Government Science and the Prestige Economy

Alternative status hierarchies can maximize the return to human capital

Should government fund science? We may believe that the private sector can produce the right levels and types of scientific funding, for normal capitalist reasons. Businesses have incentives to pursue research paths that will be profitable, unlike government, and may carry out their work more competently.

The most common argument for state involvement is that there are certain kinds of basic research that the market will underprovide. There can be knowledge that is expensive to produce, but with diffuse potential benefits that no one firm can sufficiently capture. The state is therefore available to solve a collective action problem. It helps pay for the research that no company has an incentive to provide but that will hopefully make society better off.

I would argue that there is furthermore a category of research that may never have an economic payoff but is valuable for its own sake. My go-to example here based on my own interests is the David Reich lab at Harvard, which uses ancient genomics to reconstruct the deep human past. It costs a lot of money to find ancient fossils, and develop new ways to extract DNA from them and models to interpret the resulting data. Does anyone become tangibly wealthier when we learn which modern races have Neanderthal or Denisovan blood, or that the caste system has created substantial genetic stratification in the modern Indian population? Probably not. But I still want to know these things.

The Large Hadron Collider (LHC) in Bern took 10 years and cost nearly $5 billion to build. It was a project of the European Organization for Nuclear Research, which is funded by its 23 member states, along with certain non-members including the US and Japan.

In 2012, physicists working with the LHC discovered the Higgs boson, the last missing piece of the Standard Model of particle physics. This has no practical applications for now, and although it might in the future, it is impossible to do a cost-benefit analysis to know if the LHC will end up being worth it in purely economic terms. There is currently a debate over whether to build the Future Circular Collider (FCC), which would be three times the size of LHC and cost $30 billion. The time horizons here would be very long, involving a two-step plan that would begin with preliminary experiments in 2045 and end with ultimate completion in 2070. Think here also of the NASA search for distant exoplanets that may contain the signatures of life, or the James Webb Space Telescope, which launched in 2021 and has been peering back to the first galaxies while refining our theories of cosmic evolution.

So we have two arguments for science funding: the need for basic science, which can be justified through appealing to its potential economic benefits, and the idea that fundamental knowledge about the human story and our very universe that has no market value is good in and of itself. I think these are both compelling arguments, but I would add one more consideration, which is that creating different status hierarchies can be a good way for societies to maximize human capital output.

The Wrong Way to Do Fiscal Conservatism

Before we get there, however, we need to consider some of the conservative and libertarian critiques of the idea of government-funded science. First, one often hears a case based on fiscal conservatism. Yet I think that this attitude doesn’t hold much water at all, given that the amount of money spent by the federal government on science is pretty small in the grand scheme of things. The National Science Foundation (NSF) has an annual budget of around $9.5 billion, and the National Institutes of Health (NIH) — by far the largest funder of biomedical research — spends under $50 billion. Even when you include all federal research and development spending — across agencies like NASA, the Department of Energy, and the Department of Defense — it amounted to just under $200 billion, or about 3% of the total federal budget, in 2023. For the sake of comparison, we will spend $1.6 trillion on Social Security payments this year. We spend $130 billion a year on food stamps, though the Big Beautiful Bill would bring that number down.

We could perhaps also consider some funding that goes towards higher education as public support for science. State and local governments spend about $300 billion on higher education a year, and, the Department of Education pays out $121 billion in student grants and loans. Even if you include all government higher education spending as funding for science, take that number, and add in the official R&D figures, you get government spending on science at all levels to be about $620 billion a year. That’s still significantly less than the defense budget, which is expected to soon exceed $1 trillion. Of course, it would be crazy to count all higher education spending as support for science. A lot of it goes towards funding the humanities, and the aims of education spending are generally to benefit the students and help their career prospects rather than push the frontiers of knowledge. Perhaps financial aid helps educate an engineering student at Purdue who eventually becomes a great innovator based on what he learned. But providing grants and loans for STEM majors is a very indirect way to support “science” and not what we would choose to do if the main goal was to produce valuable research.

All of this is to say that the fact that we spend so little on science in the grand scheme of things should influence how we analyze the issue. Social Security isn’t the worst government program in existence at a conceptual level, since there are things government does that are much more damaging than cutting checks, like banning the construction of new housing. But the fact that entitlements are such a huge part of the budget should make them the targets of informed fiscal hawks. Of course, there are people on the right who love to go after silly sounding research grants and imply that they are bankrupting the country while ignoring Social Security and Medicare, but they are either uninformed or demagoguing the spending issue.

Crowding-Out?

The second argument against government funding of science is that it “crowds out” superior private sector alternatives. The bottleneck here must be talent. Human capital is the most important input into scientific endeavors, and also a limited resource as there are only a certain number of people in society with the potential to be brilliant, or even adequate, researchers and scholars. If government is simply less competent than the private sector is at supporting research, then talent that could otherwise be better utilized by businesses will be wasted in academia and government labs.

The crowding out thesis can serve as a response to most research that claims to find a large positive return to government funded science. For example, Galkina Cleary et al. (2018) write that,

This report shows that NIH funding contributed to published research associated with every one of the 210 new drugs approved by the Food and Drug Administration [emphasis added] from 2010–2016. Collectively, this research involved >200,000 years of grant funding totaling more than $100 billion. The analysis shows that >90% of this funding represents basic research related to the biological targets for drug action rather than the drugs themselves. The role of NIH funding thus complements industry research and development, which focuses predominantly on applied research. This work underscores the breath and significance of public investment in the development of new therapeutics and the risk that reduced research funding would slow the pipeline for treating morbid disease.

What the authors did here was basically search for the name of each new drug, and when necessary the target it operates on, and see whether there were NIH-funded papers associated with them. The result: “NIH-supported publications were identified in 198 of the 210 drug searches and in all 151 target searches. Thus, NIH funding was directly or indirectly associated with every one of the 210 [new-molecular entities] approved from 2010–2016.”

This seems like a big deal. It doesn’t necessarily get at the crowd-out effect though. Perhaps all those scientists who got NIH grants could have spent their time better by, say, filling out less paperwork and taking directions from hyper efficient firms, and we would’ve had even more and better drugs without state support or involvement. We can say this about almost any study that purports to show a positive return to scientific funding.

How does one test the crowd out effect? The best work I’ve read on this topic is a 2015 paper by Azoulay et al called “Public R&D Investments and Private-sector Patenting: Evidence from NIH Funding Rules.” To understand what the authors did, it is necessary to know something about how NIH funding works. The NIH is composed of 27 different Institutes or Centers (ICs). They can be organized around body systems or disease areas, like the National Cancer Institute. Before grant applications get to the ICs, however, they go to one of 180 standing review committees, which are called study sections and organized around scientific topics such as “Behavioral Genetics and Epidemiology” or “Cellular Signaling and Regulatory Systems.” The authors are therefore able to use as their unit of analysis a D/S/T for each type of research area and time period, because they can isolate the disease (D), science (S), and time (T) of every application. For example, if a scientist sends in an application in 1990 that goes to the standing review committee for Cellular Signaling and Regulatory Systems and then the IC for cancer, we can come up with a DST marker for that application and others in the same category.

When applications are received by a study section, they get a raw score between 1.0 and 5.0, with 1 being the best. Raw scores are normalized within a study section and converted to a percentile, which are considered a “science rank.” Once this happens, the relevant IC funds applications in rank order, where ranking is determined by comparing science ranks across applications within the same body system or disease area. In this way two applications can have the same raw score for scientific merit, but one might get funded while the other doesn’t. Likewise, two DSTs might have similar scientific potential while one gets more funding than the other. As the output variable, Azoulay et al. looked at biomedical innovation as measured by patents and the papers that they reference, which in turn reference grants. The authors' main innovation is to use grants that fall just above the funding threshold (a “windfall”) within each IC's payline window as an instrument for DST funding, while controlling for raw score and science rank. Here’s a figure showing how it works.

This is all very complicated, but the point to keep in mind is that the methodology allows us to compare outcomes in research areas based on funding, after controlling for scientific merit. In the example above, you might expect cancer-cell signaling to lead to more patents than cancer-tumor physiology because there are more quality grants going to the first; the methodology used allows us to estimate how much the windfall funding makes a difference.

To get at the issue of crowding out, Azoulay et al. count private sector patents in a research area, including those that do not reference an NIH grant. It may be that public investment simply displaces private research that would have happened anyway. To investigate this, the authors look at not just those patents that directly cite NIH-funded publications, but also all patents in the broader intellectual neighborhood of a DST, even if they do not reference NIH work. This includes patents citing similar publications, derived from keyword overlap, regardless of citation chains. By measuring total innovation activity in a field, they can determine whether NIH funding increases the net number of private-sector patents or merely substitutes for private investment.

They estimate that a $10 million increase in NIH funding for a research area leads to 2.3 additional private-sector patents. If crowding out were a significant factor, the increase in total patenting would be negligible or zero. Instead, the fact that the total number of related patents exceeds the number of ones that are directly linked implies some degree of crowd-in: NIH funding not only stimulates public science but also enhances the productivity of related private-sector R&D in the same research area.

Azoulay et al. also attempt to quantify the economic value of this effect. Under a conservative estimate — valuing each patent at $1 million — 2.3 patents imply $2.3 million in private benefit per $10 million in NIH spending, or about 23 cents on the dollar. Using a higher estimate of $11.2 million per patent, and accounting for the fact that NIH-linked patents tend to be cited more often, the return could be as high as $25–35 million in private value. These figures capture only the private sector return; they do not include broader spillover benefits or public health gains.

While this research design is impressive, I still think there’s a potential issue with crowding out. It may be true that funding a research area leads to more patents and drugs in that area. But might this just be a result of scientific energy being redirected to certain fields over others? Perhaps if the NIH funds, say, research on treating cancer through focusing on cellular signaling pathways, more scientists go into that field and produce important patents, but they are less likely to work on curing heart disease, and the total societal benefit of NIH funding may be small or even negative. Unfortunately, the paper has no way to address this concern. Most biomedical scientists are not so narrowly specialized that they do not have room to shift their focus based on funding priorities.

Perhaps the most prominent advocate of the crowding out theory is Terence Kealey. Based on this article, his 1996 book The Economic Laws of Scientific Research, and this debate, he mostly cites comparisons across eras and historical and contemporary cross-national data. Kealey argues that before we had massive scale government funded science, innovation was happening at a faster rate. Also, the US and the UK had higher GDPs per capita than Germany and France through the first and second Industrial Revolutions, despite the latter two nations spending more on basic science. On this last point, this reasoning doesn’t rule out a spillover effect, in which they were subsidizing work that could be utilized by the rest of the world. Germany in particular was considered the leading scientific power in the late nineteenth and early twentieth century. The American drug giant Merck & Co, for example, was originally a US subsidiary of the German company Merck. Greater economic growth in the US and UK could have been due to other factors, like for example the fact that America had a much lower level of government spending as a percentage of GDP.

Just looking at who spent more on science and then connecting that to economic health isn’t a very convincing methodology. There’s simply too much going on to draw straightforward conclusions.

I therefore consider the Azoulay et al. paper to be decent evidence with some caveats, and Kealey’s arguments suggestive but in the end not very compelling.

Prestige Economies Are Powerful and Necessary

Despite the empirical evidence being unclear, I think that the facts that we spend so little on science and that the returns are potentially so large should incline us towards supporting more funding. The idea that government is just worse at doing things is one that I accept as a general matter. But so is the idea that there is basic science that might have outsized returns.

Two considerations, however, push me over the edge towards the pro-funding side. First of all, as alluded to above, I see certain forms of knowledge as good in and of themselves. Genetic paleohistory, behavioral genetics, historical linguistics, particle physics, cosmology, and other sciences that are fundamental in telling us who we are and explain our place in the universe are valuable for their own sake. I don’t care if the average voter would rather have food stamps or more giveaways to the elderly, or even tax cuts.

I also think that there is a flaw in the crowding out theory, in that it assumes that talent is interchangeable between government-funded research and the private sector. This misses an important source of variation in human nature. As I’ve previously written:

In a broad sense, there are two things that smart and hardworking people can do with the opportunities that wealth and modern levels of technological development provide. They can try to maximize income, wealth, and control over tangible resources, or seek meaning through their careers. Think of someone who becomes a petroleum engineer and works his way up to executive level management, in contrast to an individual who spends the entirety of his twenties earning a philosophy PhD, and then perhaps another half decade looking for his first real job. Nobody is purely a wealth-maximizer or a meaning-maximizer; most of the upper class is somewhere in between, but that doesn’t mean we can’t speak in terms of different prototypes.

Take a professor of medicine who works for a R1 research institution. Does he apply himself just as much if forced into the private sector? In some cases yes, but not in others. This seems right when I think about my own motivations. I work very hard doing research and writing articles. If I applied my intelligence with just as much effort to a more lucrative corporate job, I’m sure I could produce a lot of economic value for society and end up wealthier. Yet I don’t think I would put in nearly as much effort towards becoming some corporate bigwig at Walmart.

My experience tells me that a lot of academics and researchers are like this. Their utility function is simply too different from that of people who work in the private sector for there to be a complete crowding out effect from government-funded science. Does the guy who wakes up every day excited to explore string theory or the nature of the early cosmos put in the same hours to make an iPhone camera incrementally better? In some cases, maybe, but definitely not in others.

Recently, Tyler Cowen spoke with brain surgeon Theodore Schwartz, and the host noted that his guest had produced an article or book chapter every two weeks for 35 years. Despite Schwartz already being a tenured professor, his Google Scholar page lists 11 items for 2025 alone as of this writing.

In academia and across many different fields, you will often find people who pile up citations like this. I don’t know about Professor Schwartz, but it seems to me that many who succeed in academia put in a remarkable amount of effort year after year contributing new articles and research in order to impress their peers and hope that they will be remembered after they die, even though the marginal returns for any particular article are very small.

Personally, I want to be remembered too. I’m glad I have a Wikipedia page, even though it’s largely a hit piece. History will understand me better. Maybe it’ll just be my grandchildren and a few scholars of early twentieth-century American thought, but it beats one’s memory falling into complete oblivion, the fate of the vast majority of humanity.

I remember once meeting a very nice middle-aged couple waiting for a table at the Cheesecake Factory. They were some kind of hybrid East Asian pairing, and they had four sons. The couple had the sense of wonder and open friendliness you often find among some religious communities, particularly Mormons. The husband told me that he did some kind of logistics work for a pistachio factory. I remember thinking that if that were my job I would not be nearly as cheerful, as this seems like it would provide absolutely no sense of meaning.

But then I thought about this in the context of my sense that he was probably religious, which likely provided all the enchantment he could possibly want out of life. The point here is that some people who are functionally wealth maximizers behave as they do not because they don’t have a need for meaning, but because they get it from outside their job, most commonly through religion. If you think God has put you where you are for a reason and is constantly sending you signs and messages and looking out for your well being, and you in turn are glorifying him just by spreading the word and forming a family, then even the pistachio factory can be a fun place.

For those of us who aren’t religious but need meaning in life, the prestige economy maximizes the potential to do good. This is perhaps a benefit of the peer review system, where individuals are judged by members of the community that they care about. Government-funded science similarly distributes money through the decisions of committees made up of fellow researchers and academics. Prestige economies have a downside, of course. The antimonopoly movement, gender studies scholars, MAGA influencers, and antifa rioters are also participants in their own alternative status hierarchies. But we should think about how to make sure that the prestige economies we support are ones that are pro-social. It is good to have a status hierarchy where someone gets prestige for figuring out how to cure cancer, but not one where you get ahead by coming out on top in the Based Ritual.

In terms of motivating behavior, prestige economies are powerful forces. Usually not as powerful as markets, since most people primarily want money and are most comfortable in status hierarchies in which wealth determines one’s position. At the same time, there is a substantial, disproportionately high IQ and agentic, portion of the population that needs a prestige economy in order to truly flourish. A centimillionaire will on average attract more attention from women than a Pulitzer-prize winning reporter, but the latter will do better with those who have opted-in to the same status hierarchy.

I don’t think a pure libertarian model can explain how much innovation actually comes out of government, including certain communist successes in science and engineering. The Soviet Union, despite its lack of private enterprise and market incentives, produced an extraordinary range of scientific achievements: it was the first to launch an artificial satellite (1957), the first to send a human into space (1961), and maintained world-leading research in fields like mathematics, theoretical physics, and nuclear engineering. Figures like Lev Landau in physics and Andrey Kolmogorov in probability theory made foundational contributions that are still central to modern science.

Of course, these regimes are universally bad at improving the living standards of their populations. But when it comes to getting smart people to do amazing things at a very highly abstract level, the story isn’t nearly as bad. In addition to Soviet successes, the American university system just in the last few decades has seen remarkable breakthroughs like CRISPR gene editing, mRNA vaccine platforms, and the emergence of CAR-T cell therapy for cancer treatment. Perhaps all the scientists responsible for these accomplishments would have done things just as amazing if forced to work in firms, but it seems quite risky to assume this would be the case.

I was talking to Bryan Caplan about this theory, and he argued that creating new status hierarchies divorced from market logic is exactly what we don’t want to do. Instead of being forced to engage in useful, productive work, academics and those employed by government research labs can set their own priorities, working on topics they enjoy, without regard for any potential societal benefit. I think that’s a potential danger, and this argument is a good reason for perhaps not funding the humanities and many kinds of social science as much as we do.

Yet I would argue that there’s probably a pretty strong correlation between how interesting scholars find a research area and the degree to which it falls into the category of basic science, or at least what I classify as knowledge that is good for its own sake. It seems like the utility academics in the hard sciences gain comes either from the inherent importance of a problem or from their belief that what they’re researching might one day change the world, which can bring them prestige and glory. On a funny note, when I told Ruxandra about his argument that scientists should be encouraged to only do things that provide economic value, she scoffed “Does Bryan Caplan provide anything of economic value?” Now, he has in fact written books that have sold well. But even if he limited himself to academic papers and blog posts, I’d still be glad that the government supports Bryan being able to make a living writing about why everyone should be a libertarian.

These arguments about crowding out are admittedly pretty speculative. As discussed above, none of the empirical research on the returns to funding science is that convincing. We can’t do controlled experiments, and comparisons between countries and eras have too many variables to be able to tell us much of anything. Yet based on my experience and thinking about the motivations of individuals who go into high-status, low-pay careers, the idea that being free to pursue their own interests will get a lot more effort out of them makes sense to me, as does the argument about basic research being under provided.

None of this means that government-funded science as practiced can’t be greatly improved, or that we shouldn’t stay on guard against all the ways public initiatives can be corrupted. Academics spend too much time on paperwork and the IRB and various ethics requirements are way too strict, a problem that happens to be worse in universities than the private sector. Here is a more obvious case of a crowd out effect, where smart and talented people are spending their time jumping through useless bureaucratic hoops. The ideal approach here is a combination of more spending on science in addition to simplified methods for distributing money and, once projects are funded, less regulation regarding what is done with it.