Although Washington cannot be said ever to have had a coherent technology policy, the United States has nevertheless done well compared with most nations in supporting and sustaining innovation. Since World War II the three primary elements of federal technology policy have been managing intellectual property rights, providing support for industrial R&D, and funding research and training in academic science and engineering. According to economic analysts, Washington’s sustained efforts in these three areas have generated a significant share of U.S. economic growth. But recent developments in each policy area—changes in property rights law, reductions in federal direct support for R&D, and a cutback in undergraduate science and engineering—pose worrisome questions for the U.S. research enterprise.
This essay focuses chiefly on trends in federal and private support for R&D, particularly changes in the standard “science indicators”—the share and composition of support for different categories of R&D. In some respects the evidence appears contradictory—and at first heartening. Although federal spending has fallen 18 percent in inflation-adjusted dollars since its peak in 1987, private spending has increased more than enough to offset this decline. But a closer look at the data reveals that changes in the science enterprise, particularly in the property rights regime and in the demand for scientists and engineers, call into question the reliability of these indicators. The raw data turn out to be misleading and provide little grounds for complacency.
What Do the Trends Tell Us?
Consistent with economists’ findings that most economic benefits of R&D, especially fundamental research, do not and cannot accrue to those who perform it, a cornerstone of U.S. science policy has been that public support must be the dominant component in fundamental research and also play some role in industrial R&D. For the past half-century, Washington has thus paid for most university research, conducted fundamental research in national laboratories, and provided financial support for industrial research and development, especially in defense industries.
Both federal and state governments also provide R&D tax breaks, for example allowing businesses to deduct most R&D spending immediately rather than following a depreciation schedule. A federal research and experimentation tax credit also lets firms deduct part of their incremental R&D expenditures from their taxes. In a recent NBER working paper, Bronwyn Hall and John van Reenan find that the R&E tax credit has had a substantial incentive effect, with each dollar of subsidy probably increasing private-sector R&D by about one dollar. The U.S. Office of Management and Budget estimates the “outlay equivalent” cost of the credit at $2.23 billion in 1999—roughly 30 percent higher than it was during 1988-92, but still not enough to make up for the decline in direct federal support of industrial R&D, much less to account for the rise in private R&D.
Between 1989 and 1999 total U.S. real spending on research and development rose more than 36 percent. (All budget comparisons here are corrected for inflation. Dollar amounts are reported in 1992 constant dollars.) R&D spending as a share of gross domestic product grew from 2.6 percent to 2.79 percent—higher than any year since 1967. (See box on data sources on page 13.) As figure 1 [below], private industry’s share of R&D funding rose from about 50 percent in 1987 to more than 68 percent in 1999.
The declining federal share of R&D is directly related to the post-Cold War fall in defense spending. Direct federal R&D support to industry (mostly to defense industries) fell 50 percent between 1987 and 1999. But the Cold War’s end did more than lower defense-related R&D. It also broke the political consensus in support of science R&D in general, thus shrinking the federal role in fundamental research and dramatically redistributing R&D spending by disciplines and industrial applications.
In most fields of basic research, the federal presence is declining. Only in life science has federal support increased significantly. Total spending on basic research (see figure 2 below) rose more than 28 percent between 1991 and 1999, thanks largely to industry and universities. The federal increase was small—only $2.2 billion—and of that, life sciences accounted for $1.7 billion.
Although total federal academic spending has grown 55 percent since 1987, the federal share of academic research has been falling for 30 years and by 1999 was down to 57 percent. The share of federal support for university research going to life sciences grew from 58 percent in 1990 to an estimated 63 percent in 2000. Physical sciences, math and computer sciences, and engineering—traditionally fields whose research support has come mostly from Washington—have declined as a share of the federal academic budget from 27 percent to 24 percent. Although federal support for these activities grew, it did so slowly—less than 2 percent a year on average.
Industrial R&D has undergone similar changes in public and private support. In addition to changes in the public and private share of total R&D, the technological focus of R&D activities of the public and private sectors has diverged. Federal support for industry-performed R&D remains concentrated in defense areas such as missiles and aircraft (44 percent), engineering services (14 percent), and scientific instruments (19 percent). By contrast, company-supported R&D in the late 1990s was directed toward information technology, including computing, electronics, communications equipment, and computing services (more than 40 percent); drugs and medicine (9 percent); transportation manufacturing (14 percent, of which only 4 percent is in aircraft and missiles); and scientific instruments (7 percent). Industry data for basic research spending are incomplete, but industry spending appears to be distributed broadly, with about 20 percent in drugs and medicine and the remainder scattered widely across the technical fields related to physical sciences.
Industrial R&D, which drives innovation in the private economy, is now spread relatively evenly among different fields and still emphasizes applications of the physical sciences and engineering. Washington, by contrast, concentrates its industrial R&D support on missiles and aircraft, in which industry now invests relatively little. And federal spending on fundamental research increasingly focuses on life sciences, which is the science base of neither the defense sector nor most of the industries where private R&D is growing rapidly.
Interpreting the Trends
Although economic analysts have examined extensively the relationship between spending and innovation, their conclusions are not straightforward. Nor, for that matter, is the relationship. R&D is always risky, and even if it is successful, realizing its benefits can require long delays, complementary innovations, and even major changes in economic or social institutions. The National Science Foundation’s authoritative R&D data series measure spending and employment in R&D at different times. For trends in the data to be meaningful indicators of innovation, the relationship between spending and innovation must be consistent over the years so that a dollar of R&D activity leads to roughly the same amount of innovation. But the changes in the R&D enterprise noted above, as well as changes in other R&D institutions, imply different relationships between spending and innovation and thus affect our interpretation of trends in R&D spending.
One seemingly simple issue is measurement. The NSF data on R&D spending in the private sector are based on a survey of firms, and for several reasons the way firms report R&D spending today may not be consistent with earlier responses. Firms may, for example, reclassify as research certain ongoing activities, such as market research and business software development, because increasing R&D spending enables them to claim a larger federal tax credit. With profits growing over the booming 1990s, firms are especially likely to reclassify more activity as R&D to maximize tax advantages.
And the R&D survey now includes many nonmanufacturing firms for which the traditional classifications of basic, applied, and development research have less meaning. Whereas an auto maker’s R&D is reasonably well defined and structurally separated from production, the distinction is much more ambiguous in the service sector. Arguably all spending by a software development firm could be R&D. Hence, data on both the level and composition of R&D effort are probably less reliable as R&D support shifts from the public to the private sector and, within the private sector, from manufacturing to services.
Recent moves to strengthen intellectual property rights law compound the problem. Since 1980, IP rights have been beefed up in several ways, such as by making genetic information and business software patentable and by cutting back “fair use” exceptions to copyright protection. Stronger intellectual property rights are likely to increase private R&D investment and innovation, in part because innovations may be more valuable to the inventor and in part because the acquisition of innovations by other firms may be more crippling.
Stronger IP protection can generate more R&D if it shifts some of the benefits of innovation to the innovator. U.S. patent applications and patent grants are now at record highs. The patent rate per dollar of R&D is higher than at any time since 1977. But the shift in the IP regime makes interpreting these numbers ambiguous for both the short run and the long run. The short-run problem is that R&D may be employed on projects with lower social payoffs on average, thus reducing the productivity of R&D. The long-run problem is that even if stronger IP did not lower R&D productivity, overly broad IP rights could stifle the next wave of commercial innovations. For example, patents granted (as they have been) to discoveries of genes with no known commercial applications are essentially patents on basic scientific knowledge. Though that knowledge might be the basis for further useful research, the rights of the original patent holder discourage future use of the gene.
Another reason why the productivity of R&D investment may be falling relates to the supply of skilled labor. If that supply is inelastic, an increase in private R&D spending would raise the salaries of science and engineering employees, rather than real R&D effort and hence innovative activity. Of course this response would require an external change in the market, since generally employees’ compensation is related to their marginal productivity—but an external event such as the change in the IP regime could, as noted, shift the private marginal productivity of scientists and engineers. Austan Goolsbee, in his article in the May 1999 American Economic Review, finds evidence in an analogous situation. He studied the labor market for researchers in the early 1980s, when the boost in federal support for defense R&D also supplied an external shock to the market for scientists and engineers, and concluded that most of the spending increase simply raised wages.
In fact, the supply of science and engineering labor has increased little for three decades. Colleges and universities are not meeting the demand of students who want to study science and engineering (see box). The most recent statistics show a small increase in the number of research workers, but only because of immigration. Not only is immigration politically controversial at home, but the recent U.S. success in attracting foreign scientists and engineers has led other countries to try to keep technically trained workers at home or to get them to return after completing their education.
Finally, the effect of the increase in industry spending depends on the relationship between private and public R&D. Is a one dollar increase in private funding equivalent to a one dollar fall in public funding? Few analysts think so. Some point out that federal R&D drives up the cost of scientists and engineers, so that reducing it enables industry dollars to buy more innovation, at least in the short run. Cuts in defense R&D and the exodus of scientists and engineers from the national laboratories probably spurred the private R&D expansion in the early 1990s, but these sources of labor are running dry. Others argue that public R&D so complements the private R&D effort that private R&D spending must overcompensate for the decline in federal support.
While economic research has failed to establish a clear relationship between private and public R&D expenditures in general, it does support the historical foundations of U.S. science policy with respect to foundational research: public research enhances private innovation, particularly in technologically related areas. This observation both underscores our concern about the divergence between public and private activities and questions the significance of the increase in private R&D spending.
In sum, the relationship between spending and innovation is sensitive to the composition of R&D as well as to its institutional underpinnings and is fundamentally not comparable across R&D regimes. The apparent increase in R&D does not, over the long run, promise a comparable increase in innovation.
Implications for the Future
The 1990s have witnessed a return of optimism about the American economy. Unemployment is down. Strong economic growth has resumed. And measured private R&D spending is booming, suggesting that the foundations for future growth are in place.
So what is the problem? First, the federal government is a shrinking presence in R&D and is cutting back nearly everywhere except in biology and medicine. Second, the spending figures probably overstate the rise in private R&D spending, but even if private R&D has substituted for the fall in public spending, the private sector is less likely to support R&D that has broad spillover benefits throughout the economy. Third, bottlenecks in the higher education system and uncertainties about the sustainability of immigration may restrain real R&D effort regardless of the spending propensities of both the public and private sectors. The long-run implication is slower technological progress and hence slower growth.
Our policy recommendations flow from these conclusions. First, federal spending on basic research should not be so heavily skewed to the life sciences, but should focus instead on areas of physical science and engineering that are more important to most of the economy, and to information technology in particular. Second, federal and state governments should expand higher education in science and engineering to accommodate would-be undergraduate majors. Finally, because the data on national R&D effort may be misleading about long-term trends and because the relationships among science indicators, the intellectual property regime, and innovation and growth are poorly understood, Washington should fund a major National Science Foundation research effort to get answers to these crucial questions.
Commentary
Is U.S. Science Policy at Risk?: Trends in federal support for R&D
December 1, 2001