting the best science” also matters greatly, and often can be in conflict with the first criterion. I’m far faster at doing work and writing papers than I was when I was an assistant professor, and I think I’m better at mentoring, too. Giving me a grant now is going to get more science and impact per dollar than giving it to a younger equivalent of me. But we might not have current me if NSF hadn’t been so good at supporting earlier me. An important consideration is timescale: to get the best science out this year, ignoring career stage makes sense; to get the best science out ten years from now, we can’t grind our seed corn: we need to protect early career folks.

What to do

A few suggestions are below, but not all will be feasible. And it’s also important to recognize that the support that does exist (postdoc research fellowship program, in biology, graduate research fellowship program, CAREER, the core programs, and more) also reflects a ton of public and behind the scenes work – I’m sure there’s been lots of pressure to cut these even further, and so there are a lot of quiet wins for science already. None of these suggestions are “work harder” – the people involved are already maxed out, I suspect. But these might be policy changes that can help.

• Set a beginning investigator floor: NSF already tracks “beginning investigators,” those without previous NSF grants. Look at the average annual number of grants to beginning investigators from 2017-2024 (across two different administrations, so it’s a bipartisan average), and use that count (not the proportion) as the intended floor for the next grant cycle. For example, if the average was that 3,000 awards annually had a beginning investigator, have that the target this year, even if the total number of awards possible drops to, say, 4,000. The tenure system was by no means easy or fair in 2017-2024, but it was sufficiently functional, with new faculty able to get grants. This keeps it working for those people first. It also creates an incentive for more senior faculty to mentor and collaborate with junior faculty, as junior faculty-led proposals will have a clear greater chance of getting funded. I know junior people still are getting funded, so maybe this is already in place (or just reflects a preference for proposals from beginning investigators), but making it a public commitment would ease some concerns from junior faculty and encourage cross-generational collaboration.
• Report person-months by career stage funded by a proposal: I’ve led a couple of well-reviewed but ultimately not successful NRT proposals, which are focused on graduate training, and a big part of those is listing how many students will benefit directly from program funding. We already compute person-months in our NSF budgets for the postdocs, faculty, staff, and more. Just extract that as a field and 1) provide it as a table to reviewers and 2) include it in the API output so places like grantwitness or rnsf can consume it and show it. A reviewer can say, “Proposal A costs the same as Proposal B but supports an undergrad for a summer, 24 person-months for PhD students, and a three year postdoc, while B only supports summer salary for the PI.” It could be that Proposal B has more impactful science and should still be supported, but a table like this would just surface the information about sustaining the workforce in a way easier to parse. Having this information returned in a machine-readable way from the API or grants search page will help build the case for funding. NSF already tracks and releases info on which congressional districts are associated with a grant; this just lets local media and others say, “In our district, grants supported 15 postdocs, 20 grad students, 50 undergrads, all doing cutting edge STEM research.” Heck, if this becomes available in an API, I promise to make a web page for each state and congressional district in the US listing this info (I do something similar for info on US colleges already (it even got in the NYTimes), I already wrote an R package that consumes NSF’s API – launching something like this would be just a fun Saturday morning for me).
• Continue re-tuning existing programs so they maintain support: Historically, the Postdoc Research Fellowship in Biology has been a great way for new PhDs to pursue important research. This year, it is narrowly tuned to the “Intersection of Artificial Intelligence and Biological Sciences to Strengthen and Safeguard Biotechnology Innovations.” I have no actual knowledge of what caused the focus, but I have guesses (these areas seem popular with those making decisions above NSF). I personally think that having a broader focus is more likely to lead to more important discoveries, but if I had a choice between arguing for a narrow focus to keep the program alive versus having a broader program killed, I’d choose the former (rule of three: protect the postdocs first). Not everyone’s research will fit in this lifeboat, but some will be saved, and that also leaves more seats open in other lifeboats. Maybe there are other programs that are feasible in this environment. A broad geology postdoc research fellowship program probably isn’t feasible, but what about one targeting discovery of critical minerals?
• Support NSF folks doing the work: People go into science to make discoveries; some people then temporarily (rotators) or permanently put aside most or all of their own discovery work to become managers helping to distribute funds in the best way possible, just to enable our community to do even more work. “Best” is very hard, and we all have slightly different understandings of that. And even in the happiest of times the job requires saying “no” to 80% or more of the proposals that come in, often rejecting months of work by teams of colleagues. I don’t know what it’s like working there now, but I do know that many people have left, some were fired and some eventually rehired, the agency lost their whole building, and more – I suspect it’s even more fraught than usual. There have been changes at NSF that many colleagues outside NSF haven’t liked, but I think we have to assume that often these are out of the hands of, and have even been argued against, many of the people who we contact most commonly at NSF. For example, under the new proposal guide reviewers no longer see safe fieldwork plans – these only go in to NSF after a proposal is moving forward for funding. I think safe fieldwork is important, and it involves more dimensions than just physical health (see more), and I believe having reviewers see the plans helps encourage better community norms. But approaching NSF with, “why don’t you care about safety anymore?” is less kind (and likely to be less effective) than “I appreciate the steps NSF has taken to encourage field safety, but I worry that with the new policy…”. It gives the people we’re talking with info to pass up the chain that “the community really wants to continue peer review of safety plans” and also invites a conversation – I look at these plans periodically as a reviewer but the people in NSF have a better sense of how they work in aggregate, what the constraints are, and perhaps how past arguments for and against review have fared. I remain impressed by the professionalism throughout NSF: they all seem to be faithfully executing official policy (for example, I’m sure that even if they had argued internally for a broader PRFB, once the decision for an AI+biotech focus came down they will work hard to evaluate proposals using that metric). I believe we have to trust (or at least use as a very strong prior) that they are making the decisions we would make if we were in their shoes, because they basically are samples of us (though their population is enriched for those willing to perform this service full time). Our colleagues working at NSF face highly constrained optimization problems, often with barriers that those of us on the outside don’t see. So while I think it’s good for us to give suggestions and ideas, and even pushback on policies we see as problematic, it has to come from a place of gratitude to those who are doing this work every day.

Thanks again to the people at NSF for all you’re doing to sustain science and the people who make it happen.

---

To subscribe, go to https://brianomeara.info/blog.xml in an RSS reader.

Footnotes

1. The lower grant number but less reduced number of awards comes partly from how NSF is changing how it awards money for multi-year grants, now tending to award it all at once rather than releasing it a year at a time. This gives NSF more flexibility for the future: rather than having a lot of their funds obligated to pay out for grants awarded in past years, it could eventually just pay for grants of the year.