Skip to main content

Wartime Innovation: Lessons From the Office of Scientific R&D

A look at the "government startup" that drove U.S. science and technology during World War II.

The modern U.S. research enterprise has some of its roots in World War II, through the U.S. government's Office of Scientific Research and Development or OSRD. Quickly authorized by President Franklin Roosevelt in June 1940 only days after the fall of France, the entity that became OSRD was tasked with pushing the technological envelope to support the coming war effort. Over the next four years OSRD managed a wide-ranging research portfolio that produced major advances in fields as diverse as radar, nuclear fission, optics, rocketry, jet propulsion, electronic control, vaccines, antibiotics, antimalarials, and human physiology.

In ongoing work, Daniel Gross of Duke University's Fuqua School of Business and Bhaven Sampat of Columbia University's Mailman School of Public Health are exploring OSRD's structure and success, and the long-term influence it had on American science, innovation, and manufacturing. We recently got to sit down (virtually) with Gross to talk about their recent paper covering OSRD's formation, function, and some major takeaways.

What was the state of American science before the war, and the state of federal support for that science?

Daniel P. Gross, Duke University
Daniel Gross, Duke University

I think the best starting point will be to position ourselves in 1940. People who are in the know recognize the U.S. is going to be drawn into this war, although the public may not know it yet. In June of 1940 four top U.S. scientific administrators – you have Vannevar Bush, president of the Carnegie Institution of Washington, former vice president of MIT; Carl Compton, president of MIT; James Conant, president of Harvard; Frank Jewett, president of Bell Labs and of the National Academies – they all come together realizing that the U.S. military is lacking the technological sophistication it would need to be successful in the war, and they take this concern to Roosevelt. They're essentially authorized on the grounds of a one-page proposal to create the National Defense Research Committee [NDRC], and to use the President’s discretionary budget to start funding research. A year later, this effort was expanded into the Office of Scientific Research and Development or OSRD.

This was a radical move for its time. In 1940 the U.S. government wasn’t a significant funder of R&D, even though we have the sprawling federal science funding apparatus today. There was some research being performed in military scientific labs. There was some funding of agricultural research through the USDA. But it was peanuts compared to what happened during World War II. At the time, research was primarily performed inside firms or funded by philanthropies like the Rockefeller Foundation.

That said, some important fundamental advances in science had nevertheless been taking place in the first part of the 20th century, especially in the run-up to World War II. This was the case in chemistry. It was the case in physics. Much of it was coming from Europe, and especially from Germany – remember that Germany was at the frontier of science prior to World War II. And prior to World War I and in the interwar period, the U.S. scientific establishment was deepening. For the most part, it wasn’t federally funded.

What are some of the most notable characteristics of the OSRD model that allowed it to succeed?

I think you have to start with the fact that OSRD just had exceptional administrative and scientific talent working for it. I wouldn’t want to downplay the importance of research management here. Vannevar Bush was an outstanding scientist, administrator, and ambassador of science, and that kind of talent is rare. What you had on top of that [were] the best scientists in the country working in a decentralized but ultimately coordinated way. You had Nobel Laureates past and future working on problems of military importance.

Vannevar Bush, Head of OSRD

One really notable feature of the agency is that it was very entrepreneurial. There was no template to follow. It had to create not only the organizational structure that it needed to succeed, but even, for example, a funding mechanism. There was no such thing as a federal R&D contract. There was no precedent for procuring research – they ultimately weren't procuring an output, they were procuring research services that may or may not succeed. And so they had to come up with the contract language for that.

More broadly, it was a nimble and innovative organization. There was relatively little red tape and few administrative hurdles. I think of it as a government startup, which is ironic because although it was born from a one-page proposal and a committee of eight, it grew to a staff of 1,500. So, it was a fast-growing startup, but it was always willing to take risk and sufficiently funded to afford to. It was also less worried about waste or the equitable distribution of research funding across contractors, and more focused on performance.

I don’t think you can talk about OSRD without highlighting the pervasive coordination that took place across research efforts that OSRD itself managed, across agencies, internationally with other Allied countries, and ultimately with the military user, which not only helped OSRD identify problems to work on but also get feedback on the technology it was delivering.

You quote Bush saying, “Most of the worthwhile programs originated at the grass roots.”

Exactly. And this is an interesting contrast to the peacetime research funding model, in which NIH and NSF fund proposals that are developed by individual investigators and reviewed by panels of peer reviewers. There was some of this at OSRD too. Medical research funded by the CMR [Committee on Medical Research, an OSRD division] was typically investigator-proposed. But the bulk of OSRD funding went to applied research and technology development, where you had staff at OSRD collaborating with military liaisons to figure out what are the important science or engineering problems and how are we going to develop a plan of attack. For example, soldiers can’t see the enemy in darkness. How can technology change that? And that plan of attack would involve a specification of a problem, identification of potential contractors, and then OSRD could leave it up to them to figure out the best solution. That’s the researcher or engineer’s role.

I’ll mention two more things. One is that OSRD effectively had a single buyer, the U.S. military. The other thing is that they took an end-to-end approach to R&D management, supporting everything from research and development, to manufacturing, to deployment. There may have been advantages to this integrated approach. It was self-reinforcing. You had researchers who were thinking about how their technology would actually function in the field. Some of them were actually deployed into battle to find out, and to make sure it was being used correctly. These folks could bring back feedback, or insight on human factors, or ideas for new projects when they returned.

You point out in the paper that OSRD did fund some basic science, but for the most part, they had to take scientific knowledge as given. You quote Conant as saying the time for basic research is before a crisis. Can you talk about how pre-existing basic scientific knowledge hooked into what OSRD was doing and how they fit together?

In the radar case, the basic understanding of the integral pieces that go into a functional radar system were known. Generating, transmitting, reflecting, receiving radio waves – that’s fundamentally what a radar system does. So that part of the problem was understood. What was missing was the ability to generate microwaves at a high enough frequency to be able to track fast moving objects and very distant objects, as the military needed to do. And as a result, their effort in the radar program was less focused on how does this technology work, and more on how to actually generate high frequency microwaves, and then how to make the technology useful in field settings, such as with features like visualization, or radar sets for land, versus sea, versus air.

The basic science of atomic fission, on the other hand, actually was pretty recently discovered when World War II kicked off. Nuclear fission was first achieved in 1938 by Otto Hahn. But to be able to create and harness a controlled chain reaction required a bit more fundamental research that took place at the beginning of the NDRC effort. But that pretty quickly evolved again into an applied research effort, which was to develop enough fissile material to make a bomb. That’s when the transition from R to D began to take place. Of course that eventually was transferred over to the Army and became more of a production problem than a scientific problem.

Cyclotron at Berkeley Rad Lab.
Cyclotron at Berkeley Rad Lab, a participant in the Manhattan Project. | Department of Energy

In the penicillin case, again, penicillin was first discovered in the 1920s, and its molecular structure was identified in the 1930s. When we're on the eve of U.S. entry into World War II, the challenge was once again development and production. There are two paths to mass production of penicillin: one is the synthetic path, one is the natural path. For the synthetic path, the main challenge is being able to actually synthesize the compound, and that's still ultimately a drug development problem. For natural penicillin, it was about fermenting Penicillium notatum mold at scale. Either way, basic science was taken as given, and the opportunity of treating infectious disease with an antibiotic like penicillin was possible because of the basic science that preceded the war.

So I would wrap all of this up by saying what basic science delivers in the moment of crisis is the ability to state the scientific or engineering problem, and at least have an idea of the form or forms that a solution might take. The R&D problem is then more a matter of developing these ideas into functional technology that can be made and used at scale than it is about understanding exactly how do we specify the problem and where do we even begin trying to solve it.

Science can inform technology, but technology can also inform science. Were you able to identify any areas in which OSRD technological efforts had any kind of positive feedback on basic science beyond the war and the war effort?

Great question. We haven’t gotten all the way there yet – we’re actually working on this question now. One of the things we want to examine empirically is to what extent the war effort changed the direction of research in the life sciences, particularly in medical research, and whether it essentially unlocked new areas for subsequent follow-up work. There’s a quote from Irvin Stewart, who was the secretary of the OSRD and wrote an official history of the organization, where he said that of CMR’s medical research programs, “The shift in emphasis and in direction was enormous. Many subjects of minor importance in peacetime become of controlling importance in war, and some were even born of war.”

I think our hunch is that after the war, we might see an acceleration of basic research in some of these subject areas. For example, antibiotics and the treatment of infectious disease – I think that's going to be a clear example. Other subjects that weren’t a particular focus of research, at least not in the U.S. pre-war, but may receive more attention post-war, could include the treatment of malaria, which was very important during the war itself. Or things like altitude sickness or oxygen deprivation and its effects on the body. Or nutrition. These may all be fields where the OSRD had longer run effects on U.S. biomedical research, including basic research.

In an earlier paper this year, you and Dr. Sampat identified some of the groundwork that OSRD helped to lay for research, innovation, and manufacturing hubs around the country. What would you say are the major legacies of OSRD?

OSRD had both direct legacies and broader influence on innovation and science policy in the post-war era. OSRD funded several examples of what we would now call “big science” labs, like the MIT Rad Lab, the Applied Physics Lab at Johns Hopkins, the Jet Propulsion Lab at Caltech. Some were new – the Rad Lab was created in 1940 – while others already existed but were relatively small in scale until World War II, and the OSRD grew them significantly. Many of these later became federal contract research centers, what we now call FFRDCs. The Rad Lab, although it was wound down at the end of World War II, spun out the MIT Research Lab of Electronics and Lincoln Labs. The MIT Instrumentation Lab, which was another important World War II contractor, became Draper labs, which is still in Cambridge, too.

The receiver lab at the MIT Rad Lab
Receiver lab of the MIT Rad Lab, 1941. | National Archives

The fact that we have significant federal support for research and development today is a legacy of World War II and the success of the war effort. Even the way that research is administered by the federal government is a legacy of the war. The R&D contract, indirect cost recovery, patent clauses that govern disposition of intellectual property rights in federal R&D contracts – these were all OSRD inventions. The short form, the long form patent clauses, which we mentioned in the paper, still exist today, you know, in somewhat modified form. You can even go into the U.S. code and find some variant of them there that looks pretty similar to what they looked like in the 1940s.[1]

What we show in the other paper – which we’re still working on – is that OSRD-funded research changed the direction of U.S. innovation in the aggregate, and set in motion the growth of technology hubs around the country, especially seen in patenting but also reflected in job growth in associated manufacturing industries. How and why is the focus of our ongoing work.

How comparable is the OSRD integrated research model with the present COVID-19 effort today or other recent such efforts like Zika? Do you see more recent analogs for that model?

I think we have only a limited window into how [Operation Warp Speed] is being run...Given what we do know, I think there are parallels but also some differences. And parallels in particular to the OSRD’s drug development efforts, not surprisingly perhaps. Similarities would include the fact that they are both more applied than basic. They both take a portfolio approach to individual scientific problems, in this case vaccine development or therapeutics, engaging multiple firms and multiple drug candidates. There is as much, if not more, emphasis on production and distribution as there is on R&D. There is production capacity being built at risk, and there’s information sharing across the parties involved.

That said, I would also pick on a number of differences. In my view – and I believe in the view of a lot of economists – U.S. COVID vaccine and therapy research was underfunded, given the scale of the threat and its impact on human life and the economy. BARDA could have had five or ten times the budget and funded even more research, including on countermeasures like mask-wearing and social distancing, which spans the physical and social sciences. OSRD had a very broad portfolio.

The other salient difference between World War II and today – and to me, a quite poignant one – was the degree of coordination, especially internationally. International coordination has not been a prominent feature in the development of COVID-19 vaccines. I think it’s a real missed opportunity. I understand that that’s shaped by a number of geopolitical considerations.

We obviously are dealing with several public challenges now. What’s the applicability of an OSRD model for some of these other public challenges?

I think the answer to “Can you apply this model to other problems” will inevitably be “it depends.” It depends on the nature of the problem. For problems where you have a single user that can control implementation, and can support this tightly integrated, end-to-end approach and give directed feedback on everything from the specification of the research problem to the solution and how it fits the field conditions, the OSRD model can be quite effective. In settings where users are diffuse, it is more difficult to successfully implement this model, especially when solutions require some kind of behavioral change on the part of users.

In the context of COVID, the hospitals and clinics are the organizations that will most likely be the ones administering a vaccine, or who are the ones administering treatments. How do you coordinate with thousands of hospitals and understand what are the problems they face in deploying vaccines and therapeutics and adapt the technology to their needs? In this case, an industry association can be a useful conduit. Moreover, vaccine administration is relatively standardized. The logistical hurdles may include distribution and cold storage, but these are solvable operational problems. The government can, for example, fund the purchase of freezers or seek to develop vaccine variants for use in different settings such as large, urban hospitals versus small, rural ones. The pandemic context isn’t perfectly suited to the OSRD model, but arguably it comes closer than most.

For something like climate change, a solution in many ways requires changes in individual behavior to reduce carbon emissions to the level that will make global climate patterns more sustainable. The OSRD model might not work so well in this context because it’s a complex, systemic issue that interacts with human activity at home, at work, and in between.

What are the major lessons from OSRD?

One key lesson is that good science and technology policy in a crisis might look quite different than S&T policy in regular times, including heavier spending. In a crisis, the returns to a quick resolution are so extraordinarily high that they far outweigh the cost of any potential waste, if some research is unsuccessful or if we get multiple solutions, as we have with COVID-19 vaccines. In regular times, however, we might be more concerned with minimizing wasteful duplication.

Another important lesson is the value, especially in a crisis, of having the mechanisms in place to support coordination. One of the features of OSRD I haven’t mentioned yet is that its leadership had personal relationships before the war. They knew each other, they worked well together, and those relationships and the trust that they already came in with were an asset too.

That actually leads me to a broader, final lesson. The really big lesson is that in a crisis, the assets you need the most are the hardest to build in real time, and the episode points to the importance of investing in strategic assets to draw on during crises. Trust and leadership are among them. In this case, Bush’s ability to win Roosevelt’s trust and ultimately the trust and respect of the military was crucial to OSRD’s success. But this also includes a deep well of basic research. It includes a talented scientific workforce. It can also include knowing who the top scientists are for different problems, so you can find them quickly when you need them. And it can also include ensuring supply chains to make sure that we have the inputs we need to get the research done and to get the technology manufactured and deployed. The investments we make in regular times will shape our ability to harness science and technology in future crises.


[1] See Federal Acquisition Regulations 52.227-11 and 52.227-13.