⚛🧬 The importance of government R&D: A Quick Q&A with economist Andrew Fieldhouse

'We find sizable, persistent productivity spillovers from nondefense government R&D'

Readers of Faster, Please! (and those who’ve read my 2023 book, The Conservative Futurist) know that I’m a big-time proponent of increased federal R&D spending as a catalyst for economic growth and technological innovation. For (too) many on the right, especially, the notion of government investment is a difficult swallow. But it shouldn’t be. 

As I recently wrote over at the AEI website: “Unlike private R&D that tends to focus on developing specific products, basic research focuses on fundamental discoveries that may not have immediate commercial value, but create a foundation for wide-ranging innovations such as the internet and GPS. Public R&D funding also helps address a market failure where private companies underinvest in research due to inability to capture all the benefits.” 

In that piece, I also cite two important papers worth pondering. One, “Science and Innovation: The Under-Fueled Engine of Prosperity,” finds that “a sustained doubling of all forms of R&D expenditure could raise U.S. productivity and real per capita income growth rates by an additional 0.5 percentage points per year over a long time horizon.” The other paper, “Public R&D Spillovers and Productivity Growth,” finds that “the decline in public R&D in the US can explain a third of the deceleration in [total factor productivity] from the 1950s to present days.”

(By the way: Total factor productivity, or TFP, is a measure of economic efficiency that represents the portion of output growth not explained by increases in inputs such as labor and capital. It’s often used as a shorthand for technological progress and innovation in an economy.)

Two great papers, and now here’s a third: I recently came across a paper coauthored by economist Andrew Fieldhouse, “The Returns to Government R&D: Evidence from U.S. Appropriations Shocks,” that explores why federal R&D can be a great investment. The researchers found that nondefense government R&D investment significantly boosts productivity over time. Since the Second World War, for example, the United States has seen returns as high as 150 to 300 percent. From the abstract:

Based on a narrative classification of all significant postwar changes in R&D appropriations for five major federal agencies, we find that an increase in nondefense R&D appropriations leads to increases in various measures of innovative activity and higher business-sector productivity in the long run. We structurally estimate the production function elasticity of nondefense government R&D capital using the SP-IV methodology of Lewis and Mertens (2023) and obtain implied returns of 150 to 300 percent over the postwar period. The estimates indicate that government-funded R&D accounts for one quarter of business-sector TFP growth since WWII, and imply substantial underfunding of nondefense R&D.

To get a better sense of these findings, I asked Fieldhouse a few quick questions.

Fieldhouse is a visiting assistant professor in the Adam C. Sinn ’00 Department of Finance at Texas A&M University, where his research centers on the returns to public R&D spending, US business cycles and growth, and US housing credit policy and secondary mortgage markets. He formerly served as a federal budget policy analyst with the Economic Policy Institute and The Century Foundation.

(Fieldhouse’s responses reflect his own views independent of his coauthor, Karel Mertens, the Federal Reserve Bank of Dallas, and the Federal Reserve System.)

1) What is the impact of R&D spending on total factor productivity?

It depends on what kind of R&D spending we’re talking about! In our recent paper, we find that nondefense government R&D spending has a highly significant, positive effect on total factor productivity (TFP) starting about eight years after funds are appropriated; and we estimate that the aggregate returns to nondefense government R&D are roughly 150–300 percent, depending on the regression specification and various robustness checks.

Conversely, we find no evidence that federal defense R&D spending has had a statistically or economically significant, persistent effect on TFP since WWII, at least within the 15-year horizon we study. (Why the difference? See question three, below.) Firm- or industry-level studies generally find that private returns to firms’ R&D investments are around 20–30 percent, although these studies generally don’t capture broader spillover effects beyond those firms or their industries (unlike our aggregate results).

Jones and Summers (2022) calculate that the “social return” to total US R&D spending, including aggregate spillovers, is roughly 67 percent. Our results broadly square with their finding: We estimate higher returns for nondefense government R&D and our results suggest the returns to defense R&D are likely lower, which would imply an average social return to total R&D spending of 67 percent if the social returns to private R&D were roughly 30–40 percent, which seems plausible. [Note: Government R&D spending refers to work funded by the government, but not necessarily performed by government agencies — much is performed by firms, universities, and nonprofits.]

2) To what do you attribute the fall in US federal R&D expenditures following the 1960s?

The rapid growth in government R&D spending during the 1950s and early 1960s — largely for the Department of Defense and NASA — was driven by national security concerns and geopolitical competition with the Soviet Union. Relative to GDP, federal R&D spending peaked in 1964, after the US nuclear triad was deployed and astronauts seemed likely to beat cosmonauts to the moon. Congress slashed NASA’s budget even before the moon landing, and concerns about inflation restrained federal R&D funding throughout the 1970s. Research funding for the National Institutes of Health (NIH) and National Science Foundation (NSF) has gradually risen relative to GDP, but these agencies never received nearly the levels of R&D funding that were devoted to winning the Cold War. Over time, growth in mandatory spending and debt-financed tax cuts have exerted downward pressure on the overall nondefense discretionary budget — which currently funds most federal R&D — further constraining R&D spending. And when we pivot to fiscal consolidation, the US budget is a slow ship to turn, but it’s relatively easy to temporarily fund fewer research grants to universities or delay planned R&D projects — especially since the benefits of R&D spending are often diffuse and realized gradually.

The Obama administration envisioned a large increase in NIH funding from the Recovery Act as a down payment on sustained investments, but the Budget Control Act of 2011 and ensuing sequestration spending cuts instead reduced NIH funding and other federal R&D spending. We’ve seen unusually fast growth in federal R&D spending during the Biden administration, but federal R&D spending is currently just 0.8 percent of GDP despite that recent uptick, down from 1.8 percent in 1964.

3) What differentiates the impact of defense vs. nondefense R&D shocks? What specific characteristics of nondefense R&D make it more effective in boosting productivity compared to defense R&D?

We find sizable, persistent productivity spillovers from nondefense government R&D, but no such effects for federal defense R&D; the latter results are instead imprecisely estimated and broadly inconclusive. We think there are three likely explanations for this divergence in results. First, defense R&D tends to be classified, deliberately impeding knowledge spillovers, at least over the 15-year horizons we study: A slower diffusion of knowledge from (initially) classified R&D could yield productivity spillovers at longer horizons than we can feasibly study with post-war data.

Second, defense R&D is heavily weighted towards weapons development, whereas nondefense R&D funds much more basic and applied research, which is understood to have bigger productivity spillovers than development work; over our sample, an average dollar of defense R&D funds 86 cents of development, 11 cents of applied research, and just three cents of basic research, whereas an average dollar of nondefense R&D is split about evenly across basic research, applied research, and development. Lastly, our methodology cannot be extended back through WWII — policy variation that might have improved the precision of our defense R&D estimates. Similar to our findings, the Congressional Budget Office (CBO) currently assumes that there are no private-sector spillovers from defense development work, only spillovers from the small portion of defense R&D going to basic and applied research.

4) Do we need an external threat, such as international conflict, to spur public investment in R&D? How could we promote the need for R&D before tragedy strikes?

Geopolitical crises have fueled most of the large increases in federal R&D investments to date. We’ve seen three sustained increases in defense R&D since WWII: one driven by the Korean War, Sputnik crisis, and ICBM race; a second by the Soviet invasion of Afghanistan and President Reagan’s military buildup; and a third by 9/11, the Global War on Terror, and the Iraq War. The only sustained increase in federal nondefense R&D was spurred by the Sputnik crisis and Moon Mission. But there have been (less reactionary) exceptions to this trend, such as Congress doubling the NIH’s budget around the turn of the century. Policymakers might more consistently take a more forward-looking view of federal R&D spending if CBO started dynamic scoring said economic and budgetary effects, in keeping with their analysis of public infrastructure investments. Hopefully more academic research on the longer-term benefits of government R&D spending helps as well!

5) How might your findings on the complementarity between public and private R&D investments inform today's public R&D funding conversations?

We find no evidence that government R&D discourages private-sector R&D spending in the aggregate; to the contrary, we document a modest complementarity between government nondefense R&D and private-sector R&D and an even stronger degree of complementarity for federal defense R&D. A dollar increase in defense R&D spending “crowds in” roughly 52 cents of privately funded R&D, versus 19 cents from a dollar increase in nondefense government R&D; relatively more defense R&D work is performed by private firms, which could explain this greater degree of complementarity, whereas relatively more nondefense R&D is performed by universities and federal agencies themselves. These related increases in private-sector R&D investments should be accounted for in TFP, if measured correctly, so our estimated effects of government R&D spending on TFP should not be biased by spillovers to private R&D spending. That said, crowding in privately funded R&D contributes to economic growth and innovation. Zooming out, economists often think of tax credits and subsidies as key policy levers for encouraging private-sector R&D investment, but our work suggests that appropriations for federally funded R&D are another such policy lever for stimulating private R&D spending. 

---

---

The Conservative Futurist: How To Create the Sci-Fi World We Were Promised

---

---

Micro Reads

▶ Business/ Economics

• Is A.I. Unstoppable? Some Hollywood Craftspeople Fear the Answer Is Yes. - NYT 

• Snow Belt to Sun Belt Migration: End of an Era? - San Francisco Fed

• Donald Trump’s promise of a golden age for oil is fanciful - Economist 

▶ Policy/Politics

• The global chip war could turn into a cloud war - FT 

• Elon Musk, Reid Hoffman and Other Tech Billionaires Brawl Over Politics - NYT

• Google Sees AI as The Key to a Health-Care Revolution - Bberg 

▶ AI/Digital

• AGI: ARC Prize Offers $1 Million to Inspire AI Development - IEEE Spectrum

• From sci-fi to state law: California’s plan to prevent AI catastrophe - Ars Technica 

• The Heterogeneous Productivity Effects of AI - SSRN

• AI Companions Reduce Loneliness - Harvard Business School

• Rodney Brooks’ Three Laws of Artificial Intelligence - Rodney Brooks  

▶ Biotech/Health

• Will implants that meld minds with machines enhance human abilities? - NS

• Breakthrough in Anti-Aging: IL-11 Deactivation Boosts Lifespan by 25% - Scitech Daily

• The movement desperately trying to get people to have more babies - Vox

▶ Clean Energy/Climate

• Bill Gates-Backed Fusion Power Startup Type One Raises $53.5 Million - Bberg

• The New Gods of Weather Can Make Rain on Demand—or So They Want You to Believe - WIRED

• Deep Sea Mining Could Bring a Greener Future But at a Cost to the Ocean - Bberg

• First-Ever Grid-Scale Wave Energy Device Funded by the US Department of Energy Installed by the Navy - The Debrief

• Microwave technique recovers 87% of batteries' lithium in 15 minutes - New Atlas

• The Energy Permitting Reform Act of 2024: What’s in the Bill - Bipartisan Policy Center 

• Kairos Power Begins Construction on Hermes Low-Power Demonstration Reactor - Kairos Power

▶ Robotics/AVs

• Rodney Brooks’ Three Laws of Robotics - Rodney Brooks 

▶ Space/Transportation

• AI put in charge of setting variable speed limits on US freeway - NS

• Chinese E.V. Makers Are Upending Thailand’s Auto Market - NYT 

▶ Up Wing/Down Wing

• Why did the U.S. miss the battery revolution? - Noahpinion 

▶ Substacks/Newsletters

• A California bill could seriously limit open-weight AI models - Understanding AI

• California's Other Big AI Bill - Hyperdimensional

• Chatbots and Vibes - AI Supremacy

• Towards a Sane Compromise on AI Part 4 - From the New World

• LLMs breach a threshold - Strange Loop Canon