Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 59
--> 6 Patents and Publicly Funded Research Francis Narin CHI Research Inc. Introduction I have been involved in measuring science and technology for a very long time. I was at the Illinois Institute of Technology (IIT) Research Institute in the late 1960s and was principal investigator in a study called TRACES, with which some of you may be familiar. TRACES was an early attempt to trace events of technological importance back to their origins in science. We carded out studies of magnetic ferrites, the contraceptive pill, the technique of matrix isolation in chemistry, and a few other advances. In all cases, we used experts—individuals who knew the literature on these subjects—to trace back and identify the events that led up to the technology. We traced the evolution of the ideas and technologies over relatively long periods of time. We classified the events as non-mission-oriented research, mission-oriented research, and development and application, and we tried to quantify these various stages. That, in fact, was the real contribution of the original study—the attempt to quantify how and when these different stages of scientific research affected technological development. Sometime after the TRACES study the National Science Foundation (NSF) became interested in bibliometrics. In 1970 we were awarded an early project in bibliometrics, and then around 1971 or 1972, NSF began producing the Science Indicators report, and we worked on that. In fact, CHI Research has basically produced all the literature and patent citation data that has ever been included in Science Indicators, from the 1972 report to Science and Engineering Indicators 2000, which we have not started working on yet. The next Science and Engineering Indicators report is due out in January 1998, and we are just about through with this report. In short, we have been interested in the connection between science and technology for a long time. CHI Research is involved in three different kinds of bibliometric research and analysis. We do a substantial amount of work in what is in a sense classical bibliometrics, which is examining how scientific papers cite other scientific papers. In these studies we are usually examining how well the United States is doing, or how well a given university is doing—that is, how many papers they produce, how often these papers are cited, and whether they are the key papers in their scientific field. A larger amount of classical bibliometrics is done in Europe than in the United States. A group in the Netherlands
OCR for page 60
--> at Leiden is very active in this area, as is the Science Policy Research Unit at Sussex University in the United Kingdom. Similar work is done in Manchester, England. A number of other groups in Europe also carry out large scale bibliometrics research. In part this is because many European countries have a centralized responsibility for science (centralized in one agency or group) and within the European Economic Community there are central groups responsible for the function of the scientific establishment. Europeans thus have an administrative motivation to look at science as a whole and to try to develop tools for measuring performance. We also do a lot of work in technology-oriented areas and in "patents citing patents." Most of this work is done for private clients and is not publicly available. We do a lot of competitive intelligence work and cross-licensing work as well as technology tracing. We have written papers about this work, but generally we do not say much about it. Linkage of Patents to Scientific Papers In this article I discuss the third type of bibliometrics(the linkage of patents to scientific papers. To carry out such studies, we have standardized more than 1 million references to science from the front pages of U.S. and European patents (see Figure 6.1). We put them into a standard "journal, volume, United States Patent 4,713,814 [Inventors] Andrusch et al. (Germany) Dec. 15, 1987 [Assignee] IBM (Armonk, NY) STABILITY TESTING OF SEMICONDUCTOR MEMORIES References Cited U.S. PATENT DOCUMENTS Firms Inventors 3,995,215 11/1976 IBM Chu et al 324/158 4,004,222 1/1977 Semi Corp. Gebhard 324/158 4,418,403 11/1983 Mostek Corp. O'Toole et al 365/201 4,430,735 2/1984 Burroughs Corp. Catiller 371/25 4,502,140 2/1985 Mostek Corp. Prochsting 371/21 4,503,538 3/1985 Robert Bosch GmbH Fritz 371/21 Other references cited: Wiedmann, IEEE Journal of Solid State Circuits, vol. SC-19, no. 3, pp. 282-290, Jun. 1984 FIGURE 6.1 Front page of a U.S. patent.
OCR for page 61
--> page, year" form so that we can match them to a scientific bibliography. Take a look at the front page of a U.S. patent—and this can be done easily; IBM has a wonderful Web site for U.S. patents (http://patent.womplex.ibm.com)—and look at the form of the scientific references. The nonpatent reference is a tribute to the creativity of the American inventor and the patent attorney. The first word can be anything. It can be a journal; it can be a name; it can be part of a title. It can be virtually anything, and it takes a lot of time and effort to turn these references into something that can actually be matched to a scientific paper. At the moment we are standardizing about 5,000 references a week. This allows us to link patents (technology) to publications (science). We need to consider a number of the characteristics of the linkage from patents to scientific publications. First, we are discussing the central references—the ones on the front page of the patent, placed there by both the applicant and the examiner, and passed by the examiner. What we find is that patents are citing papers at a rapidly increasing rate. It has increased by 200 percent in just 6 years for a patent system that has grown by just 25 to 30 percent over that same period. Everything else is changing slowly in the patent system, except the way in which it links to science. The linkage is very subject specific. Patents in biotechnology primarily cite publications in clinical medicine and biomedical research, especially in basic biomedical research; patents in chemistry cite chemistry and chemical engineering publications; patents in computing and communications tend to cite engineering and applied physics papers, especially those published in the journals of the Institute of Electrical and Electronics Engineers. So the linkage is very subject specific, just as citing in a scientific paper is—a chemistry publication largely cites chemistry papers, with a few citing biology or physics papers (i.e., most citations are to a very narrow section of the literature). Citation patterns common in the scientific literature are similar to those in the patent literature. The linkage between patents and scientific publications is also national. U.S.-invented patents heavily cite U.S.-authored scientific papers. German-invented patents cite German scientific papers, Japanese patents cite Japanese papers, and so on. Patent citations are not homogeneous; they are quite national. Perhaps the most startling finding is that 73 percent of the science citations on the front pages of U.S. industry patents are to publicly funded science, that is, to scientific publications from universities, government laboratories, government-funded research and development centers, and other public laboratories. From this it is clear that publicly funded science is having a major impact on U.S. technology. Figure 6.2 shows that inventors in every country in the U.S. patent system are increasingly linking their patents to science and that the patents of U.K. and U.S. inventors are linked particularly strongly to science. Part of this is because the United States and the United Kingdom are heavily involved in biotechnology and drug and medicine patents, which are the most science-linked component of the patent system. But even when adjusting for the specific area of technology, a U.K. patent is more likely to cite scientific papers than is a Japanese patent in the same area. In the U.S. patent system, half of the patents are granted to U.S. inventors and half to foreign inventors. Thus, of the approximately 100,000 patents granted each year, 50,000 or so have U.S. inventors. Of the remaining 50,000 patents, 20,000 have Japanese inventors, and another 20,000 or so have Western European inventors. The other 10,000 are from Canada, Taiwan, Korea, and smaller countries. As an aside, we were examining some patent data the other day and found that both Korea and Taiwan are rapidly approaching the United Kingdom in number of patents granted in the United States—the number of patents issued in these two Asian countries was insignificant 10 years ago. The nationality of the inventor is determined based on the address of the inventor, not on the country in which the company is headquartered. An IBM patent invented in Switzerland, for example, is included in the Swiss patent count, not the U.S. patent count.
OCR for page 62
--> FIGURE 6.2 In the U.S. patent system, the linkage of patents to science is increasing, particularly for the patents of U.S. and U.K. inventors.
OCR for page 63
--> FIGURE 6.3 Average number of references to scientific papers in 1991-1995 patents. The above data quantifies the position of each of these areas on the technology continuum. In this regard, it is interesting to note that human genetics patents have 10 to 15 scientific references per patent. If one looks broadly at the United States and Japan, the Japanese tend to improve on their earlier technology rapidly, often 1 to 2 years more rapidly than U.S. inventors. Japanese patents thus often cite very recent patents. We call this the technology cycle time, and we interpret this as showing that Japanese inventors are masters at rapid incremental adaptation. As noted above, there is a strong science linkage in the patents of U.S. inventors, more so than in Japan. We interpret this is an indicator that the United States is on the leading edge in technology. One of the strongest points of the U.S. technological system is its ability to take science and incorporate it into technology rapidly: The amount of science that is going into these patents from the public sector is increasing at an incredibly fast rate. The science linkage varies greatly with the technology, a point illustrated in Figure 6.3. The average U.S. patent has just one science reference. Patents in the automotive area have almost none; patents in chemistry have close to two. In drugs and medicine, the average U.S. patent has 6 science references, and in human genetics technology, the average U.S. patent has somewhere between 15 and 20 science references. (The number of papers cited also depends on the year that the patent was granted. In fact, this rate is increasing so rapidly that I have to be careful how I phrase my comments.) Genentech's patents, for example, are extremely science linked. The average Genentech patent has more than 25 science references. Genentech has three or four patents on TPA [tissue plasminogen activator], its blood-clot-dissolving agent, with more than 400 related science references. So the distribution is highly skewed, with lots of science linkages in biomedicine, a substantial number in chemistry and some of the advanced areas of electronics, and almost none in the mechanical area. Figure 6.4 illustrates the point I made about subject specificity. The figure shows data on clinical medicine and biomedical patents, plotting the number of references to clinical medicine and biomedical, chemistry, physics, and engineering research journals for the United States, the United Kingdom, Japan, and Germany. In every case, the great majority of the science cited in the patents comes from publications in clinical medicine and biomedical research journals, with some from chemistry journals. For chemical patents we would find the same thing, except that most of the citations would be in chemistry journals, with references to biomedical as well as physics journals. For computers and communications, almost all of the cited papers are in the engineering and physics literature: applied physics, solid-state
OCR for page 64
--> FIGURE 6.4 In the sample of clinical medicine and biomedical patents shown, most of the science cited was published in clinical medicine and biomedical research journals. physics, IEEE, and other engineering journals. As noted above, there is the same kind of subject linkage between patents and the underlying science that is found in the scientific literature itself. Figure 6.5 shows is that there is a strong national component to the citing of scientific publications in patents. The bars are the percent of references from each country' s patents to its own scientific papers, divided by the percent of papers they have. For example, German scientists have about 8 percent of the papers in the Science Citation Index. Roughly 16 percent of the science references in German-invented patents go to German papers. The ratio of 16 percent to 8 percent gives rise to the Germany-Germany bar, which is roughly of height 2. If every country's patents used the world scientific literature homogeneously and there were no significant national component, then every bar would be at 1. This is clearly not the case. U.S.-invented patents heavily use U.S. science, although decidedly less so than Japanese-invented patents use Japanese scientific papers. I want to point out that, if the same analysis were done on the scientific literature to look at how scientific papers cite other scientific papers, the resulting figure would look similar to Figure 6.5. A
OCR for page 65
--> FIGURE 6.5 Citing of scientific publications in patents in 1993-1994 showed a strong nationalistic trend. German scientific paper is two to three times more likely to cite an earlier German paper than expected, based on the volume of German scientific papers in that area. That is, a German scientist will cite his or her own earlier papers, along with the papers of colleagues in the German universities. It is also likely that the area that a scientist is working in is strong in German science rather than weak in German science. All of these effects give rise to a strong nationalistic trend in the citations. Looking at patents in the U.S. patent system, the pattern is exactly the same for German-invented patents. German-invented patents are two or three times as likely to cite an earlier German-invented patent as would be expected by adjusting for the size of the German technological literature. So there is also a strong national component in the patent citations. In general, the cited papers are written in English. This is a set of citations matched to the Science Citation Index, and about 91 percent of the papers in the SCI are in English, so the difference is not primarily because of a language problem. I am sure that some of the bias is attributable to the language barrier, but this must be only a small part of the effect, as most of the scientific literature cited is in English.
OCR for page 66
--> By the way, in our analysis of the patent citations, we did not take out the references to the inventor's own publications. If we took those out, it would change the results a little, but not a lot. For university patents, inventors often cite their own papers. In industry, we find some self-referencing, and we occasionally find it in biomedicine, but it is not found in most other areas. If we corrected for self-references, it would not change the fundamental statistics of the relationship. Figure 6.6 covers the fundamental finding that we made that has recently hit the popular press. It shows the increase in patent-science linkage in three ways. First, there is an increase in the number of papers that are cited, which is illustrated on the left side of the figure. The number of papers cited has increased from 11,000 to 30,000. The central bars show that the number of citations of those papers has increased even more rapidly, because some papers are cited in 3, 4, 5, or sometimes 10 different patents—that is, the number of citations is higher than the number of papers. Citations have increased even more rapidly than the increase in the number of papers cited. The last part of Figure 6.6 shows the number of support sources acknowledged on the cited papers. We have gone to the library and looked up the sources of support (NIH, NSF, DOE, and so on) for nearly 50,000 papers. The number of such acknowledgments is increasing even more rapidly, which says, of course, that there is an ever larger number of acknowledgments in each paper cited. The number of papers that acknowledge two, three, or even four different sources of support is increasing rapidly, possibly because much more collaborative research is going on. Every measure of collaboration that we have ever looked at, from how often papers are co-authored to how often they acknowledge different agencies for support to how often patents cite the papers, is increasing at a steady rate. This is really remarkable. We are talking about the difference between 1987-1988 and 1993-1994 patents—just 6 years! And we find that all of these markers have increased substantially—in a patent system that has increased in size by approximately 30 percent over this period. One of the points I make above is that a large fraction of the scientific references in patents is in biomedicine, and probably 60 percent of all the publications cited in patents are biomedical papers. However, the biomedical literature is extremely large; there are large numbers of papers in clinical medicine and biomedical research. So what we did in Figure 6.7 was adjust for the total number of published papers. This allows us to see how often, in a normalized sense, the different kinds of science are cited. What we found is that biomedical research, the ''basic" field in biomedicine, has more citations per paper than does clinical medicine. Interestingly enough, in biomedicine the citation is preferentially to the basic papers—a patent is much more likely to cite a paper in the Journal of Biological Chemistry than in the Journal of the American Medical Association. This implies that it is basic science that is driving biotechnology, not clinical applications. Note also that there is quite a lot of patent citing to the other sciences, which have far fewer papers than biomedicine. For chemistry and engineering and technology, the papers are not that much less cited, on a per paper basis, than biomedical papers. The number of papers in these fields is much, much smaller than in the biomedical field, but the papers are being heavily used, on a per paper basis, just as is the literature in biomedicine. I want to make a point about the institutions whose papers are cited in patents, as well as the support sources acknowledged in these papers. These publications come from the most prestigious mainstream universities and companies; see Table 6.1. In chemistry, MIT, the University of Texas at Austin, Harvard, DuPont, Berkeley, Bell Labs, and IBM are the most heavily cited institutions. In general, the papers cited in patents come from basic journals. In biomedicine, the research is primarily supported by NIH. In chemistry the National Science Foundation supports far more cited papers than any other agency (see Figure 6.8), followed by NIH (the National Institute of General Medical Sciences [NIGMS] and the National Cancer Institute [NCI]) and DOE. Now, we have not adjusted for research budgets at the different institutions. It would be easier to
OCR for page 67
--> FIGURE 6.6 Three indications of the increase in the linkage of patents to science. normalize by the number of papers published, but we have not done that either. I suspect that when normalized by the number of papers published, the number of industrial papers cited would increase sharply, because industry cites its own papers, as well as university papers. I do not know what the impact would be if we normalized by research budgets. The general rule is that the number of university papers published correlates with the budget, with a correlation coefficient of approximately 0.7. We
OCR for page 68
--> FIGURE 6.7 In the sample shown, biomedical research had more citations per paper than did clinical medicine. TABLE 6.1 Institutions That Originated U.S. Scientific Papers from 1981 to 1991 That Were Cited Most Frequently in U.S. Patents, 1993 and 1994 Biomedical Chemistry Physics Harvard University 2,506 Massachusetts Institute of Technology 171 AT&T Bell Laboratories 854 National Cancer Institute 1,279 University of Texas at Austin 171 IBM Corp. 566 Veterans Administration 1,033 Harvard University 160 Stanford University 300 University of California, San Francisco 930 DuPont Co. 142 Bellcore 174 Stanford University 920 University of California at Berkeley 139 United States Naval Research Laboratory 167 University of Washington 845 AT&T Bell Laboratories 130 Lincoln Laboratory 150 Massachusetts Institute of Technology 756 IBM Corp. 122 Massachusetts Institute of Technology 133 Scripps Clinic and Research Foundation 690 Merck & Co., Inc. 102 University of Illinois at Urbana-Champaign 120 University of California at Los Angeles 642 Cornell University 96 University of California at Santa Barbara 110 Massachusetts General Hospital 625 Texas A&M University 95 Cornell University 106
OCR for page 69
--> FIGURE 6.8 Funding organizations acknowledged in the chemistry papers cited in 1993-1994 patents. Note that these are the numbers of support acknowledgments given in the cited papers, which is not the same as the number of funded papers, since intramural papers often do not list a funding source. Thus, the number of funded papers might be undercounted. would like to look at these issues in more detail but have not had time to do it. We hope to do this in the future. Sources of Science for U.S. Patents I would like to touch on one last point, namely, Where is the science coming from for U.S. industrial patents? To answer this question, we first removed the government patents (for example, NIH's patents), the patents issued to universities, and foreign-invented patents from the list. The result of this analysis is shown on Figure 6.9, the top part of which is for biomedicine. What we found is that about 10 percent of the science base of U.S. biomedicine comes from the U.S. drug industry (private). Fifty-five percent comes from U.S. publicly funded science—that is, from universities, medical schools, government laboratories, federally funded research and development centers, and other public science sources. The other 35 percent is foreign, and the distribution of foreign citations is such that roughly 15 percent of that is industry. Most of the foreign citations refer to public science also. In fact, if U.S. and foreign private companies are taken together, 15 percent of the science cited in U.S. industry drug and medicine patents comes from the private sector, and all of the rest comes from the public sector. To a large degree the same thing is true for chemical patents. The one area that is quite different is computers and communications. Here a large fraction of the cited papers were written by research scientists from industry. This occurs because Bell Labs, IBM, and other major U.S. companies (and overseas, Fujitsu, Hitachi, and other companies) publish many papers, and those papers are heavily cited in patents. There is therefore a fairly large private-sector contribution in computers and communications (34 percent). However, when we put it all together, since most of the citations are in biomedicine, what we find—and this is the bottom line of this whole discussion—is that roughly 73 percent of all the
OCR for page 70
--> FIGURE 6.9 Sources of science base for patents in drugs and medicine (top), chemicals (center), and communications equipment and electrical components (bottom) for 1993-1994 patents. science papers cited on the front page of U.S. industry patents had their origins in publicly funded science—in universities, federal laboratories, federally funded laboratories, hospitals, and research institutes. Discussion Audience Member: Do you have any direct evidence that high citation count is related to commercial importance? Francis Narin: The answer is no. We have one particular client, a large industrial client we have been working with for almost 10 years supporting their cross-licensing. The client calls us at least once a day, and we have people there with whom we work on a regular basis. We have repeatedly asked them if they would tell us which of their patents are most important commercially and which ones are not. They simply will not identify the important ones. However, there is one interesting study that was done by Professor Michael Sherer at Harvard and just submitted for publication. He studied German patents that were applied for in 1977 and had been renewed for 18 years (full term). He went to the companies and got them to say what the economic value of the patents was, in millions of Deutsche marks. He then looked at the relationship between economic value and whether the patents were cited, both within the German patent system and for the patent equivalents filed in the United States. It was clear that the patents that had large economic value had many more citations than the ones that did not. It is nearly a step function. Most of the patents are not highly cited, but a small number are, and most of those were valuable patents. Also, there is a very small set of patents that has been adjudicated by the courts as pioneering patents. Those patents are cited five times as often as other patents.
OCR for page 71
--> I have the impression (and that is all it is) that when we look at patents, if the patents are cited three times as often as the average (after 10 or 15 years, the average patent is cited five or six times, depending on what area it is in), that is, 15 to 20 citations, then that patent is likely to be of technological and economic importance. This set is, perhaps, 10 percent of the patents, but I have no hard data. No one has ever been willing to provide the material with which we could do that analysis. However, companies must think patents are important. It costs at least $10,000 to get a patent, and there are 100,000 or more new patents every year. Audience Member: Does the cost of getting a patent affect the decision to apply? Francis Narin: We do not get involved in it, but most of our clients have committees and groups that do that, and it is a very difficult decision. You are talking about lots of ideas that come up, and it is going to cost at least $10,000 to obtain and maintain a patent. The clients clearly have to be selective. I think they select on the basis of whether the patent will protect their position, but I do not know of any studies or hard data on that. It is a tough question, and one thing you do notice is that when companies are prosperous, they tend to obtain more patents (they have more patent attorneys and will get more patents). When things get tough, they cut back. Those are interesting artifacts in the system. Audience Member: Aren't you exposing your technology by getting a patent? Francis Narin: That is true, but it also provides protection. Remember, a patent is a bargain between yourself and the government. The government protects your rights to the invention. Audience Member: But you are making it public. Francis Narin: The reason the government grants the patent is so that it will be public, so other people can improve on it. That is the whole idea of a patent. What you do see is, in a company that has a strong patent position, there are clusters of interlinked patents. When companies have a weak patent policy, we say that they have a "chicken pox" patent strategy. There are two or three patents here, one over there, and so on, and they do not connect to one another. But when you have a company with interlinked patents, and Alza, a company with highly specialized technology in drug delivery, is one of our classic examples, all their patents are incredibly interconnected. They have built a patent structure that would make it very, very difficult for somebody else to penetrate into that technology. Audience Member: Is the difference between U.S.- and foreign-invented patents due to differences in the patent systems? Francis Narin: Yes, that is another aspect of this issue that I didn't talk about at all. For patents filed in the U.S. system, you expect that the parameters will be the same. However, if the patent originates in Germany, almost always there is a German priority patent, and that certainly influences the way it is written. I think that, in fact, there is a genuinely strong connection between university and government research and industry research in the United States, and that the connection is much stronger (especially historically) than the linkage between, for example, the universities and industry in most areas in the United Kingdom, not necessarily in biotechnology but in most other areas of the United Kingdom. There is a classic paper of perhaps 30 years ago when somebody looked at the British industrial chemical journals and British university chemistry and found that there were very few citations from the
OCR for page 72
--> industrial chemical journals to academic chemistry—a complete disconnection between the two communities. I don't think you would have found that in the United States even then. Audience Member: Are there differences between big and small companies? Francis Narin: Yes, but I have not looked in detail at big versus small companies to see which ones are more science linked, except that in the biotech area the small companies are. The companies that have in the past had this "chicken pox" patent pattern generally have not been companies that had a very strong technological base. Audience Member: Do different groups of companies benefit from public science differently? Francis Narin: We haven't made that cross-link. I do know that in the biotech area the smaller companies are much more science linked than the big ones, and you can understand why. A big company has lots of old technology that it has to protect; a company like Merck will have lots of process patents that are not at the leading edge of biotechnology, whereas at a biotech start-up, everything is based on leading-edge research. The interesting question is whether those differentiations will give some way of predicting whether a company is going to do well. Audience Member: Do patents relate to company success? Francis Narin: We are just beginning to explore that. We are starting to look at IPOs, initial public offerings. One interesting aspect is whether companies that have been successful and had successful IPOs in biotech are the ones linked into public science. I think that they will be, but I don't yet have any hard data. Right now I am trying to get a project going to take a look at that. Audience Member: Wasn't science developed by industry long before it was supported by public agencies? Francis Narin: I think the term "science" was used differently in those days. If you read the history of the last century, they were really talking about technology and not science. For example, during the industrial revolution in the United Kingdom, they used the term "science" to describe what we would call "technology."
Representative terms from entire chapter: