Science as a Public Good

Sessions 5 & 6 · Tue Sep 15 & Thu Sep 17

Why markets underprovide basic research, how scientific norms solve part of the problem, and the institutions that emerged from Vannevar Bush’s 1945 report.

A warning from 1898

On the evening of September 7, 1898, Sir William Crookes walked onto a stage in Bristol and told an audience of thousands of scientists that Europe was going to run out of food.

The problem was nitrogen. Crops need it to grow. Until the late nineteenth century, the world’s supply came mainly from guano deposits on Pacific islands and saltpeter mines in Chile, both finite, both nearly exhausted. Crookes had done the arithmetic. At current rates of population growth and nitrogen consumption, he calculated, the wheat supply would collapse sometime after 1931. “England and all civilized nations,” he warned, “stand in deadly peril.”

Then he said something unexpected. He did not call for population controls or rationing. He called for chemistry. Nitrogen makes up 78 percent of the atmosphere, he pointed out. If someone could find a way to fix atmospheric nitrogen into a form that plants could use, the crisis would be averted. He was describing, without knowing it, a problem that would define the next half-century of industrial chemistry.

Fourteen years later, the German chemist Fritz Haber and industrialist Carl Bosch had done it. Their process for synthesizing ammonia from nitrogen and hydrogen, the Haber-Bosch process, is now one of the most important industrial processes in the world. It consumes roughly 1 percent of global energy. Without it, roughly half the people alive today could not be fed. Haber won the Nobel Prize in Chemistry in 1918. Bosch won it in 1931.

Here is the question this session is about: Why didn’t a private company solve this problem before Haber?

The nitrogen crisis was well-known by the 1870s. The financial returns to solving it would have been enormous. Private firms had every incentive to invest in the research. And yet, for decades, nothing happened. It took a publicly funded chemist working at a German university, supported by government research infrastructure, to crack it.

To understand why, we need to look more carefully at what kind of thing scientific knowledge is.

Spend 5 minutes on the Crookes story before revealing that Haber solved it. Let students sit with the question “why didn’t a private company do this?” — write it on the board and come back to it after the public goods framework. This story sets up the entire session.

Connection to Crawford’s “Solutionism” (Module 5 reading): Crookes is the archetype of what Crawford calls a solutionist — pessimistic about the problem, optimistic about the solution. Worth flagging briefly.

What kind of thing is knowledge?

Consider two objects. The first is a bag of nitrogen fertilizer. If you spread it on your field, your neighbor cannot spread that same fertilizer on theirs. Economists say fertilizer is rival: one person’s use reduces what is available for others.

Now consider the knowledge of how to make nitrogen fertilizer — the Haber-Bosch process. Once Haber published his results, every chemist in the world could study them. Carl Bosch’s industrial scale-up at BASF drew on Haber’s published findings. Every fertilizer plant built since has drawn on the same knowledge. Haber’s use of the idea did not reduce Bosch’s, or anyone else’s. Knowledge is nonrival.

Knowledge is also often nonexcludable. Once a discovery is published (and the norms of science strongly push toward publication), it is very hard to stop others from building on it. You cannot un-read a paper.

Together, these features make scientific knowledge a public good: nonrival and nonexcludable. And this creates a fundamental economic problem.

If you cannot capture the full benefits of producing something, you will tend to produce too little of it.

This is why private firms underinvested in nitrogen chemistry for decades. The expected return to any single firm was far less than the social value of the discovery, because most of that value would immediately spill over to competitors, farmers, consumers, and governments worldwide. Basic research is especially prone to this problem: the gap between the private return and the social return is widest precisely because the knowledge is most fundamental and most freely shared.

Haber-Bosch was eventually developed by a private company (BASF), not directly by government. Does that contradict the public goods argument? What conditions allowed a private firm to succeed where others hadn’t, and what does that tell us about when markets can and cannot solve the problem?

Which of the following best explains why markets underinvest in basic research?

  1. Basic research is too cheap to be profitable.
  2. Firms cannot produce scientific knowledge.
  3. The social return to research exceeds the private return — benefits spill over to those who did not fund the work.
  4. Scientists have no incentive to do research.

Common confusion here: students often think “BASF did it, so markets work.” The nuance is that BASF benefited from Haber’s publicly-funded university research, from government-funded infrastructure, and from the wartime industrial mobilization context. The public goods problem explains underinvestment in early-stage research, not in commercial scale-up (applied research). Nelson (1959) is precise on this distinction.

What institutions filled the gap?

If markets underprovide basic research, two kinds of institutions have historically filled the gap: public funding and scientific norms.

The public funding story in the United States traces directly to Vannevar Bush’s 1945 report, Science, The Endless Frontier. Bush had overseen American wartime science (radar, the Manhattan Project, penicillin production) and had seen what organized scientific effort could do. Writing to President Truman in July 1945, he made a simple argument: the same logic that justified wartime investment in science applied in peacetime. The benefits of basic research, in health, security, and economic growth, were so large and so diffuse that private actors could never capture enough of them to justify the investment. Government had to step in.

Bush’s report led, eventually, to the creation of the National Science Foundation in 1950 and the major expansion of the National Institutes of Health. The pattern it established, government funding of investigator-led basic research carried out mainly at universities, shaped American science for the next seventy-five years.

But public funding alone doesn’t explain how science works. Scientific research is not centrally planned. No government agency decides what questions chemists or physicists should ask. So how does a decentralized system of researchers, scattered across thousands of institutions, manage to produce cumulative, reliable knowledge?

The answer, as Nelson (1959) emphasizes, lies in the norms of open science: results are published, methods are disclosed, priority goes to whoever publishes first. These norms look strange from an economic perspective (why would you give away your discoveries?), but they make sense once you see them as a collective solution to the public goods problem. Because scientists compete for priority and recognition rather than for secrecy, they have strong incentives to disclose. And because they disclose, knowledge accumulates. Each generation of researchers can build on what came before without having to rediscover it.

Bush’s model — government funds basic research; private firms develop applications — has been called the “linear model” of innovation. But is it accurate? Can you think of cases where the relationship between basic research and application ran in the opposite direction, where commercial pressure drove fundamental discoveries rather than following from them?

This is a great discussion for setting up the DARPA session later. DARPA explicitly rejects the linear model — it starts from a desired capability and funds the science needed to get there. The contrast is: Bush’s NSF model = investigator-driven (researchers ask their own questions); DARPA model = mission-driven (define the goal, fund the path).

Key institutional timeline for class: - Bush report: July 1945 - NSF founded: May 10, 1950 (five years of congressional wrangling — Bush wanted civilian control, Congress wanted political oversight) - NIH expanded: 1946–48 - DARPA (as ARPA): February 1958, response to Sputnik — a fundamentally different model

How much do governments actually spend?

The chart below shows government and private R&D expenditure as a share of GDP across major economies since the mid-1990s. Notice how the United States, Japan, South Korea, and Germany cluster around 2–3 percent, while the post-Sputnik surge in US spending — which peaked around 1964 at nearly 2 percent of GDP from federal sources alone — is not captured in data this recent.

Source: UNESCO UIS / World Bank via Our World in Data (2026). Requires an internet connection to render. Note: this chart shows total R&D (public + private). The public-private split varies significantly — the US federal share has declined from ~65% of total R&D in 1960 to around 20% today.

Two things to draw out from the chart: (1) the cross-country variation — South Korea and China have increased rapidly, while US federal R&D as a share of GDP has actually fallen since the 1960s; (2) the public/private mix matters as much as the total — a country could have high total R&D but most of it in commercial development rather than basic research. The Nelson public goods argument applies specifically to basic research, not to the total.

Connections

Builds on: The public goods framework introduced here runs directly from the discussion of externalities and market failures that students encounter in the UOE readings (Week 1–2). Knowledge spillovers are the clearest example of a positive externality that markets undersupply.

Sets up: The DARPA session in Module 4 takes the Bush linear model and asks whether mission-driven science does better. The Solutionism reading (Module 5) picks up the Crookes story directly and shows what happened next.

Arc note: This session is the economic foundation for two later arguments the course makes. First, that science policy choices have large, long-run consequences (Module 4). Second, that the right response to a resource constraint or market failure is institutional design, not despair (Module 5 — Crawford’s solutionism is exactly the attitude Vannevar Bush embodied). Students who leave without grasping the public goods argument will struggle to engage with both those later discussions.

Vannevar Bush submitted his report Science, The Endless Frontier to President Truman in 1945. Which institution did his report most directly lead to creating?

  1. DARPA (Defense Advanced Research Projects Agency)
  2. The National Institutes of Health (NIH)
  3. The National Science Foundation (NSF)
  4. The Manhattan Project

According to Nelson (1959), why do scientific norms encourage researchers to publish their findings rather than keep them secret?

  1. Publishing maximises the researcher's private profit from the discovery.
  2. It allows other scientists to check the work for errors.
  3. The norm of priority — recognition goes to whoever publishes first — gives researchers an incentive to disclose rather than hoard.
  4. Publication is legally required for research receiving public funding.

Review cards

Work through these cards now, then Orbit will schedule them for review over the coming weeks. Studies show that spaced retrieval practice produces dramatically better long-term retention than re-reading.

Reading guide

Required

Nelson, Richard R. “The Simple Economics of Basic Scientific Research.” Journal of Political Economy 67, no. 3 (1959): 297–306.

What to look for: Nelson makes a tight economic argument in ten pages. Track the logical structure: (1) basic research produces knowledge; (2) knowledge has public good properties; (3) therefore markets underprovide it; (4) therefore universities and government fill the gap. Does each step hold?

Key argument: The social returns to basic research systematically exceed private returns because knowledge is nonexcludable — competitors benefit whether or not they paid for the research.

Bring to class: Nelson was writing in 1959, a year after Sputnik and during the height of Cold War science funding. Does his argument depend on that context — or is it timeless?

Bush, Vannevar. Science, The Endless Frontier (1945). Transmittal letter + Chapter 3, “The Importance of Basic Research” (~8 pp.). Free at nsf.gov

What to look for: Bush is making a policy argument to a president, not a theoretical argument to economists. Notice how he grounds the case for public funding in concrete wartime examples before pivoting to peacetime. What does he assume about how science works?

Key argument: The benefits of basic research are so broad and so unpredictable that no private actor can capture enough of them to justify the investment — government must act as the funder of last resort.

Bring to class: Bush says the government should fund science but scientists should choose the questions. Is that the right division of labor? Who should decide what gets studied?