2005/04 Research & Review

In What Areas of Technology is Knowledge Transfer by Means of Patent Effective?

TAMADA Schumpeter
Fellow, RIETI

Purpose of the study: to enhance understanding of the relationships among science, technology and innovation

The importance of industry-academia collaboration (i.e., cooperation between companies and public research institutions, including universities) has been a frequent topic of study. Since the second half of the 1990s, a range of legislative and administrative steps have been taken to promote such collaboration, including enactment of the Business Innovation Law and the Technology Licensing Organization (TLO) Law, along with budgetary and tax incentives. Meanwhile, national universities and research institutes have undergone major reforms including their transformation into incorporated administrative agencies. Collaboration between industry and academia, which was once was a taboo in Japan, has now become an important theme that is being discussed as a means for universities and other public research institutes to contribute to society.

Collaboration with academia is also crucial for the business community. With the shortening of the time between development of new science and its application in industry, scientific knowledge is becoming increasingly important as a problem-solving tool in many sectors. To develop a highly sophisticated product or a production method, it is often necessary to aggregate and combine multiple elemental technologies. It is no easy task, however, to master all the advanced scientific knowledge behind these elemental technologies. Companies therefore typically focus on particular areas for in-house research and business projects, while seeking collaboration with universities and other external organizations to develop advanced knowledge in other areas. Against this backdrop, many Japanese companies have come to attach greater importance to collaboration with academia.

As such, the importance of industry-academia collaboration is now generally recognized throughout society. But there has been little discussion of concrete issues, for instance, the type of collaboration that should be sought between particular universities or a public institutes and particular companies. Universities are diverse, offering various fields of study - mathematics, physics, chemistry, biology and others. Technologies are just as varied and include biotechnology, nanotechnology, information technology (IT) and many more. Furthermore, when we look at the relationship between science, technology and industry, we can see a number of possible combinations. For instance, an encryption technology utilizing a mathematical theory may improve the productivity of the financial industry. The development of a jet engine technology that improves fuel efficiency by applying principles of fluid dynamics may enhance the productivity of the transport industry. Thus we can assume that science, technology and industrial innovation are related to one another in a three-dimensional network structure. But we have yet to fully understand how these three elements are linked, and whether or not the strength of the linkages differs between them.

Methodology: Measuring the science linkages in all fields of technology

This paper uses "patents" - a novel and advanced technological idea whose usability has been recognized in industry - as a barometer of technology, and "research papers" - research findings by universities and institutions that have been tuned into explicit knowledge and released to the public - as a barometer of science. By examining the number of research papers cited in patent documents in each technological field, I conducted a quantitative analysis of the relationships between technology and science. Specifically, I have taken the following steps.

First, using patent gazettes published from 1994 to 2001, I created a database containing some 880,000 patents. Then, out of some 650,000 patents published from 1995 to 2000, I selected 1,200 samples falling under four technology categories (i.e., 300 samples each for biotechnology, nanotechnology, information technology (IT) and environmental technology) that have been designated as priority areas under the government's Second Science and Technology Basic Plan. In addition, I randomly selected 300 patents from the entire population as control samples for comparison. Then, I visually examined these 1,500 sets of patent information - 1,200 relevant to the four priority technology categories and 300 from the randomly-selected samples - to extract research papers cited therein (see 2004/12 Research & Review, Keizai Sangyo Journal, METI, for further details).

Obviously, however, there is a limit to the practical applicability of this method: It is extremely time- and labor-intensive; thus, it is virtually impossible to manually extract the cited literature from all the patents covered in this survey. In order to clear this hurdle the work procedures had to be automated, but there is no standard method of citation in patent documents. A filtering program based on a simple algorithm would therefore fail to extract some of the cited literature, or it might mistakenly extract certain strings of letters that are not actual citations.

In order to overcome this problem, I developed unique software based on a finite-state machine algorithm, using information concerning the visually-extracted citations as a master. Using this software for automated counting, I determined the number of research papers cited in all patents issued from 1995 to 1999.

Each of the over 650,000 patents is assigned a technology category (hereafter referred to as a "main IPC code") based on the International Patent Classification (IPC), under which technical inventions are classified into eight sections with each section further subdivided into smaller units in a hierarchical manner. Specifically, each section is divided into classes, each class into subclasses, each subclass into main groups, and then, each main group into subgroups.

I then classified the patents into mutually exclusive subgroups and counted the number of research papers cited in each patent. Then, for each sub-group, I added the number of cited papers and divided the sum by the number of patents included in that sub-group, with the resulting number being an indicator of the intensity of technology-science linkage for that technological category.

Findings: Strength of science linkage differs markedly by technological category

Japanese patent documents do not always provide a list of cited papers on the front page because such listing is not mandatory. Therefore, to gain a comprehensive understanding of the relationship between technology and science, it is necessary to create a filtering program that automatically extracts cited papers from the full text of patent documents, including patent specifications. Through machine learning, I was able to develop a program that has attained a very high level (98% or more) of both precision and recall ratio (see Table 1). Similar attempts have been made in Europe but the levels of precision and recall rate remain around 70%. The levels achieved through this research are thus the highest in the world. The program allows automated extraction of research paper citations and preceding patents from all the patent data contained in the database, thus enabling comprehensive analysis of science linkage for any given IPC technology category.

 
Patent reference
Non-patent reference
Category
Patent cited
Noise
Leakage
Non-patent cited
Noise
Leakage
Environment/Unexamined
Environment/Examined
531
1296
1
0
0
5
55
73
2
0
11
3
Mixed/Unexamined
Mixed/Examined
1355
2342
7
14
6
2
435
672
6
11
18
32
IT/Unexamined
IT/Examined
234
977
2
5
4
6
46
115
8
2
2
18
Biotechnology/Unexamined
Biotechnology/Examined
875
1322
14
25
6
13
3420
4267
112
22
31
32
Nanotechnology/Unexamined
Nanotechnology/Examined
476
1867
4
1
2
1
83
213
11
3
11
14
Total
11275
73
45
9379
177
172
 
Recall ratio: 99.6%
Recall ratio: 98.2%
 
Precision ratio: 99.4%
Precision ratio: 98.1%

Table 1: Results of the automated reference extraction system

In this research, I measured the intensity of science linkage for each of some 600 technology categories with respect to patents published in patent gazettes between 1995 and 1999 falling under these categories. In Figure 1, the horizontal axis shows technology subclasses and the vertical axis shows the average science linkage for each subclass. As evident from Figure 1, the intensity of science linkage differs substantially for each technological category. For instance, in the case of "C12N Microorganisms or enzymes; compositions thereof," the number of research papers cited per patent averages nearly 15, which is roughly 30 times the average for all categories (0.5 per patent).

Figure 1: Significant difference in science linkage among different technological classifications

Figure 1: Significant difference in science linkage among different technological classifications

Table 2 shows the top 20 subclass categories of Japanese patents in terms of the average number of research papers cited per patent. The category with the strongest science linkage by this measure is "C12N Microorganisms or enzymes; compositions thereof" at 14.6," followed by "C07K Organic chemistry, peptides" at 12.2. "C12Q Measuring or testing processes involving enzymes or microorganisms, composition or test papers thereof, processes of preparing such composition, condition-responsive control in microbiological or enzymological processes" has the third-strongest linkage at 7.6. The average for all categories is 0.5.

Table 2: Top 20 subclasses by average science linkage

Table 2: Top 20 subclasses by average science linkage

These results concerning Japanese patents are similar to the trends observed in European patents as shown in research conducted by Michel et al of the European Patent Office. Of the top 10 IPC subclass categories of Japanese patents, which have been automatically extracted and ranked in order of average number of research paper citations, six subclass categories are also found among the top 10 subclasses of European patents (enclosed in boldface boxes). In addition, the top three subclass categories are identical for Japanese and European patents. Japanese and European patents thus show a similar pattern in the intensity of science linkage. This suggests the differences in the relationship between technology and science are not attributable to patent nationality but rather that the degree of reliance on scientific knowledge differs depending on the type of technology being patented.

Subclass categories with strong science linkage are mostly related to biotechnology but also include those related to nanotechnology. Of the top 20, ranked fifth is "G03C, Photographic-sensitized material, photography (e.g., motion pictures, X-ray photography, multicolor photography, stereoscopic photography), auxiliary photographic processing methods"; while "G09C, Ciphering or deciphering apparatus for cryptographic or other purposes involving the need for secrecy"; "G06E, Optical calculating machine"; "G10L, and Analysis or synthesis of speech, speech recognition" rank 11th, 18th and 19th respectively. These four subclass categories belong to Section G (physics) in the IPC system and these subclasses are assumed to be IT-related technologies.

Consideration: Fine-tuned policies are needed for each technology field

The results of this research suggest the intensity of linkage between science and technology differs significantly by technology. The degree of science linkage, as measured by the number of research papers cited in patent documents, is particularly strong in biotechnology. Among other types of technology, those related to photographic-sensitized material, cryptography, optical computing, and speech recognition also show strong science linkage.

These results are consistent with other research findings. For instance, according to remarks by made David Mowery, a professor at the University of California, at a symposium held under the auspices of the Corporate Innovation System Renaissance Project, 90% of the $15 million earned by the university from patent royalties come from the field of biotechnology. Stanford University's Office of Technology Transfer (OTL) owes most of its success to a patent for gene splicing technology. Also, many of the successful university-oriented venture businesses, in which university professors corporatize their scientific knowledge, operate in biotechnology and IT-related fields.

According to Goto and Nagata (1997), whereas the validity of patents is one of the most highly valued factors in appropriating innovation in the pharmaceutical industry, patent protection ranks only as the sixth most important factor in terms of average score in a survey 826 American companies. For these firms, in order of importance, the five most important factors were: lead time, secrecy of technological data, management of production facilities and know-how, management of distribution and service networks, and complexity of production and product design.

The findings above suggest that different innovation mechanisms are working in different fields of technology. In the case of pharmaceuticals and agricultural chemicals, a company which has a substance patent for a particular innovation tends to be able to claim and secure its rights based on the molecular architecture of the resultant chemical compound even if another company produces such a product through a different process. By contrast, in the case of aircraft and automobiles, for example, numerous parts and components, along with various patents and production technologies are involved in producing a final product. A great deal of emphasis is also placed on the quality of after-sale services. Meanwhile, in the case of electronics, the pace of technological innovation is so rapid that priority is given to the speed of new product development through modularization, and to increasing product differentiation. Considering all these factors, it is quite natural that different types of innovations are appropriated in different ways.

Such differences in innovation processes by type of technology must be fully taken into account when launching university-industry collaboration. For instance, in sectors such as pharmaceuticals and certain areas of IT, where a patent provides high appropriability of innovations and both manufacturing scale and production costs are relatively low, it is possible to transfer technology through patent licensing and thereby to launch a university-based venture business. In such sectors as aerospace and automobiles, however, where products are highly complex, manufacturing costs are high, and a sophisticated service network is indispensable, major companies will likely continue to dominate. Meanwhile, in areas where process innovation is important, companies will be able to maintain competitiveness by holding their proprietary technology in absolute secrecy, rather than by revealing such technology through patents. Companies falling into this last type include Okano Industrial Corp., a Tokyo-based company renowned for its precision molding technology, and Sharp Corp. which is known for its strategy of maintaining the secrecy of its process technology for liquid crystal displays.

Needless to say, it is pointless and even harmful to blindly press university researchers to apply for patents. In certain fields of technology, such researchers can hardly expect licensing revenues and the additional time and effort required to file patent applications may hamper their research activities. Where universities can especially demonstrate their strength is in developing capable human resources equipped with knowledge and expertise by offering high-quality education, and by transforming their tacit knowledge obtained through research into codified knowledge in the form of research papers, thus making public goods in the form of knowledge accessible to all. As discussed above, the appropriation of innovation by means of patents and the transfer of technology through licensing are likely to be most effective in the pharmaceutical industry and some sectors of the IT industry.

>> Original text in Japanese

Reference(s)
  • Mansfield, E. (1991), Academic Research and Industrial Innovation, Research Policy, 20 (1):1-12.
  • Michel, J., Bettels B. (2001), Patent citation analysis, Scientometrics, 51(1):185-201.
  • Narin, F., Hamilton, K. S., Olivastro, D. (1997), The increasing linkage between U.S. technology and public science, Research Policy, 26(3):317-330.
  • OECD (1990), University-Enterprise Relations in OECD Member Countries. OECD: Paris.
  • Tamada et al., RIETI Discussion Paper Series 04-E-034, 2004.
  • Goto Akira, Odagiri Hiroyuki, et al., "Saiensu-gata Sangyo (Science-based Industries)," NTT Publishing Co.
  • Goto Akira and Nagata Akiya (1997), "Inobeshon no Senyu Kanosei to Gijutsu Kikai (Appropriability of Innovation and Technology Opportunities)," National Institute of Science and Technology Policy.
  • Tamada Schumpeter, Kodama Fumio, Genba Kiminori (2002), "Nihon Tokkyo ni okeru Saiensu Rinkeji no Keisoku: Inyo Bunken Detabesu Kochiku ni yoru Idenshi Kogaku Gijutsu Bunya Tokkyo no Bunseki (Measuring of Science Linkage in Japanese Patents: Analysis of Patents in Genetic Engineering Technologies by Establishing Database of Cited Litarature)," Kenkyu Gijutsu Keikaku (Journal of Science Policy and Research Management), 17 (3/4) : 222-230.
  • Nakayama Ichiro, "'Puro-Patento' to 'Anchi-Komonzu' ('Pro-Patent' and 'Anticommons,'" RIETI Discussion Paper Series 02-J-019.
  • Ministry of Education, Culture, Sports, Science and Technology, "White Paper on Science and Technology 2001," National Printing Bureau, Tokyo.

August 1, 2005

Article(s) by this author