The real cost of science

In the two weeks since my last post, the US government shutdown has not ended. Federally funded science is slowly winding down as programs spend their last allocations. Less money means fewer discoveries, but exactly how many fewer? How much knowledge will not be created because of the shutdown?

It is hard to measure the relationship between a society’s investment in science and that society’s improvement because of those discoveries. There are a few reasons. First, the impact of any one discovery is hard to predict, and may only become clear in the long term. A mathematical advance written on a napkin with pencil may have negligible cost, but lead to applications in other fields in the following decades. On the other hand, billions of dollars invested in human disease research may lead to negligible improvements in public health. Second, measuring improvement is a difficult task – is it increase in human lifespans or income, prevention of war, or something less definable? For these reasons, funding agencies typically focus on target areas (e.g. cancer treatment, nuclear weapons, biodiversity conservation) and spend widely and broadly until the aim is achieved, or until societal interest refocuses on other topics. It’s not easy to guess how much money will be needed to solve a given problem.

But one way to measure societal benefit is by scientific publication rates. More publications presumably mean more discoveries. How much money does a new publication cost society? Scheiner and Bouchie, who are officials at the National Science Foundation, just answered this question. In their paper, they took the total amount of funding for environmental biology research (ecology) and divided it by the numbers of publications resulting from this work, a number which all federally-funded researchers have to report. A scientific publication costs approximately 34,000 US dollars. In contrast, the state of California pays approximately 47,000 dollars per year for each prison inmate, and 8,700 dollars per year to educate each student.


When I read their number, I thought it was high – surely the average study doesn’t require so many resources. But many kinds of expenses are factored in to the NSF number. For example, most universities take approximately 50% of each federal grant in overhead, to pay for administration, pensions, building upkeep, utilities, and so on. That means the $34,000 represents something closer to $20,000 in ‘actual’ expenses. Salaries also come out of this number, leaving something probably closer to $10,000 in research money, based on allocations in grants that I’ve seen funded. So where does the rest go?

An ecological theorist may appear to not need any more resources than a pencil and paper (and generously, a wastebasket). However, they still have costs. Attendance at conferences requires registration fees and travel expenses; publication in open-access journals can cost multiple thousands of dollars. Computers and software resources have to be purchased. Student researchers have to be paid.

For field or experimental ecologists, the biggest source of expenses for ecologists is in the prosecution of their work. Field trucks, station fees at biological field stations, chemicals, glassware, notebooks, GPS units, plane tickets to remote parts of the world, research permits, and the list goes on. Here are two examples from my recent work in Peru paid by Yadvinder Malhi:

costs 1

First, weather stations – here, an aerial tower installed in a remote location in the eastern Andes, probably costing several thousand dollars to build, more to ship to this country, and even more to transport and install on the side of a lonely mountain and maintain each year.

costs 2

Or second, a large pile of timber, purchased to build a research platform, plus the services of dozens of people to carry that timber miles into the forest, and then later to assemble the structure.

These expenses all add up, and get us closer to the $34,000 number. To me, as a finishing PhD student, the numbers seem large, but I suspect they will begin to seem smaller as I dream up increasingly complex and large-scale projects. I also don’t know how they compare to other fields. My intuition is that fields like molecular biology, genomics, and experimental physics have far higher per-publication costs, but I don’t know for sure.

The next dangerous questions then become: how much impact do publications in each field make? Would it be worth refocusing research spending on low-cost, high-impact science? Does scientific spending produce more benefit per cost than my other two examples (prison incarceration or K-12 education)? Or are there some areas worth investing in, whatever the cost? I’ll leave that more complex issue for another post, and instead hope that there will soon be federal money for any of these areas.

4 Comments Add yours

  1. Nice post Ben. I should read their paper for more details and specifics. My first reaction is that the ‘mean’ may be based on a long-tailed distribution for project costs. For example, if their sample includes studies with heavy infrastructure like Astronomy then the real cost may reflect those expensive fields. Ecology in general is much cheaper (and theory even cheaper!). I did a quick back of the envelope calculation for funding to our lab and total publication output and I get a number around $15-8K per publication (total grant funding divided by publications). On face value then, ecology appears much cheaper or we are more efficient (!). This number includes the cost of University overhead (51%) so the actual amount funding the science is definitely lower on the order of $7-4K. However, that overhead does come back to benefit us as our lab does rely on infrastructure at the university and elsewhere, mainly funded by the NSF or other governmental institutions, that is not included in my estimate. Also, our lab has been enabling science (and publications) of others via access to data (,, as well as various code published etc.etc.) plus we do public outreach and teaching on the side so these numbers don’t reflect the synergistic effect of NSF investment in our lab. So, the straightforward direct calculation (NSF funds in) / (publications out) is not as a direct measure of cost per benefit as one would like. Most of these costs, as you mention, are actually ‘personnel costs’ – salaries etc. This means that NSF funding does go into some rabbit hole – we are funding people and the livelihoods which have a direct and immediate economic impact. Nonetheless the important point remains – Science is not cheap and leadership in Science critically relies on solid government assistance.

    1. bblonder says:

      Brian, thanks for your thoughts. I agree with all of your points about the indirect benefits of NSF support, and the key role of government leadership.

      The data from that paper are entirely from NSF’s DEB, so it’s exciting if the lab is coming in at a lower figure than the $34K average. For my own work, a back-of-the-envelope calculation comes in at about $13000/publication – including my graduate fellowships and all the grants I’ve received to-date. I think that in general you would be hard pressed to find any graduate student doing work that costs more than $34K/study!

  2. Aaron Hogan says:

    A very thought provoking post there Ben! As usual, it is a pleasure reading your posts.

    I read recently that the next generation is termed the “starving generation”. I think back to the Baby boomers, Gen X’ers, Gen Y’ers, and others and it appears that they did not have it all that bad compared to what is in store for future generations. They are calling this new generation the “starving generation” because they will be forced to create an increasing amount of opportunities with resources at all-time lows. But before I start sounding like your grandpa, I would like to point out the recent technological advances and benefits that have arisen in the last few decades. The challenge is being able to implement these and undiscovered technologies in creative and novel manners to help solve the starvation problem. We know that we (scientists) need to be subsidized to accomplish this, and with an increasing squeeze on resources, competition for funding is off the charts. I do advocate for cost-effective science that pays dividends. Perhaps, we can try to conserve science a bit by taking conservative approaches to answering scientific questions, the same way previous generations had an obligation to do the same. In theory it sounds like something everyone would do, right? However, history has told and continues to tell a different story. Regardless of whether we want to or not, all sectors will be starved a little because none of them is willing to take one of the team, arguably science included. The ability to working together across disciplines and narrow interest, under increasingly limited resources availability, will prove to be trying task into the future.

    1. bblonder says:

      Thanks for writing, Aaron. I worry about your conservation-of-science idea, mostly because I think we’ve collectively built a world where we look to technological solutions for world problems, like the starvation issue you mention. We need to sustain scientific investment to increase knowledge faster than that knowledge itself creates more problems. Perhaps there are other solutions – in your example, voluntary reduction of fertility rates, more wealth sharing, whatever – but it seems clear to me that we are not good at solving our problems in that way. Rather, we need to keep chasing new ideas to provide new solutions. When we ration these ideas, we let the problems get ahead of us. The long-term outcome of this game is hard to predict – collapse, would be my best guess – but I don’t see a better way forward.

      That’s all written in generalizations. Anyway, I agree with you that we need cost-effective science to produce dividends. But I also think there is an important space for ‘blue-sky’ research to think up unthunk ideas, to push forward the conceptual boundaries of human inquiry so that some later scientist can piece together these paradigms and make a practical innovation. The balance between these aims, of course, is hard to choose…

Leave a Reply