Does Federal Research and Development Funding Promote Innovation and Productivity?
Fieldhouse and Mertens (2026) examine whether government-funded research and development (R&D) increases innovation and economic productivity. They analyze quarterly U.S. data from 1948 through 2021 and identify major changes in federal R&D appropriations that were largely unrelated to short-term economic conditions. They find that increases in nondefense R&D funding produce substantial long-term benefits. A funding shock that eventually raises government R&D capital by 1% increases business-sector productivity by about 0.2% and labor productivity by roughly 0.25% after 15 years. They estimate returns of 140%–210% and find that nondefense government R&D accounts for at least one-fifth of postwar productivity growth.
Why This Article Was Selected for The Policy Scientist
Government R&D is an important policy issue because technological innovation is a major source of long-run productivity growth, higher living standards, and economic competitiveness. The question is especially timely amid renewed concern about weak U.S. productivity growth and the appropriate federal role in supporting scientific research. This study makes an important contribution by estimating the aggregate returns to public R&D rather than examining individual programs or industries. Its unusually long postwar dataset is a major strength and supports broader inference, although generalizability beyond the United States remains uncertain. The narrative identification strategy and long-horizon causal estimates are methodologically strong and substantially more persuasive than conventional multivariate regression.
Full Citation and Link to Article
Fieldhouse, A. J., & Mertens, K. (forthcoming). The returns to government R&D: Evidence from U.S. appropriations shocks. American Economic Review.
Central Research Question
The central question is whether government funding for research and development produces causal, long-run gains in innovation and economic productivity. The analysis addresses a fundamental identification problem: government R&D spending may rise or fall in response to economic conditions that independently influence productivity. A simple correlation between public R&D and subsequent economic performance therefore cannot establish that government research caused the observed gains. The study seeks to isolate changes in federal R&D appropriations that are plausibly independent of short-run economic conditions and trace their consequences over periods as long as 15 years.
A second question is whether the economic effects differ between defense and nondefense R&D. This distinction matters because knowledge generated through civilian research can diffuse broadly through universities, firms, patents, publications, and trained scientists, whereas military research may involve classified technologies or more limited dissemination. The study consequently estimates separate effects for defense and nondefense appropriations and examines several mechanisms through which government research might ultimately affect private-sector productivity.
Previous Literature
The study contributes to a large literature on the “social” returns to R&D—that is, economic returns that include benefits extending beyond the organization conducting the research. Hall, Mairesse, and Mohnen (2010) review extensive evidence showing that R&D generates substantial spillovers. Jones and Summers (2022) similarly use aggregate postwar evidence to estimate the social returns to research investment. Bloom, Schankerman, and Van Reenen (2013), using changes in federal and state R&D tax incentives, estimate a 55 percent gross social return to private R&D.
A related literature examines particular government programs. Azoulay et al. (2019) find that NIH funding causes increases in private-sector patenting. Myers and Lanahan (2022) document substantial private R&D spillovers from the Department of Energy’s Small Business Innovation Research program. Gross and Sampat (2023) show that wartime research programs contributed to the subsequent development of technology clusters, while Kantor and Whalley (2025) identify persistent local manufacturing effects associated with NASA research during the moon program. Babina et al. (2023) find that reductions in federal university research grants decrease high-technology entrepreneurship and scientific publications.
The study also builds on research concerning public investment and productivity, including Aschauer (1989) and Munnell (1990), and on methodological work concerning fiscal shocks. Ramey (2011), for example, emphasizes the importance of correctly identifying when information about future government spending becomes available. The present study extends these literatures by attempting to estimate the aggregate causal effects of government R&D across programs rather than concentrating on a particular agency, industry, or geographic area.
Data
The analysis assembles an unusually long historical dataset covering the postwar United States. The principal quarterly estimation sample extends from 1948 through 2021. The researchers construct a new record of congressional appropriations supporting R&D at five major federal agencies: the Department of Defense, Department of Energy and its predecessors, NASA, National Institutes of Health, and National Science Foundation. Together, these agencies typically represent roughly 90 percent of federal R&D spending.
The appropriations series is constructed from the Budget of the U.S. Government and related primary sources. The researchers identify 257 significant changes in real R&D appropriations between 1947 and 2019. They then examine budget documents, congressional committee reports and hearings, presidential statements, CQ Almanac accounts, and contemporary newspaper coverage to determine the motivations underlying these changes.
The analysis combines these appropriations data with measures of government R&D capital, business-sector total factor productivity (TFP), GDP, labor productivity, public infrastructure, private R&D, and other economic variables. Additional measures capture innovative activity, including economically valuable patents, STEM doctoral degrees, the number of researchers engaged in R&D, and technology publications.
Methods
The empirical strategy is designed explicitly for causal inference. The authors first use narrative historical analysis to distinguish changes in R&D appropriations associated with short-term economic conditions from changes driven by other considerations. Of the 257 significant appropriations changes examined, 46 are classified as endogenous to short-run macroeconomic conditions. Baseline estimates exclude these changes.
The researchers then use Jordà (2005) local projections to estimate how economic outcomes evolve following an unexpected R&D appropriations shock. The models include lagged measures of productivity, government and private R&D capital, capacity utilization, stock-market indicators, defense-spending news, and other variables intended to remove predictable economic influences. Because research may take years to produce commercially useful knowledge, responses are estimated over horizons extending to 15 years.
Finally, the study uses the SP-IV methodology developed by Lewis and Mertens (2024) to estimate the production-function elasticity of government R&D capital. This structural approach relates the dynamic response of government R&D capital to the corresponding response of productivity. The researchers conduct numerous robustness tests using alternative controls, specifications, estimators, shock definitions, and statistical inference procedures.
Findings/Size Effects
The principal finding is that increases in nondefense government R&D produce substantial but delayed gains in private-sector productivity. A funding shock that ultimately raises government R&D capital by 1 percent increases business-sector TFP by approximately 0.2 percent. The effect emerges gradually and becomes statistically significant primarily at horizons of roughly 8 to 15 years. Labor productivity rises by approximately 0.25 percent after 15 years, while potential output eventually increases by approximately 0.2 percent.
The innovation measures provide evidence about the mechanisms preceding these productivity gains. The patent innovation index, which weights patents according to their estimated economic value, rises by as much as 2 percent following a positive nondefense R&D shock. The number of new STEM Ph.D. recipients eventually rises by more than 1 percent, while the number of researchers engaged in R&D increases by as much as 0.5 percent during intermediate horizons. Technology publications also increase.
Nondefense appropriations disproportionately increase basic research and generate persistent increases in research performed at universities and government agencies. Public and private R&D also appear to function more as complements than substitutes: higher public R&D does not simply displace privately financed research.
The structural estimates imply a production-function elasticity of total government R&D capital of approximately 0.11, corresponding to an elasticity of approximately 0.06 for nondefense R&D capital. Growth accounting indicates that nondefense government R&D accounts for at least one-fifth of postwar business-sector TFP growth. Slower growth in public capital collectively explains approximately 0.36 to 0.43 percentage points of the roughly one-percentage-point productivity slowdown following the 1960s.
Most strikingly, the estimated average return to nondefense government R&D ranges from approximately 140 to 210 percent, depending on specification and calculation method. A central estimate is approximately 171 percent. These estimated returns substantially exceed conventional estimates for public infrastructure and private R&D.
The results for defense R&D are markedly different. The estimates provide little evidence that increases in defense R&D produce persistent gains in TFP within the 15-year horizon. The authors caution that this does not establish that defense research has no productivity spillovers. Classified research, slower diffusion, different research composition, or effects occurring beyond the available horizon could make those benefits substantially harder to detect.
Conclusion
The study concludes that nondefense government R&D has historically generated large and persistent productivity spillovers throughout the U.S. economy. Its effects are not immediate. Instead, additional appropriations gradually expand the stock of publicly financed knowledge, increase measurable innovative activity, enlarge portions of the scientific workforce, and eventually raise private-sector productivity and potential output.
The distinction between defense and nondefense research is central to interpreting these results. The strong productivity effects are concentrated in nondefense R&D, which places comparatively greater emphasis on basic research and permits wider dissemination of scientific knowledge. Defense research produces substantially less conclusive long-run productivity effects within the period the researchers can reliably examine.
The findings also place the postwar decline in federal R&D investment within the broader history of slowing U.S. productivity growth. Government R&D appears to have made a quantitatively important contribution to productivity growth since World War II, with estimated social returns considerably exceeding those associated with many other forms of investment. At the same time, the estimated returns represent historical averages across a long period encompassing major changes in scientific institutions, federal research priorities, and the structure of the U.S. economy. The magnitude of future returns therefore need not equal the historical estimates. The study ultimately provides evidence that government-funded knowledge capital has been an important component of long-run U.S. productivity growth while identifying the composition of public research as an important subject for further causal investigation.
Comments