3
IMPACTS ON THE U.S. ECONOMY AND QUALITY OF LIFE
What is known about the contribution of research to GDP, productivity, wages, employment, and private sector R and D? Is there a basis for setting a target for aggregate research spending? How can the flow of knowledge from research into particular economic activities be measured? These were some of the questions addressed during the session of the workshop on the direct economic benefits of research spending. Three speakers looked at such issues as R and D’s influence on productivity gains, the association between research activity and local labor market conditions, and citations in industrial patents to publicly funded research. These have been the principal avenues for measuring economic benefits of research
FEDERAL RESEARCH AND PRODUCTIVITY
From the 1950s to the 1970s, many studies examined the broad outcomes of federal R and D, but fewer studies have occurred in recent decades, said Carol Corrado, Senior Advisor and Research Director in Economics at the Conference Board. She presented recent results from investigations of the relationship between R and D and productivity, taking mostly a “30,000-foot perspective.” She also emphasized a key prospective change in the U.S. national accounts. Starting in 2013, R and D spending will be capitalized as an investment instead of being treated, as it is now and has been historically, as an intermediate expense. This means that both private and public R and D will raise bottom-line GDP and national saving.
According to Corrado, the total U.S. R and D investment level has been stable since the 1980s as a share of GDP. Since 1959, the share of
all R and D investment funded by the public sector has declined relative to that funded by the private sector, with rough stability in both sectors since about 2001. The total nominal R and D investment in 2007 was $407.5 billion, with business at $269.6 billion, government at $117 billion, universities at $10.6 billion, and nonprofits at $8.4 billion.
Corrado investigated the R and D intensity of eight industries over two time periods: the 1990s and the 2000s. When the R and D intensity of each industry matched Total Factor Productivity (TFP) estimates, as it did for the 1990s, R and D can be interpreted as the sole driver of productivity gains. The 1990s data also show that the computer industry, which was heavily subsidized by federal R and D, outperformed the others. In fact this industry seemed so exceptional that Corrado removed it to look solely at the other seven industries for more general trends. But even excepting computers, R and D appeared to be the sole driver of the productivity gains of the 1990s.
However, the same comparison showed that R and D contributed only 30 percent to the average industry productivity gain in the 2000s, Corrado said. This analysis had too little data to draw firm conclusions, according to Corrado. The analysis also was not able to measure the impacts of investments in the life sciences on human health, though the Bureau of Economic Analyses (BEA) is working to introduce a healthcare satellite account. Also excluded from this analysis was educational services, which may require a geographically localized approach.
The productivity growth of the 1990s suggests that the Internet and demand for networked devices were key drivers of economic activity in that decade, said Corrado. Government played “a classic role” in supporting new technology when several private companies worked with NSF to set up the first T1 telephone data line in 1987. This federal R and D created infrastructure and also helped to close “valleys of death” in the commercialization of research.
Corrado also called attention to the dwindling share of manufacturing in the U.S. economy. What does it mean for policy if the United States moves to an economy characterized by “designed in California, made in China”? she asked.
Finally, she observed that innovation is “more than science.” Studies suggest that firms innovate based on intangibles such as product design, new business processes, and staff knowledge building, not just new research results. An estimate for 2001 put R and D’s share of
spending on all of these intangibles at just 16 percent, although R and D dollars could influence the outcome of spending on other intangibles.
Corrado said that the source of innovations needs to be better understood. For example, Virgin Atlantic holds a patent on the design of its first class cabins, which is one example of how the notion of a science and innovation policy can be broadened. The role of diffusion, which could help explain the changes from the 1990s to 2000s in the industries she analyzed, also needs more intensive study.
INDIRECT ECONOMIC BENEFITS OF RESEARCH
Government research expenditures are increasingly justified in terms of economic benefits such as job creation. But the practical benefits of research are disputed even by some scientists, said Bruce Weinberg, Professor of Economics and Public Administration at Ohio State University, and there is little accepted methodology for estimating these benefits.
Weinberg focused on “indirect benefits.” He described these as the “productivity spillover benefits” beyond particular products or processes that develop out of research. Examples include a better trained workforce that generates higher productivity, solutions to industrial problems, new infrastructure, or hubs for innovation. Even if these spillover benefits turn out to be smaller than the direct benefits, “they are important and are increasingly driving the discussion about the cost and benefits of research.”
One way to estimate the economic benefits of research is through job creation, but Weinberg noted that “this poses deep fundamental and practical problems.” For example, if a job pays $50,000 a year, the value of the job to a person is really that amount minus what a jobholder would have been earning on another job. Also, as wages go up in science jobs, people may move to science from other occupations, which moves jobs from one sector to another rather than creating jobs.
Instead, Weinberg suggested focusing on outcomes—wages or productivity—in places where more science and research is carried out. What should be estimated, he said, is whether research leads to more productive industries in local economies.
Weinberg related measurements of research in particular cities to economic metrics of those cities. He asked whether wages and employment are better in cities where more research is being done. He
also looked at measures of innovation such as patenting in cities with more science.
Based on preliminary results for U.S. metropolitan areas, a positive correlation exists between wages, employment, and academic R and D, he said. The results indicate that a 1 percent increase in academic R and D is associated with roughly 120,000 more people employed and $3 billion more earnings in a metropolitan area. Weinberg cautioned, however, that these results are far from definitive because of confounding factors. For example, science-intensive cities may be different from other cities, or workers may have different abilities across cities. “The literature hasn’t really addressed the underlying challenges convincingly,” he said.
“If I were to summarize the literature, I would say there is some evidence that science or research impacts wages, industrial composition, and patenting, but these estimates are weak,” Weinberg concluded. For the future, it is important to think about productivity spillovers not simply in terms of job creation but by doing studies that “unpack the mechanisms by which science and research impact economic outcomes.”
BEYOND CITATIONS AND PATENT REFERENCE COUNTS
A common way to measure knowledge flows among universities, government laboratories, and firms is through citations in patents to patent references (PR) assigned to universities, federal laboratories, or research institutes and citations to non-patent references (NPR) with an author affiliated with a university, federal laboratory, or research institute. Such references provide “rich data that can be used across industries and firms and over time,” said Michael Roach, Assistant Professor of Strategy and Entrepreneurship at the Kenan-Flagler Business School at the University of North Carolina.
However, patent citations also suffer from some limitations, Roach acknowledged. Not all inventions are patented or even patentable, so such studies are limited in what they can observe. Similarly, not all knowledge flows are citable or cited. Firms may not want to disclose important developments, or industrial authors may overuse citations, which is a trend Roach has found in his research. As a result, citations likely mismeasure knowledge flows, either randomly or with a systematic bias.
In particular, NPR citations capture knowledge flows through channels of open science (such as publications), direct use of technological opportunities in new R and D projects, and knowledge flows to firms’ applied research. NPR citations do not but should capture knowledge flows through contract-based relationships, intermediate use in existing projects, and knowledge flows to firms’ basic research activities. All things considered, Roach concluded that citations likely understate the impact of public research on firms’ performance.
Roach described a study done with Wesley Cohen (Roach and Cohen, 2011) that used the Carnegie Mellon R and D Survey of manufacturing firms to measure a firm’s use of public research. The “key takeaway,” according to Roach, was his calculation showing that the unobserved contribution of public research to innovative performance is comparable to what is observed. They estimate that observed knowledge flows account for about 17 percent of firms’ innovative performance while unobserved flows account for about 16 percent.
Future research should concentrate on NPRs, Roach said. Though such data are costly to obtain, they are one of the best measures available to measure knowledge flows. He suggested that the National Bureau of Economic Research and the U.S. Patent and Trademark Office make NPR data more readily available to scholars.
Other external data could be used to measure knowledge flows, such as NSF’s recently expanded Business R and D and Innovation Survey (BRDIS). Also, the origins of citations need to be better understood. “We need to be looking at the micro level,” Roach said, echoing points made in the previous panel. Research needs to look at inventors, scientists, and firms—“trying to get inside that black box.”
Alfred Spector of Google commented on Corrado’s description of the change in national accounts making R and D a capital investment. Spector noted that firms currently expense research because they do not know what the results of the research will be. Corrado replied that while some business accountants are resisting the change, those who favor it say it can provide a “holistic picture of how and where firms make their investments. … What you set aside today to generate future consumption—in other words, what you forego today—is your
investment.” She explained that national accounts do not have to line up with firms’ accounting practices.
The session moderator, Bronwyn Hall, said that publicly held firms use Financial Accounting Standards Board (FASB) policy for expensing R and D. An advantage is that expensing R and D offsets current income. The problem from an economic analysis perspective, Hall said, is that “in the United States, the value of firms even when the market is down is substantially higher than the value of their tangible capital assets.” When one looks for what explains the difference, “capitalized R and D is the first thing” one sees.
In response to a question about how research funders can generate more positive spillover effects from research, Weinberg pointed out that research funding is more likely to have positive effects in nearby location than distant locations. Improvements in dissemination could enhance information flows, and there are many ways to study the impacts of this dissemination.
