R&D
and the Patent Premium
NBER Working Paper No. 9431 (January 2003)
by Ashish Arora (Carnegie Mellon), Marco Ceccagnoli (Insead), and Wesley M. Cohen (Duke)
NBER Working Paper No. 9431 (January 2003)
by Ashish Arora (Carnegie Mellon), Marco Ceccagnoli (Insead), and Wesley M. Cohen (Duke)
--summarized by Robert M. Hunt* (Federal Reserve Bank of Philadelphia)
Summary
This is an important paper that presents many interesting estimates of the economic value and effects of patents. It is also a rather long and technical paper. The authors are careful researchers and describe their modeling assumptions and estimation approach in great detail.
They ask a critical question: what is the relationship between the effectiveness of patent protection, firms’ utilization of the patent system, and their investments in research and development (R&D)?
In standard theoretical models, stronger patents imply that there should be both more patent activity and more R&D. But these implications are sometimes reversed in more complicated models, and theory alone can rarely convey a sense of the magnitudes or sensitivity that are important for policy recommendations.
Using data collected in a 1994 survey by Carnegie Mellon, the authors are able to derive estimates of the economic value conferred by patent protection on patent holders and the resulting effect on a firm’s R&D decisions. Assuming that firms are rational in their decisions to apply for patents, the authors find that patents significantly increase the private value of inventions. They also find a positive association between the effectiveness of patents, as reported in the Carnegie Mellon survey, and private R&D spending.
The Approach
The authors derive a system of three equations that explain firms’ R&D investments, the number of resulting innovations, and the firms’ subsequent patenting decisions. The challenge here is that several important variables, like the number and value of inventions, and the cost of obtaining patents, are not readily observable. Inferences about the number and value of inventions are made by assuming an explicit production function for inventions, and that the distribution of patent premia follow the normal distribution.
Variables from the Carnegie Mellon Survey (see Cohen, Nelson, and Walsh 2000) help to do the rest of the work. In addition to gathering a host of information on the characteristics of firm and their R&D labs, that survey asked R&D managers about their R&D spending, the number of patents they apply for, and the percentage of product innovations they sought to protect with patents. R&D managers were also asked to report the percentage of product innovations where patents helped to protect the competitive advantage that their inventions gave their firms.
Results
The authors derive two measures of the patent premium: an expected premium and a conditional premium. The former is the amount a firm would expect to gain if it patented all its innovations, while the latter is the amount the firm would expect to gain if it patented only those innovations it would actually choose to patent.
They find that, on average, the expected premium is 0.52—in other words, obtaining a patent on any given innovation is likely to reduce its value by about 50 percent. The expected premium varies considerably across industries—it is only 0.01 for electronic components, like semiconductors, but as high as 1.34 for biotech firms.
But firms do not apply for patents on all their innovations. They do so only where the expected increase in the value of the invention exceeds the cost of patenting. The conditional premium earned on innovations satisfying this incentive constraint is estimated to be 2.25—in other words, a patent doubles the value of inventions that firms would actually choose to patent.1
The authors consider two measures of the effect of patent protection on R&D spending. The first is a counterfactual exercise. Suppose that patents did not exist. To capture this, we substitute the value of unpatented inventions for patented ones in the equation that explains firm R&D investments. The result is a 25 to 35 percent decline in R&D spending. Of course, this calculation assumes that all other economic and strategic factors remain the same.
As an alternative, the authors measure the incremental effect of an increase in the patent premium, holding all other variables constant. They find that a 10 percent increase in the patent premium raises R&D spending by 6 percent and the number of patent applications by 12 percent. The effects are most pronounced for firms in the drug, biotech, and medical equipment industries. This elasticity is somewhat difficult to interpret, the authors caution, because the model does not address how we might generate such an increase in the patent premium.
The paper reports many other interesting facts and results. These include the following:
• The average firm reports that a patent helps to protect the competitive advantage conferred by a product innovation about 38 percent of the time. This measure varies significantly across industries. It is highest in drugs, biotech, and medical instruments. It is lowest for food and tobacco products, and for electronic equipment.
• Firms apply for patents on about one-third of their innovations, but this rate varies significantly across industries. Patent propensity is highest in drugs, biotech, and medical instruments. Manufacturers of computer and office equipment apply for patents on about 40 percent of their innovations. Patent propensity among firms in the semiconductors and other electronic components industries is only about 20 percent.
• The authors estimate that there are, on average, 5.6 patent applications per innovation. Drugs and biotech generate the lowest number of applications per innovation (about 2) while semiconductors, transportation equipment, and rubber products generate the most (7 to 9).
• The authors estimate that, on average, a 10 percent increase in firm R&D spending yields a 5 percent increase in the number of innovations.
• The productivity of firm R&D is positively influenced by information spillovers from universities and government labs. The estimated effect is quite large. The evidence for spillovers from rival firms is mixed.
• The authors estimate that the cost of obtaining patents on an innovation represents about 35 percent of the value of an unpatented innovation. This estimates clearly reflects more than the observable cost of filing for patents.
Qualifications
The authors are careful in describing their approach and characterizing their results. For example, they point out that their model and data do not address questions about the potential relationship between the effectiveness of patents, or the patent premium, on firm entry or industrial structure.
The authors are also concerned about the effect of strategic interactions between firms on the relationship between the strength of patents and the incentive to invest in R&D. For example, the authors point out that their patent effectiveness variable does not seem to reflect R&D managers’ concern about defensive patenting—the practice of obtaining patents as bargaining chips against potential litigation initiated by rivals.
Their estimation technique attempts to control for such factors. In a more static environment, we can think of holding constant the effect of any strategic interactions. But in a more dynamic environment the effect of strategic interactions is likely to change, perhaps because patent law itself is changing. The regression approach used in this paper may not adequately identify these effects. In the authors’ words, “our results also point to the need for a more fully elaborated model incorporating competitive interactions that can help disentangle the various effects.”
Conclusion
This paper represents an important step forward in empirical research on the effects of patents. The authors present estimates of the incremental value conferred by patent protection and the implied effect on firms’ incentives to conduct R&D. Both effects are economically significant.
Sources
Cohen, Wesley M., Richard. R. Nelson, and John P. Walsh, “Protecting Their Intellectual Assets: Appropriability Conditions and Why U.S. manufacturing Firms Patent or Not,” NBER Working Paper No. 7552 (February 2000).
1. Note that these coefficients are gross of the estimated cost of obtaining a patent. If we net out the estimated cost, the expected premium is 0.17 and the conditional premium is 1.91.
*The views expressed here are those of the author and do not necessarily represent the views of the Federal Reserve Bank of Philadelphia or the Federal Reserve System.
This is an important paper that presents many interesting estimates of the economic value and effects of patents. It is also a rather long and technical paper. The authors are careful researchers and describe their modeling assumptions and estimation approach in great detail.
They ask a critical question: what is the relationship between the effectiveness of patent protection, firms’ utilization of the patent system, and their investments in research and development (R&D)?
In standard theoretical models, stronger patents imply that there should be both more patent activity and more R&D. But these implications are sometimes reversed in more complicated models, and theory alone can rarely convey a sense of the magnitudes or sensitivity that are important for policy recommendations.
Using data collected in a 1994 survey by Carnegie Mellon, the authors are able to derive estimates of the economic value conferred by patent protection on patent holders and the resulting effect on a firm’s R&D decisions. Assuming that firms are rational in their decisions to apply for patents, the authors find that patents significantly increase the private value of inventions. They also find a positive association between the effectiveness of patents, as reported in the Carnegie Mellon survey, and private R&D spending.
The Approach
The authors derive a system of three equations that explain firms’ R&D investments, the number of resulting innovations, and the firms’ subsequent patenting decisions. The challenge here is that several important variables, like the number and value of inventions, and the cost of obtaining patents, are not readily observable. Inferences about the number and value of inventions are made by assuming an explicit production function for inventions, and that the distribution of patent premia follow the normal distribution.
Variables from the Carnegie Mellon Survey (see Cohen, Nelson, and Walsh 2000) help to do the rest of the work. In addition to gathering a host of information on the characteristics of firm and their R&D labs, that survey asked R&D managers about their R&D spending, the number of patents they apply for, and the percentage of product innovations they sought to protect with patents. R&D managers were also asked to report the percentage of product innovations where patents helped to protect the competitive advantage that their inventions gave their firms.
Results
The authors derive two measures of the patent premium: an expected premium and a conditional premium. The former is the amount a firm would expect to gain if it patented all its innovations, while the latter is the amount the firm would expect to gain if it patented only those innovations it would actually choose to patent.
They find that, on average, the expected premium is 0.52—in other words, obtaining a patent on any given innovation is likely to reduce its value by about 50 percent. The expected premium varies considerably across industries—it is only 0.01 for electronic components, like semiconductors, but as high as 1.34 for biotech firms.
But firms do not apply for patents on all their innovations. They do so only where the expected increase in the value of the invention exceeds the cost of patenting. The conditional premium earned on innovations satisfying this incentive constraint is estimated to be 2.25—in other words, a patent doubles the value of inventions that firms would actually choose to patent.1
The authors consider two measures of the effect of patent protection on R&D spending. The first is a counterfactual exercise. Suppose that patents did not exist. To capture this, we substitute the value of unpatented inventions for patented ones in the equation that explains firm R&D investments. The result is a 25 to 35 percent decline in R&D spending. Of course, this calculation assumes that all other economic and strategic factors remain the same.
As an alternative, the authors measure the incremental effect of an increase in the patent premium, holding all other variables constant. They find that a 10 percent increase in the patent premium raises R&D spending by 6 percent and the number of patent applications by 12 percent. The effects are most pronounced for firms in the drug, biotech, and medical equipment industries. This elasticity is somewhat difficult to interpret, the authors caution, because the model does not address how we might generate such an increase in the patent premium.
The paper reports many other interesting facts and results. These include the following:
• The average firm reports that a patent helps to protect the competitive advantage conferred by a product innovation about 38 percent of the time. This measure varies significantly across industries. It is highest in drugs, biotech, and medical instruments. It is lowest for food and tobacco products, and for electronic equipment.
• Firms apply for patents on about one-third of their innovations, but this rate varies significantly across industries. Patent propensity is highest in drugs, biotech, and medical instruments. Manufacturers of computer and office equipment apply for patents on about 40 percent of their innovations. Patent propensity among firms in the semiconductors and other electronic components industries is only about 20 percent.
• The authors estimate that there are, on average, 5.6 patent applications per innovation. Drugs and biotech generate the lowest number of applications per innovation (about 2) while semiconductors, transportation equipment, and rubber products generate the most (7 to 9).
• The authors estimate that, on average, a 10 percent increase in firm R&D spending yields a 5 percent increase in the number of innovations.
• The productivity of firm R&D is positively influenced by information spillovers from universities and government labs. The estimated effect is quite large. The evidence for spillovers from rival firms is mixed.
• The authors estimate that the cost of obtaining patents on an innovation represents about 35 percent of the value of an unpatented innovation. This estimates clearly reflects more than the observable cost of filing for patents.
Qualifications
The authors are careful in describing their approach and characterizing their results. For example, they point out that their model and data do not address questions about the potential relationship between the effectiveness of patents, or the patent premium, on firm entry or industrial structure.
The authors are also concerned about the effect of strategic interactions between firms on the relationship between the strength of patents and the incentive to invest in R&D. For example, the authors point out that their patent effectiveness variable does not seem to reflect R&D managers’ concern about defensive patenting—the practice of obtaining patents as bargaining chips against potential litigation initiated by rivals.
Their estimation technique attempts to control for such factors. In a more static environment, we can think of holding constant the effect of any strategic interactions. But in a more dynamic environment the effect of strategic interactions is likely to change, perhaps because patent law itself is changing. The regression approach used in this paper may not adequately identify these effects. In the authors’ words, “our results also point to the need for a more fully elaborated model incorporating competitive interactions that can help disentangle the various effects.”
Conclusion
This paper represents an important step forward in empirical research on the effects of patents. The authors present estimates of the incremental value conferred by patent protection and the implied effect on firms’ incentives to conduct R&D. Both effects are economically significant.
Sources
Cohen, Wesley M., Richard. R. Nelson, and John P. Walsh, “Protecting Their Intellectual Assets: Appropriability Conditions and Why U.S. manufacturing Firms Patent or Not,” NBER Working Paper No. 7552 (February 2000).
1. Note that these coefficients are gross of the estimated cost of obtaining a patent. If we net out the estimated cost, the expected premium is 0.17 and the conditional premium is 1.91.
*The views expressed here are those of the author and do not necessarily represent the views of the Federal Reserve Bank of Philadelphia or the Federal Reserve System.
© 2003. Verbatim copying and distribution of this entire article for noncommerical use are permitted provided this notice is preserved.
