Are large firms withdrawing from investing in science? By Ashish Arora, 1 Sharon Belenzon, and Andrea Patacconi 2 Basic research in science and engineering is a fundamental driver of technological and economic progress. Historically, privately funded research has contributed significantly to scientific progress. National Science Foundation data indicate that industry funded about 22% of all basic research in the United States in 2009, compared to about 53% by the U.S. Federal Government [1]. Scientists at companies such as AT&T, Du Pont, General Electric, IBM, and Xerox have produced important breakthroughs, earning several Nobel prizes in the process [2]. But while the scientific contributions may have been significant, they did not always translate into higher profitability. Qualitative accounts indicate that, starting in the 1980s, many corporate labs were closed, downsized, and redirected toward more immediate commercial needs [2 4], although not all observers consider this a cause for alarm [5]. Measuring the extent to which large corporations moved away from basic research and toward commercial applications is challenging. NSF data also show that the share of basic and applied research in corporate R&D in the United States has declined from 28% in 1985 to 21% in 2009 [1]. However, classifying expenditures as either research or development is difficult. Further, these data are not publicly available at the firm level. This makes it difficult to assess whether the aggregate trends reflect changes in the behaviour of existing firms or other factors, such as a change in the industrial mix of reporting firms. We develop publication-based indicators of scientific research and relate it to market based indicators of value [6]. We link scientific publications in hard science (including engineering science) journals from the Web of Science to publicly traded firms in the United States, using the affiliations of the authors. Our primary firm sample consists of 1,014 R&D performing companies over the period 1980-2007. Collectively, these firms account for 312,000 firm publications scientific articles where at least one of the authors is a company employee. In addition, we use firm-level information on variables such as patents, stock market value, book value of capital, and R&D expenditures. Our evidence indicates that, during the period 1980-2007, large American firms have published less over time while patenting more. Figure 1 shows that the share of R&D performing firms that publish at least one scientific article in a given year has fallen from about 17% to about 6% between 1980 and 2007, whereas the share of firms that patent has risen from about 15% to about 25%. Over this time, the ratio of R&D to sales has largely remained stable. A similar pattern emerges when we examine changes within firms. Firms are reducing publication output at about 3% per year, controlling for sales and R&D expenditure, but are maintaining patenting rates over time. As shown in Figure 2, these patterns are not driven by any particular industry, but are present across the board: The share of firms publishing in scientific journals has dropped in every major industry, including pharmaceuticals, chemicals, electronics and machinery. Figure 3 shows that the implied value collectively attached by investors to scientific capabilities declined, while the implied value attached to patents increased. Figure 2 is based 1 To whom correspondence should be addressed. E-mail: ashish.arora@duke.edu 2 A. Arora is at the Fuqua School of Business, Duke University, and NBER. S. Belenzon is at the Fuqua School of Business, Duke University. A. Patacconi is at the Norwich Business School, University of East Anglia.
upon estimates from a regression of the log of the stock market value of a firm upon its stock of publications, where each publication was weighted by the citations it had received. In addition, we included tangible assets (book value), cumulative R&D expenditure (depreciated at 15%), log of the sales revenue (all the foregoing variables are logged), along with controls for year of observation and the industry of the firm. The regression estimates imply that in the period 1980-85, a 10% greater citation-weighted stock of publications was associated with a 1% higher market value, but only 0.5% higher market value by 1991-95. By 1996, this association had largely disappeared. The implication is that the decline in publishing is related to the value attached to research rather than to the difficulty or cost of performing research. A concern is that increases in patent protection may have induced firms to patent more and publish less, out of fear of inadvertently disclosing valuable information. However, if that were the case, firms would especially avoid publishing in applied scientific journals, as applied publications are more likely to contain commercially valuable information. We find that firm publications fell much more in basic science journals than in applied journals, using the CHI classification of journals into applied and basic [6]. Similarly, we find that firm publications fall more steeply in high-impact journals. Thus it is the composition of corporate R&D which is changing (more D and less R ), rather than simply publication practices. Firms may reduce investment in research if new discoveries are not as useful as they once were, so that it may be more profitable to exploit existing knowledge than to push the frontiers of science further. However, as Figure 1 also shows, patents filed by American corporations continue to cite scientific articles at slightly higher rates over time. Moreover, the average vintage of scientific articles cited has not changed [6]. Aggregate data reported by the National Science Foundation show similar trends for the period 1998-2010: American patents cited science and engineering publications at a steady rate between 1998 and 2010 [7]. Thus, insofar as patent citations are a valid proxy measure of the extent to which inventions draw upon science, the evidence suggests that science continues to be valuable to corporation innovation. It is also unlikely that the declining production of publications by American firms merely reflects the growing scientific might of other countries. Intuitively, if it is harder to publish but the commercial value accorded to research is stable, then papers that do make it into print should on average be of higher quality and value. However, as Figure 3 shows, we find a decline in the commercial value attributed to publications. The declining publication output represents a decline in the value of research to corporations, rather than simply greater competition to publish. The evidence presented so far raises two broad questions. First, why are large American firms withdrawing from science? And second, what are the implications for policy? As to the first, there are several possible explanations, which need further investigation. The decline may be related to growing competition in markets, including competition from overseas. As profit margins shrink, investors may demand that corporations abandon their more speculative, long-term projects, to focus on shorter-term projects whose financial performance is easier to measure. In addition, investments in science may most benefit diversified firms, which are most likely to find commercial applications for the unpredictable fruits of science [8]. Insofar as investors favor focused firms, this may also explain why investments in science are declining. A growing division of innovative labor may also be
behind the trends we document. Established firms are increasingly turning to universities and start-ups for inventions to fuel their growth [9]. These firms may invest less in internal research, and rely instead on licensing and acquisitions, to renew their product pipelines. In such a division of innovative labor, universities will have a heavier burden: More research, more translation, and perhaps even more commercial application of research. Though private philanthropy is supporting a larger share of university research, we will likely need more, not less, public funding for research. The straitened budgets of NIH and NSF are thus a cause for concern. But money should also be spent more wisely, paying attention to the unintended, and often perverse, consequences of financial incentives [10-11]. University research may also not be a perfect substitute for corporate research and, at least in some cases, the industrial labs that are being displaced may be very difficult to replicate in other organizations that lack the financial resources and market dominance of a Bell Labs or the IBM. Finally, because large firms are less willing to pay for the scientific capabilities of the firms they acquire [6], then start-ups will have to focus upon development and commercialization to attract attention from prospective acquirers. If so, the scientist turned entrepreneur may not be enough to save the day. References and Notes [1] NSF Science & Engineering Indicators 2012, Appendix Tables 4-3, 4-8 and 4-9. [2] D. A. Hounshell, Industrial Research and Manufacturing Technology. Encyclopedia of the United States in the twentieth century (1996): 831-57. [3] G. Pisano, The evolution of science-based business: innovating how we innovate. Ind. Corp. Change 19:2, 465-482 (2010). [4] D.C. Mowery, Plus ca change: industrial R&D in the "third industrial revolution". Ind. Corp. Change 18:1,1-50 (2009). [5] R. Buderi, Engines Of Tomorrow: How The Worlds Best Companies Are Using Their Research Labs To Win The Future (Simon and Schuster, 2000). [6] A. Arora, S. Belenzon, A. Patacconi, Killing the Golden Goose? The Decline of Science in Corporate R&D. Working paper no. 20902, (National Bureau of Economic Research, Cambridge, MA. 2015) [7] NSF Science and Engineering Indicators, 2012, Appendix Table 5-49. [8] R. R. Nelson, The simple economics of basic scientific research. J. Polit. Econ. 67:3, 297-306 (1959). [9] A. Arora, A. Fosfuri, A. Gambardella, Markets for Technology (MIT Press, Cambridge, 2001). [10] P. Stephan, How Economics Shapes Science (Harvard University Press, Cambridge, MA, 2012). [11] P. Stephan, Perverse incentives, Nature 484 29-31 (5 April 2012).
Figure 1: Publishing, patenting, and research: Publicly listed U.S. firms, 1980-2007 0.25 Share of patenting firms 437/1,814 firms 0.2 260/1,438 firms 0.15 213/1,438 firms Share of patent citations to science 0.1 0.05 Share of publishing firms 110/1,814 firms R&D intensity 0 Scientific publications Patents R&D intensity Patent citations to science Note: This figure presents the share of publishing and patenting firms of all Compustat firms with positive R&D expenditures, over time. R&D intensity is R&D expenditures over sales. Patent citations to science is the ratio between patent citations to leading scientific journals and the total number of citations the patent makes.
Figure 2: Share of publishing firms by selected industries, 1980-2007 0.4 0.35 0.3 0.25 Chemicals Drugs Drugs Chemicals Electronics IT Machinery Telecommunication 0.2 0.15 Machinery Electronics 0.1 Telecommunications 0.05 Information Technology 0 Note: This figure presents the share of publishing of all Compustat firms with positive R&D expenditures over time by selected industries. Industry classification is based on four-digit main SIC code.
Figure 3: Estimates of elasticity of stock market value with respect to publication and patent stocks over time, 1980-2007 0.25 Elasticity estimate of stock market value with respect to publication and patent stocks 0.2 0.15 0.1 0.05 0 [0.048, 0.161] [0.166, 0.260] [0.161, 0.246] [0.121, 0.203] [0.019, 0.108] [0.032, 0.112] [-0.013, 0.086] [-0.031, 0.045] [-0.060, 0.050] [-0.018, 0.073] [-0.028, 0.050] -0.05 [-0.076, 0.001] 1980-1985 1986-1990 1991-1995 1996-2000 2000-2004 2005-2007 Note: This figure presents estimates of elasticity of stock market value with respect to publication and patent stocks over time. The estimates are from regressing stock market value against firm s assets, R&D stock, publications stock, patents stock (citations-weighed), and complete set of dummies for year and three-digit industry codes.