[This document was published in the Journal of Technology Transfer Vol. 22 (3) 65-72, 1997 and is based on a study performed for the Agricultural Research Service of the U.S. Department of Agriculture by CHI Research, Inc.]

Abstract

One of the most difficult challenges in technology transfer is to measure the movement of knowledge from basic scientific research to industrial technology. This paper will report on a study of the linkage between science supported by the Agricultural Research Service (ARS) and patented technology. This study traced the citations from U.S. patents issued in 1987-88 and 1993-94 to scientific research papers linked to the U.S. Department of Agriculture (USDA). The number of patent citations to ARS papers, and to other USDA-supported papers has increased four-fold over the six-year period. A distinct difference also exists between the patent-cited ARS papers and patent-cited extramural USDA-supported papers: ARS papers are in more agriculturally related journals, while the extramural papers were in more basic and biomedical journals. USDA-supported papers were overwhelmingly cited by U.S.-invented patents (in a patent system in which half the patents are foreign-invented). In the primary field of ARS papers (Biology), they are cited much more often by patents than Biology papers from any other publishing organization. Since the publishing organizations and support sources of all the papers cited in these patents have now been identified, we can study the transfer of scientific results to patented technology by institution, by agency, or by any other category of patent or paper holder.

I. Introduction.

Many scientists and economists believe that public science is a driving force behind high technology and economic growth. They also believe that the transfer of publicly supported knowledge to industry is an important part of the technology transfer process. For purposes of this paper, "public science" means scientific research done in academic and governmental research institutions, and supported by governmental and charitable agencies. The core of this paper is a case study of the links between science supported by a governmental agency (the U.S. Department of Agriculture (USDA), particularly the Agricultural Research Service (ARS)) and technology represented by patents issued by the U.S. Patent & Trademark Office (USPTO).

We study the links between science and technology by examining the references to scientific and engineering papers on the front pages of U.S. patents. These science references are contained among the larger set of "non-patent references" (NPRs) that appear there. NPRs are a mixed set of references to journal papers, meetings, books, and many non-scientific sources, such as industrial standards, technical disclosures, engineering manuals, and every other conceivable kind of published material.

References to previously issued patents and non-patent references both comprise the "prior art" of the patent. These "references cited" on U.S. patents are a fundamental requirement of patent law. To be issued, a patent must satisfy three general criteria: it must be useful; it must be novel; and it must not be obvious. The novelty requirement is the main factor leading to the references that appear on the front page of the U.S. patent. It is the responsibility of the patent applicant and his attorney, along with the patent examiner, to identify, through various references cited, all of the important prior art on which the patent improves. These references are chosen and screened by patent examiners, who are "not called upon to cite all references that are available, but only the best." ("Manual of Patent Examining Procedures," Section 904.02)

Scientific journal papers represent the distillation and refinement of months or years of work. Unlike less formal means of scientific communication that also appear in patent NPRs, journal papers are usually reviewed by the authors’ peers and scrutinized by editors before publication is permitted. Also filtered out of them by the authors are the dead ends, mere progress reports, and inconsequential negative results. Consequently, scientific and engineering journal papers are the most authoritative records of research. Databases such as the Science Citation Index (SCI) contain detailed records of these papers and citations to them from other papers, allowing detailed analysis.

All this suggests that scientific and engineering papers cited by patents are a unique source of information on the transfer of scientific results to patented innovations. For example, universities and government agencies can identify potential technologies and companies with which they can collaborate in cooperative research and development agreements. If they discover they are contributing their scientific expertise to a technology, they may move to benefit more directly from their efforts by lending their scientific expertise directly to the development effort in the private sector.

This case study is a good illustration of the various insights one gets from studying this patent-to-paper linkage. Among the main results of the USDA study are the following: 1) Citations by patents to USDA-supported papers have more than quadrupled within six years; by comparison, citations from patents to all U.S. papers have increased about half that; 2) Collaboration between ARS and extramural institutions, mostly universities, has increased, and cited paper authorship is split evenly between ARS and extramural; 3) Patent-cited ARS research is predominantly applied, while cited extramural research is more basic; 4) USDA-supported papers are preferentially cited by U.S. inventors; 5) For at least one year (1988) ARS is the leader in patent-cited biological research, with ARS intramural papers cited 70 percent more than average for the biology papers of all U.S. universities and organizations.

Following sections devoted to Methodology and the Results of the study, the Conclusions will examine the possible use of such results as a quantitative indication of the transfer of knowledge from science to technology.

II. Methodology.

As part of a series of research contracts and grants with the National Science Foundation (NSF), CHI Research, Inc., (CHI) has matched journal references found on the front pages of U.S. patents to bibliographic data in the NSF’s Science Literature Indicators Database (SLID). This is a data set based on the SCI and maintained at CHI for the National Science Foundation. Through this, we could positively identify most scientific and engineering papers cited by patents. (Some references had insufficient data to match to the SCI, typically about 25 percent.) These matched papers form the basic data set.

On the 398,000 patents issued in the years 1987-88 and 1993-94, there were 430,000 NPRs. Of these NPRs, about 242,000 are science references—citations to scientific journal papers, scientific meetings, and other scientific publications. About 175,000 of these science references are to papers published in the 4000 journals covered by the SCI. CHI unified these 175,000 into a standard journal name, volume, page, and year format.

From here we confine ourselves to papers published in an eleven-year window before patent issuance date. There were 40,000 citations to papers published in 1975-85 cited by patents issued in 1987-88, and 104,000 to papers published in 1981-91 cited by patents issued in 1993-94. This window covers 80 percent of the citations. We then matched these 144,000 citations to our SLID database, and successfully identified 109,000 (76 percent) of the citations. Most of the unmatched references were incomplete or incorrect, with misspellings, wrong page numbers, etc.

The next step identified the author addresses on the papers. The SLID contains a code for each author institution, standardized for more than 3500 major U.S. publishing institutions.

The final step identified the agencies supporting the U.S.-authored papers. CHI staff searched libraries for the 45,000 matched papers that had at least one U.S.author, and that were cited by 1987-88 and 1993-94 patents. They found about 95 percent of them, and tabulated all the U.S. and foreign research support information contained in the acknowledgements. Of these 95 percent, about 63 percent acknowledged some source of external support, most frequently in academic papers. Papers with authors in governmental agencies and the private sector seldom have funding acknowledgements, since most of their funding is internal.

Several methodological points are important. The first is data complexity. One can count the data in a variety of ways: a single paper can be cited by several patents and a single patent can cite many papers. In addition, the papers have multiple authors from multiple institutions, supported by more than one agency. All this results in various ways to count and present data.

Another is the limitation to science references on the front page of the patent, ignoring references in the text of the patent. The front page references should be the most important, since the patent examiners rely specifically on them to establish the patent’s novelty, as mentioned in the Introduction. Further, from a practical standpoint, these text references are extremely difficult to extract, as they are scattered about an enormous amount of text. CHI did a brief study to check how representative are the front page science references. (Narin, Rosen, and Olivastro 1988) It showed much similarity between the front page and text references. We therefore assume that the front page references are sufficiently representative. Indeed, the use of them alone would underestimate the contribution of science to the patents.

From the linkage database of papers cited by patents issued in 1987-88 and 1993-94, we extracted those that acknowledge some USDA funding support. In addition, we identified those papers with at least one USDA intramural author and added them to the data set. More than 90 percent of the intramural papers were written by ARS researchers. These are not, however, the complete set of all agricultural science papers cited by U.S. patents, nor the complete set of all agricultural patents issued in these years. Our data are limited to patents citing USDA-related papers.

CHI is currently preparing a paper with a comprehensive analysis of the linkage between U.S. technology and public science. (Narin, Hamilton, and Olivastro 1997) Two previous CHI papers have addressed this data in preliminary form. (Narin and Olivastro 1992; Narin, Hamilton, and Olivastro 1995)

In 1987-88 and 1993-94, the USPTO issued 468 patents, citing 596 papers with at least one ARS author (including some with extramural collaborators), resulting in 771 citations in all. The same numbers for extramural, USDA-supported papers (non-ARS) are 433 USPTO patents citing 510 papers for a total 727 citations. Figure 1 shows percent increases in citations between the two two-year periods. Citations to all USDA-supported papers increase over four-fold between the two periods. Extramural citations—mostly to universities—increased more than seven-fold. Since, during this period, extramural funding was increasing for the USDA, the unusually large increase in citing to extramural papers may be a reflection of increased numbers of extramural publications.

Figure 1. Percent Increases in Citations, 1987-88 and 1993-94

Indeed, the numbers of all journals and papers have increased steadily over this time (especially in biomedicine), which may partially account for the citation increases. One can, however, adjust for this and for different citation patterns among different fields by calculating the number of patent citations per paper. For example, Biomedical Research papers are, overall, cited more than twice as often as Clinical Medicine papers.

Considerable collaboration exists between ARS and extramural scientists. Figure 2 profiles the authorship of this patent-cited paper data set. The numbers are "whole counts." For example, a paper that has two ARS authors, one university author, and one private company author will be counted in both the "ARS+University" and "ARS+Other" categories. This makes the paper totals shown here slightly higher than the actual number. Citations to papers that have only university authors are in the "University Only" column—the same for "ARS Only." The "Other" category includes private companies, other government agencies, and miscellaneous other institutions whose total is less than 10 percent of the citations. The column on the left has citations from patents issued in 1987-88; on the right are citations from 1993-94 patents.

Figure 2. Authorship of Patent-Cited Papers

Collaboration between ARS and universities increased from 20 to 27 percent, along with marginal increases in ARS co-authorship with other types of institutions. Further evidence of increased university participation in technologically important science supported by the USDA is in the "University Only" category. As mentioned earlier, some increases in numbers may come from increased extramural funding.

One way to measure the division of responsibility for these patent-cited science papers is the "fractional count" of authors on them. A "Fractional Count" is the sum of the fraction of each institution type’s authorship of each paper. For example, if the paper cited by a patent has two ARS authors and three university authors, then the ARS is credited with 0.4 citations, and universities with 0.6. Figure 3 shows the proportion of that citation sum shared by various types of institutions. The ARS and universities share the authorship of these patent-cited papers about equally, and together claim more than 90 percent of them. Each other institution type has less than 5 percent. Universities and the ARS are about equally responsible for creating the science supported by the USDA and cited in patents.

Figure 3. Proportion of Citations by Type of Institutions

Figure 4 breaks down the fields into which these patent-cited papers fall. The journal in which a paper appears determines the field. (Noma 1986) Compared to the extramural USDA-supported papers (mostly universities), ARS authors publish more often in Biology and Chemistry journals. The extramural papers, however, are predominantly published in Biomedical Research journals, and some Clinical Medicine. The difference in emphasis is between the more applied kind of research done at ARS, especially agricultural science, and the more basic research done by the universities, especially Microbiology, Biochemistry, and Molecular Biology.

Figure 4. Breakdown of Patent-Cited Papers into Fields

The journals in which research papers appear can suggest how basic or applied the research is. CHI has given journals rankings of one (applied technology), two (engineering-technological science), three (applied research), or four (basic scientific research), going from most applied to most basic. (Narin, Pinski, and Carpenter 1976) Examples of the most basic (Level 4) journals include The Physical Review and Journal of the American Chemical Society. The most applied (Level 1) journals include such publications as Journal of the Iron & Steel Institute and Journal of the American Medical Association.

Figure 5 shows how our USDA data set is distributed among the levels. On the left are citations to papers with ARS authors. Here citations to applied research papers edge out those to basic papers by 5 percent. On the right, however, about 63 percent of citations to non-ARS papers supported by the USDA are at the more basic level, consistent with Figure 3.

Figure 5. Distribution of USDA Data Set Among the Levels

Focussing on the patents, we now want to examine the geographic origin of patents citing research literature. We use the home addresses of the inventors, which is included on USPTO patents, to infer where the inventive activity occurred. This is more precise than the address of the assignee that owns the patent, which is usually the company headquarters. Figure 6 graphs the percent of citations going from U.S.-invented patents to ARS-authored papers. The counts for inventors are fractional. Thus, a patent citation with both U.S. and foreign inventors can be split. Nonetheless, these cases are a tiny fraction of patents and papers, so the division between U.S. and foreign is clear.

Figure 6. Percent of Citations Going from U.S.-Invented Patents to ARS-Authored Papers

Of all patent citations to ARS papers, 83 percent were from U.S.-invented patents, with slight increases from 1987-88 to 1993-94. From other research, we know that about the same percentages obtain for citations to all U.S.-authored papers. Both in agriculture and overall, U.S. science is preferentially feeding U.S. technology. In addition, about 60 percent of all patent citations to foreign papers come from U.S.-invented patents issued in 1987-88. This shows that U.S.-invented patents are heavily citing foreign science, but foreign-invented patents are not heavily citing U.S. science. Therefore, U.S. technology is using foreign science far more intensively than foreign technology is using U.S. science. Moreover, U.S.-invented patents are more strongly science-linked. Although 13 percent of U.S. and 10 percent of foreign patents have at least one science reference on their front page, U.S. patents have about 0.55 science references per patent compared with 0.23 for foreign patents.

Finally, we would like some comparison between our target institution and the outside world. Due to the large amount of data that needs to be processed, we picked a single year (1988) to see how the cited institution ranks among others doing work in its fields. We went back to our SLID database and extracted all patent-cited papers published in 1988 in three fields that the ARS emphasizes in its papers: Biomedical Research, Biology, and Chemistry. Since authors from multiple institutions can collaborate on a single paper, we fractionally counted the number of papers each institution has in a particular field. We ranked the institutions in order of number of fractional papers cited by all patents issued through 1995.

Table 1. Results for the Field of Biology

Our analysis concluded the following:

1.The number of USDA-supported papers cited by U.S. patents, and the consequent total citations, tripled between 1987-88 and 1993-94. Over and above mere increased publication rates, this is probably an indicator of increased reliance on basic research by all of agricultural technology, and also reflects a shift in the technology to areas that are traditionally more science-linked, such as genetic engineering.

2.Collaboration between the ARS and other institutions, especially universities, increased by one-third. Still, authorship of USDA-supported papers is evenly split between the ARS and universities. Using this data, one can track the demographics of the major contributors to a patented technology. The technological impact of collaborations could also be investigated this way.

3.Cited ARS papers are predominantly applied science, especially agricultural, whereas cited USDA-supported university research is more basic and biomedical. This no doubt originates in the agency’s funding decisions. Therefore, the consequences of these decisions can be followed through to the published science, and on to the patented technology that cites it.

4.All USDA-supported papers are preferentially cited by U.S. inventors. The U.S. patent system is roughly half foreign and half U.S., and yet more than 80 percent of the citations to these papers are from U.S. inventors. The USDA, like other agencies we have studied, seems to contribute to U.S.-based technology preferentially. This patent set, however, was extracted using USDA-supported papers as a source. Whether the U.S. has a disproportionate share of all agricultural patents has yet to be determined.

Even most of the foreign research cited by patents is cited by U.S.-invented patents. This goes on although patents in the U.S. system are almost evenly split between those held by foreign and U.S. entities. All of this suggests that a direct wholesale plundering of American science for foreign benefit is not happening. Quite the contrary, U.S. patents are heavily using both U.S. and foreign science.

5.For a single year (1988) we compared the ARS with other research institutions in scientific fields in which the ARS is heavily involved. ARS has the most Biology papers published that year and cited by patents, more than three times as many as the next institution on the list. Their papers have a high rate of patent citation—70 percent more often than average. This reflects practical biological research of great interest among technology developers. It may be an indicator of value in GPRA (Government Performance & Results Act of 1993) evaluations. This act charges federal agencies with establishing goals and measuring performance results. The contribution by science supported by public money to the nation’s technology is surely one legitimate measure of performance. The ARS is also among the top sixteen institutions in its less-emphasized fields of Chemistry and Biomedical Research.

Science references contained in U.S. patents are independent indicators that link science and technology. Despite the limitations, they effectively trace the scientific underpinnings of technical innovations—at least the science and technology that have been publicly disclosed. These indicators can be further traced back to the agencies that fund the science. The connection between money appropriated for research (much of it public) and resulting technology is now apparent.

This case study can be replicated for any one of 3500 institutions in the corporate address database, including combined groups of agencies, such as the Department of Defense. One can also approach the study from patents to science, by singling out the patent portfolios of individuals, companies, technologies, or nations, and tracing their scientific ancestry. The potential of this linkage database is far from fully realized.

The authors thank the Agricultural Research Service of the U.S. Department of Agriculture, especially Dr. Richard Parry, for this paper’s use of the study performed for them by CHI Research, Inc. (Grant number 59-0790-6-054)

Narin, Francis, Gabriel Pinski, and Helen H. Gee. "Structure of the Biomedical Literature." Journal of the American Society for Information Science 27(1), January-February 1976, pp. 25-45.

Narin, Francis, Mitchell Rosen, and Dominic Olivastro. "Patent Citation Analysis: New Validation Studies and Linkage Statistics." In Science Indicators: Their Use in Science Policy and Their Role in Science Studies, ed. A. F. J. van Raan, A. J. Nederhoff, and H. F. Moed. The Netherlands: DSWO Press, November 14-16, 1988, pp. 37-47.

Narin, Francis, and Dominic Olivastro. "Status Report - Linkage Between Technology and Science." Research Policy 21(3), June 1992, pp. 237-249.

Narin, Francis, Kimberly Hamilton, and Dominic Olivastro. "Linkage Between Agency-Supported Research and Patented Industrial Technology." Research Evaluation, 5(3), December 1995, pp. 183-187.

Narin, Francis, Kimberly Hamilton, and Dominic Olivastro. "Linkage Between U.S. Patents and Public Science." Science, Manuscript submitted for publication, 1997.

Noma, Elliot. "Subject Classification and Influence Weights for 3,000 Journals." Report to National Institutes of Health and Advisory Board for the Research Councils (England). (Available from CHI Research), May 1986.