Estimating the Long-Term Effects of Federal R&D Spending: CBO’s Current Approach and Research Needs

Posted by Sheila Campbell on
June 21, 2018

The federal government spent $116 billion on research and development (R&D) in 2017, an amount equal to about 0.6 percent of gross domestic product. That money funds basic scientific research, research that applies scientific understanding to achieve specific practical objectives even if those objectives might not have any predictable commercial appeal, and R&D that serves a governmental mission—such as securing national defense or maintaining public health. In addition to its intended purposes, CBO projects that federal spending for R&D increases macroeconomic output through its effect on private-sector productivity.

The Congress requires the Congressional Budget Office, to the greatest extent practicable, to assess the macroeconomic effects of major legislation that has been approved by Congressional authorizing committees, and to incorporate those effects into the agency’s 10-year cost estimates. In a recent report, CBO analyzed the macroeconomic and budgetary effects of changes in federal investment. This blog post expands on that discussion by explaining in detail how CBO assesses the specific effects of changes in federal subsidies for R&D.

This post also highlights areas in which additional research would enhance the agency’s capacity to evaluate the macroeconomic effects of R&D spending. Economics research has not resolved several key questions about the effects of such federal spending on macroeconomic output, lending uncertainty to CBO’s analysis. Studies that examine the relationship between R&D spending and economic growth have yielded wide-ranging results. Economics research also has yet to settle questions of complementarity—how federal R&D spending affects private R&D spending—and timing—how long it takes for changes in spending to have an effect on economic growth. (A bibliography detailing the studies and related scholarly literature that CBO has consulted for its analysis is available on CBO’s website.) Given the limitations of current research, rigorous studies that answer those key questions would reduce uncertainty in CBO’s analytic capabilities in this area and could change the agency’s conclusions.

The Stages and Components of Federal R&D Spending

Research and development occurs in three stages: basic research, which seeks to discover scientific principles; applied research, which aims to acquire knowledge for a specific practical objective; and the development of new or improved products and technology. Federal R&D spending supports a variety of work in government laboratories, universities, and the private sector, including basic research in physics and chemistry, health research, and the development of weapon systems. R&D investment builds the stock of knowledge that helps expand the economy over time, and the academic research that it funds is essential to the training of future generations of scientists. Most of the federal government’s nondefense spending for R&D goes to basic and applied research; most of its defense spending is focused on development of weapons.

Of the $116 billion in federal spending for R&D in 2017, a little over 40 percent, $51 billion, was spent for defense purposes. (This amount is lower than amounts reported for previous years because the National Science Foundation, which provides estimates of R&D spending by stage, has recently amended its treatment of development spending to focus more narrowly on experimental development.) Of the remaining $64 billion:

  • Federal spending on health R&D accounted for $34 billion. The National Institutes of Health received the bulk of those funds, directing them to research on cancer, infectious diseases, and other health problems.
  • $20 billion went to R&D on general science, space, and technology, mostly to the National Aeronautics and Space Administration (for such projects as observatories and space missions) and to the National Science Foundation (for research in such areas as physical sciences and engineering).
  • The remainder of the R&D spending went to areas including transportation, agriculture, natural resources and the environment, and energy.

Modeling Steps and Research Needs

The economic effects of R&D spending typically differ depending on the stage of R&D being performed and whether the spending is for nondefense or defense purposes. To estimate the effect of a proposed spending change, CBO takes into account what kinds of R&D spending would be affected, or, if those details are unclear, the agency uses historical averages to develop its estimates.

To determine the long-term effects of changes in federal spending for R&D, CBO considers the following questions:

  • How would changes in federal R&D spending affect R&D spending by the private sector?
  • By how much would those changes in federal R&D spending affect economic growth?
  • How long would it take to realize those changes in economic growth?

In the discussion that follows, CBO discusses its approach to answering those questions and the types of research that the agency draws upon to do so. Some of the challenges the agency faces in addressing those questions are also described, with the hope of stimulating further research by experts in the field.

How Would Changes in Federal Spending Affect the Private Sector’s Spending for R&D?

In evaluating the macroeconomic effects of changes in fiscal policies, CBO first considers the effect that those changes have on spending by state and local governments and the private sector. The private sector has been the primary source of funds for R&D in the United States since 1980. Industry spent $333 billion on R&D, or 67 percent of the national total, in 2015 (the most recent year for which data are available). The federal government provided 24 percent of the funding for R&D in that same year, and universities, colleges, nonprofit organizations, and nonfederal governments accounted for the remaining 8 percent of national R&D funds.

Studies typically find some degree of complementarity between public and private R&D spending, although the estimated amounts vary. Private industry and the federal government generally focus on different stages of R&D. Most industry-funded R&D is for product development. As noted previously, the majority of federally-funded nondefense R&D is for basic or applied research. One reason for the federal government’s large role in nondefense research is that private firms invest less in such research than its social benefits justify, both because the benefits of such research are difficult for firms to capture and because its possible commercial potential is tough to predict. The federal government plays only a small role in the development stage of nondefense R&D because private firms have strong incentives to invest in projects that are close to commercialization. CBO expects the complementary effect to be larger for federal spending on basic research that may be relevant to future commercial applications, although the agency does not expect that effect would be immediate, especially to the extent that the change in private R&D spending would occur at the applied research or development stages.

In CBO’s judgment, federal spending for R&D has a small but noticeable positive effect on the amount of money the private sector spends on R&D. However, additional research could improve CBO’s understanding of that relationship. Studies that examine one or more dimensions of the question of complementarity would be most instructive: how federal R&D affects private R&D at different stages of research or across stages of research; how those effects differ depending on the stage of research; and how long it takes for any changes in federal R&D spending to affect private R&D spending.

How Much Would Changes in R&D Spending Affect Macroeconomic Growth?

CBO expects that R&D (as well as federal investment more generally) increases aggregate economic output mainly by gradually boosting private-sector productivity in the longer term. (Federally-funded R&D has other effects as well: It increases output in the short term by increasing overall demand for goods and services in the economy, and as part of its intended purposes, it produces public benefits that are not fully captured in measures of economic growth, such as improvements in life expectancy and public health.) In CBO’s assessment, some federal investments in R&D have stronger effects on productivity than others. CBO estimates that federal nondefense R&D yields about three-fourths of the effect on output that private investment yields (from all types of investment, including R&D). CBO bases that productivity estimate primarily on researchers’ estimates of the productivity of investment in public physical capital because there are few estimates of the direct effect on output of federal investment in R&D.

CBO treats federal R&D spending for defense differently. While such research sometimes provides civilian benefits—such as with the development of global positioning satellites that were initially developed for military purposes—the majority of defense R&D does not provide spillover benefits to the private sector. As a result, such spending does not influence CBO’s analysis of macroeconomic outcomes, except for the small portion of defense investment that goes to basic and applied research, which CBO considers to have similar effects to nondefense research.

Some studies have suggested that an increase in federal R&D spending could diminish research productivity in both the government and the private sector, particularly for defense research. That would occur if greater federal R&D spending increased the demand for scientists and led to higher salaries for scientists across both public and private sectors. A tight labor market for scientists would make that wage effect more likely, and could in turn have a negative effect on economic growth or alter the amount of spending on R&D by the private sector.

Although extensive data exist on federal spending for nondefense R&D, economists have not come to a consensus about the economic (or social) returns on that spending. The knowledge or advancement of science that is produced specifically by federal spending is difficult to account for; the timeframes involved are often long; the efforts needed to achieve innovation may be increasing over time; and past performance may not offer a good prediction of future returns. A convincing synthesis of the results from the literature has proved elusive. In their attempts to reach a conclusion, researchers have pursued several different tracks: estimating the aggregate effects of federal R&D investment, studying individual industries or groups of industries, and tracking publications and patents.

  • Aggregate Studies. Researchers have examined the links between federally-funded R&D and private sector outcomes across disciplines and industries. Many studies have found that federally-funded R&D has positive economic effects, but estimates of those effects differ greatly. (The overall macroeconomic effect also depends on whether the federal spending is financed by reducing other federal spending or by borrowing additional funds, a consideration that also applies to forms of federal investment, not just R&D.) The literature on the economic returns to R&D frequently cites Edwin Mansfield’s 1991 finding of a 28 percent rate of return on academic research (which includes benefits to society that are not captured in economic activity). CBO has found that estimate to be both imprecise and not very useful in making budgetary decisions, both because it is an estimate of average returns rather than the marginal return on additional spending on R&D and because it focuses on academic research, which may also be funded by industry or the university.
  • Industry Studies. The effects of R&D spending on particular industries—agriculture, pharmaceuticals, and manufacturing—have received substantial attention from researchers. However, those industry studies typically focus on a specific portion of the federal R&D portfolio, and the results are not necessarily generalizable across industries. Studies dating back to the 1950s have estimated rates of return on various segments of basic agricultural research to be between 20 and 50 percent. Studies focusing on the impact of federally-funded biomedical research have typically found that it has contributed to new drugs and other advances in treatments for illnesses and produced high rates of return. However, critics note several reasons that their estimates may be unreliable: they often focus only on successful R&D programs; they assume the returns could not have occurred without the public R&D; and they ignore other investments beyond R&D that might have contributed to the innovation. Analysis of the manufacturing sector has suggested that the returns on federal R&D spending through that channel have been lower than the returns on private spending for physical capital such as facilities and equipment. However, that analysis also suffers from drawbacks, including a lack of information about the private R&D capital stock—the value of the tools and knowledge already accumulated by the private sector.
  • Bibliometric Approach. An additional set of research takes a bibliometric approach, tracing a line from federal spending to the publication of academic papers. That approach has the advantage of establishing a clear link between funding and research output, but it does not measure the economic value of an academic publication. Related studies connect publications based on federally-funded research to patents. Citation of a paper in a patent record establishes a clear link to the federally-funded research but does not identify the contribution of those cited papers relative to other inputs into the invention. In addition, patents are not used uniformly across industries and so the contributions of some fields of science are likely to be noted more completely than others.

After reviewing all three strands of the literature, CBO has identified a need for research to examine differences in the results from various aggregate studies and the limited applicability of industry and bibliometric studies. Studies that clearly establish the relationship between all federal spending for research and aggregate macroeconomic outcomes would be helpful, as would studies demonstrating the relevance of those outcomes for changes in future spending. CBO would also benefit from more studies that estimate the relationships between federal spending on a particular stage or specific vein of research and the macroeconomic outcomes. Such research would be very useful for isolating the macroeconomic effects of federal R&D spending.

How Long Would It Take to Realize Changes in Macroeconomic Growth?

The length of time over which federal R&D spending affects aggregate output in the economy depends on the stage of research being funded. Industry studies of pharmaceuticals have found that new drugs entered the market about 20 years after public investments in biomedical research and about 12 years after targeted, disease-specific research. In the absence of other evidence, CBO estimates that the macroeconomic effects of spending on basic R&D begin only after 20 years, and that it will take another 20 years to realize the full effect. For applied R&D, the effects are considered to begin sooner—starting after 10 years—but still taking another 20 years to realize their full macroeconomic impact. Development expenditures are expected to begin to have an impact about a year after they made and, like basic and applied research, are also anticipated to follow a 20-year path to their full effect.

Because of the narrowness of the supporting literature, CBO has identified a research need for more information on the average amount of time that will elapse before different types of R&D spending translate into viable commercial products that affect economic output. Estimates for the different stages of R&D for purposes other than pharmaceutical development would be particularly useful. Such research might lead CBO to modify its expectation of how long it takes such spending to affect aggregate economic output.

Sheila Campbell is a Principal Analyst in CBO’s Microeconomic Studies Division and Chad Shirley is CBO’s Deputy Assistant Director for Microeconomic Studies.