The products of academic research, as noted elsewhere in this chapter, include trained personnel and advances in knowledge. The former have been discussed in chapter 3 of this volume and in the
preceding sections of this chapter. This section deals with indicators of advances in knowledgespecifically:
This discussion of article outputs places the United States in the context of other countries contributing to the world scientific literature and examines that literature by field. For a description of the data used in this analysis and its limitations, see "Data Sources for Article Outputs." [Skip Text Box]
The article counts discussed in this section are based on scientific and engineering articles published in a stable set of about 4,800 journals selected by the Institute of Scientific Information (ISI) as the base for its Science Citation Index in 1981. Fields covered are clinical medicine, biomedical research, biology, physics, chemistry, earth and space sciences, mathematics, and engineering and technology. Appendix table 5-45 lists the constituent fine fields. A database covering the social sciences and behavioral aspects of psychology is being prepared for inclusion in future Indicators volumes. The database excludes letters to the editor, news pieces, editorials, and other content whose central purpose is not the presentation or discussion of scientific data, theory, methods, apparatus, or experiments.
ISI periodically updates its journal coverage, based in part on references to articles in publications not currently included in the database. Given this citations-based updating, ISI's database appears to give reasonably good coverage of a core set of scientific journals (albeit with some English-language bias), but not necessarily of all that may be of regional or local importance. This last point may be particularly salient for the engineering and technology category and for nations with a small or applied science base. In this discussion, long-term publishing trends including coauthorship patterns are based on a journal set established by ISI in 1981. Citation trends are based on a 1985 journal set. Of course, new journals are always being created, and old ones cease publication. No attempt has been made here to trace the birth and death of journals and their selection for coverage by SCI over the years. All data derive from the Indicators Bibliometrics database prepared for NSF by CHI Research, Inc.
Articles are attributed to sectors and countries by the authors' institutional affiliation, which introduces certain complexities and limitations. For example, a paper is considered to be multi-authored only if two or more authors have different institutional affiliations. The same rule applies to cross-sectoral or international collaborations. For example, a paper written by a U.S. citizen temporarily residing in the United Kingdom in collaboration with someone at his or her U.S. home institution is counted as internationally coauthored, thus possibly overstating the extent of such collaborations. On the other hand, a paper coauthored by a British citizen temporarily located at a U.S. university with another member of the faculty would not be considered internationally coauthored, thus understating the count.
In the United States, increased attention has been given to cross-sectoral collaboration in scientific and engineering research. Of particular interest has been the collaboration between industry and universities to enrich the research perspectives of investigators in both settings and to create a means for more efficiently channeling research results toward practical applications. This section discusses the sectoral distribution of U.S. articles, patterns of cross-sectoral collaboration and citation, and multidisciplinary connections of these articles.
Sectoral Distribution. About 142,800 scientific and technical articles were published by U.S. authors in 1995 in the set of 4,800 journals. Of these:
These proportions represent a slightly enhanced position for academic publications since 1981 (68 percent) and an offsetting decline in the federal share including FFRDCs. (See appendix table 5-44.)
The number of academic papers increased in all fields but biology (down 25 percent since 1981) and mathematics (down 27 percent). The decrease in biology was partially offset by a strong increase in biomedical research articles, possibly reflecting a shift in focus. No ready explanation is evident for the decline in mathematics outputs. (See appendix table 5-45 for field taxonomy.)
Industry publishing has undergone considerable change over the period, reflecting both growing interest in the biomedical fields and a decline in some more traditional areas of industry activity. Industry publications almost doubled in clinical medicine and tripled in biomedical research; these two fields combined accounted for 4,700 industry articlesor 39 percent of this sector's total in 1995, versus 19 percent in 1981. Industry publications in physics, chemistry, and engineering and technologyfields traditionally emphasized in industrial researchas well as mathematics all declined in absolute numbers during the 1990s; engineering and technology suffered a particularly steep decline during the 1980s. The precise reasons for these declines are unclear, but they may in part reflect one outcome of the restructuring and refocusing of corporate R&D activities. (See appendix table 5-44.)
Article production by the Federal Government fell and was steady overall for FFRDCs. Federal research output in biomedicine and chemistry was steady. Physics and earth and space sciences articles were up; but a declining output in clinical medicine, biology, mathematics, and engineering and technology outweighed these numerical increases. In the case of FFRDCs, increased publications in physics and earth and space sciences balanced declines in other fields. Nonprofit organizations increased publication in the biomedical fields, in which they have a combined 11 percent share. (See appendix table 5-44.)
Cross-Sectoral Collaboration. Scientific and engineering research in the United States increasingly involves investigators from several employment sectors, as evidenced by the steady increases in the number and proportion of articles with authors from more than one sector. This increase is evident for all sectors and for all fieldseven those with declining outputexcept mathematics, where the modal pattern remains sole authorship.
Just under one-quarter (24 percent) of all academic papers published in 1995 involved collaboration with one or more authors from other sectors6 percent from industry, 8 percent each from the Federal Government and not-for-profit sectors, 3 percent from FFRDCs, and 2 percent from other sectors. While this proportion may appear low, it involved roughly 25,900 articles and represented an increase from 20 percent in 1981 (20,100 articles). (See appendix table 5-46.)
The propensity of scientists and engineers employed in other sectors to collaborate across sectoral boundaries was much higher than for their academic colleagues50 percent in industry, 56 percent in FFRDCs, and 60 percent and above in the other sectors. Moreover, 1981-95 increases in cross-sectoral collaboration have been more pronounced in the nonacademic sectors, ranging from 7 percentage points for nonprofit institutions to 23 percentage points for industry. But most of the cross-sectoral collaborations involved one or more academic authors.
Intersectoral Citation Patterns. Research builds upon previous results, and references to scientific and technical articles reflect their utility in subsequent work. The distribution of such citations to U.S. scientific and technical articles largelybut not entirelyreflects the distribution of the articles themselves, with the bulk of citations going to academic papers. The academic sector contributes 71 percent of all articles and receives 71 percent of all citations. Its citation frequency in clinical medicine, biomedical research, and mathematics is slightly below its publications share; in biology, chemistry, and engineering and technology, the citation frequency exceeds its publications share. (See appendix table 5-47.)
Industry articles are cited at a higher frequency than their share would suggest in the fields of physics and engineering and technology. In recent years, however, both of these fields have experienced a decline in the number of industry articles as well as a decline in the number of citations to these articles.
Linkages Among Disciplines. Research on many scientific challenges increasingly relies on the knowledge and perspectives of a multitude of disciplines and specialties. Biologists seeking to understand cell functions supplement techniques and approaches developed internally with others developed in engineering, chemistry, and physics. Citations in scientific and technical articles that cross disciplinary boundaries are one indicator of the multidisciplinary nature of the conduct of research. The citation patterns among Science Citation Index articles provide a glimpse of connections among major fields and fine fields.
Citations in 1994-95 U.S. articles contained in SCI were aggregated by field. Of the roughly 2.3 million references, articles in the three life sciences-which accounted for 63 percent of the U.S. output-contained 73 percent of the citations, those in other sciences and mathematics 25 percent, and engineering and technology articles just over 2 percent. The distribution of these citations across major fields shows the expected high incidence of references to articles in the same broad field, ranging from 69 to 83 percent in the physical and earth and space sciences to 62 percent in biology. Articles in the combined life science fields cited other life science articles 98 percent of the time. However, the citation patterns are not symmetrical. A greater proportion of citations in the physical sciences, mathematics, and engineering and technology focuses on the life science fields than vice versa. (See appendix table 5-48.)
Examination of fine fields underscores the tight connections among the life science fields. Citations in clinical medicine and biomedical research articles are largely to other articles in these same major fields90 percent or higherwith most of the remaining citations to biology. This does not mean that their research is isolated from other major fields. About 6,200 citations in clinical medicine and 20,000 in biomedical research articles were to physical sciences, engineering, or mathematics journals. But these represented tiny portions of their total citationsnumbering 920,000 and 651,000, respectively. Pharmacy and pharmacology, for example, cite articles in chemistry journals; some biomedical specialties cite chemistry, physics, earth and space sciences, and engineering and technology. The earth and space sciences' connection to biomedical research is intriguing: 4 percent of astronomy/astrophysics citations were to this literature, which in turn received more than 200 citations from general biomedical research articles. These citation links reflect, among other things, the well-publicized adaptation of astronomy imaging techniques to medical diagnosis uses. Otorhinolaryngology articles contain references to the acoustics literaturephysicsreflecting a similar connection. (See appendix table 5-48.)
The article counts reported here indicate the volume of scientific publishing in a given field and country, and the field mixes of different countries, as reflected in this set of core journals. In interpreting these counts, note that they reflect field-to-field and country-to-country variations in publishing conventions and differing sizes of scientific infrastructures. The discussion focuses on broad trends and relationships. (See "Data Sources for Article Outputs.")
Worldwide publication of scientific and technical articles in the SCI journal set stood at about 439,000 in 1995. (See appendix table 5-49 for detailed counts.) Almost one-third of these135,000were articles in clinical medicine; biomedical research and biology accounted for an additional 107,000 articles. Articles in chemistry, physics, and the earth and space sciences numbered 61,000, 74,000, and 23,000, respectively; there were 31,000 articles published in engineering and technology, and 8,000 in mathematics. (See figure 5-24.)
Five nations produced more than 60 percent of all articles in the SCI set of journals in 1995: the United States (33 percent), Japan (9 percent), the United Kingdom (8 percent), Germany (7 percent), and France (5 percent). No other country's output reached 5 percent of the covered articles' total. The regional distribution of these articles is shown in figure 5-25.
From 1981 to 1995, the number of articles published worldwide in the SCI journal set rose by almost 20 percent, compared with 8 percent in the United States alone. This increase coincided with the development or strengthening of national scientific capabilities in several world regions, a development that was particularly pronounced after the end of the Cold War. Thus, a gradual decline in the U.S. world share since the early 1980s continued through the mid-1990s, despite continued growth in U.S. publications output. (See appendix table 5-49.) The European share rose from 32 to 35 percent over the period. It is likely that these gains partially reflect European nations' concerted policies to strengthen the science base in both individual countries and across Europe as a whole.
The article volume of the Central European statesBulgaria, Czech Republic, Hungary, Poland, Romania, and Slovakiaas a group declined through the early 1990s but rebounded to close to its 1981 level by 1995 (9,100 articles). In contrast, the article output for the nations of the former Soviet Union declined at an accelerating rate after the late 1980s, dropping from about 30,000 in 1981 to 22,000 in 1995; this decrease led to a decline in world share from 8 to 5 percent. This long-term decline in world share is not entirely attributable to the disintegration of the Soviet Bloc, although this certainly continues to contribute to the trend. Articles reflect work done one or more years earlier, and the decline has been gradual and observable over the entire period. It is likely that relative political and scientific isolation, combined with economic difficulties, has affected the conduct of scientific research in this region.
Southeast Asia's emergence as a potent, high-tech region is well-known, and data on article production present evidence of a robustly developing indigenous S&E base. The Asian nations' world share of publications rose from 11 to nearly 15 percent since 1981, but contradictory trends combined to produce this total. The number of articles produced by Japan increased from 25,100 in 1981 to 39,500 in 1995; this represents a 57 percent increase, three times the world average. Very large percentage increases over the periodthough from very low baseswere evident for China (from 1,100 to 6,200 articles) and the newly industrialized Asian economies: Taiwan (from 370 to 3,900), South Korea (170 to 3,000), Singapore (120 to 900), and Hong Kong (from 500 in 1987 to 1,100 in 1995). While these gains were realized on a small output base and the combined output remains modest, the combined world share involved rose from one-half of 1 percent to 3.4 percent in a very short timewith no decrease in growth yet evident. On the other hand, India's publications output has contracted by 33 percent since 1981, dropping from 11,700 articles to 7,900 in 1995.
Since the conduct of research reflected in these article outputs requires financial, physical, and human resourcesin short, a scientific infrastructurethe potential for further shifts in article distributions can be gleaned from a brief comparison of the economic and article outputs of selected countries. While no simple relationship exists between the relative size of a nation's GDP and its article output volume, there do appear to be some general tendencies. (See appendix table 5-50.) For the nations shown, the number of papers produced per billion U.S. dollars of GDP ranges from 2 to 54. (See figure 5-26.) Israel and a number of smaller European nations rank highest, exceeding 30 articles per billion U.S. dollars of GDP. The United States is in the middle range, with 20 articles per billion dollars of GDP. Nations with fast-developing economies have smaller than expected article outputs. There is also a large number of nations with economies that are small, or small on a per capita basis, that contribute little to the world's scientific output.
As noted earlier, for all countries combined, the life sciences accounted for the bulk (55 percent in 1995) of the articles in the SCI database. (See figure 5-24.) The nearly 20 percent increase in world articles from 1981 to 1995 was driven by increases in physics (63 percent), the earth and space sciences (36 percent), and biomedical research (30 percent). Biology and mathematics publications declined in number (by 11 and 23 percent, respectively), possibly signaling the demise of some journals in these fields. Chemistry and clinical medicine articles increased slightly (by 12 and 16 percent, respectively); while those on engineering and technology did not increase at all. Because of the large number of articles produced each year, shifts in field distribution have been small but noticeable. (See text table 5-12.) For example, the life science share fell by 2 percentage points; those of mathematics and engineering and technology fell by 1 point. Within the life sciences, biology's share fell by 3 points while biomedical research articles increased, suggesting a gradual shift in research focus. The share of physics articles increased by 5 percentage points over the period.
The roughly 142,800 U.S. articles published in 1995 accounted for about 33 percent of the world's total, up in number from 132,300 in 1981 but down from the almost 36 percent share of world total these articles then represented. This drop reflects the fact that other nations' publications output has expanded relatively faster than that of the United States. U.S. output has grown in some fields: notably-in round numbers-from about 22,000 to 28,000 in biomedical research, and from 13,000 to 18,000 in physics. It has been roughly steady in clinical medicine, at about 50,000. Declines in output occurred in biology (from 15,000 to 11,000), engineering and technology (from 12,000 to 10,000), and mathematics (from 4,000 to 3,000). (See appendix table 5-49.)
But the U.S. article portfolio is quite different from that of other major producers (see "The Science and Technology Portfolios of Nations," below); consequently, U.S. world share, and changes in world share, are field dependent. The biggest relative declines occurred in engineering and technology (7 percentage points) and biology (6 points). Smaller declines in the U.S. share (2 to 4 percentage points) occurred in clinical medicine, the earth and space sciences, and mathematics. The physics share contracted by nearly 5 points, while chemistry held steady.
Nations make implicit or explicit choices about the nature of their science and technology portfolios through their allocation of resources; the results of these choices are reflected, to some degree, in their article output data. (See appendix table 5-51.) It is clear that different nations have very different choice patterns, and that these patterns can-and do-change over time.
Figure 5-27 shows the 1995 portfolio mix of a range of countries, arrayed by the fraction of their total output devoted to clinical medicine and biomedical research (which account for about half of these articles worldwide). The differences in emphasis are striking. The United States, United Kingdom, countries of Northern Europe, several smaller Western European nations, and Chile all emphasize these fields quite heavily. At the other end of the spectrum are China and the rapidly growing newly industrialized Asian economies, India, Eastern Europe, Egypt, and Mexico, each of which has a small fraction of its portfolio in these fields.
In contrast, France, Germany, Spain, Italy, Eastern European nations, Russia, Mexico, Japan, the newly industrialized Asian economies (especially), India, China, and Egypt put far more weight than the world average on chemistry and physics. Russia, China, Egypt, and-again especially-the Asian economies are noteworthy for their concurrent emphasis on engineering and technology.
Countries tend to shift the focus of their scientific activities gradually over time. (See appendix table 5-51.) Major shifts toward chemistryand, to a lesser extent, physicsare evident for some of the world's developing nations and regions. Russia, which once had an extremely heavy stake in these traditional fields, is shifting away from them. Biology research is in relative decline around much of the world, in favor of increases in the more applied life science disciplines. Engineering and technology has lost ground in many national portfolios relative to other fields. Note, however, that the portfolios of some of these countries were very small in 1981, making relatively large percentage changes possible as publication counts have grown.
In many fields, cutting-edge science is increasingly dependent on knowledge, perspectives, and techniques that cross traditional disciplinary boundaries. Often, the scope of the problem (e.g., constructing a coordinated array of widely spaced telescopes or mapping global environmental trends), combined with complexity and cost, suggests or even dictates broad collaboration that increasingly involves international partners. Both trendsincreased collaboration and growing international cooperationcan clearly be seen in the publications data. A pervasive trend toward greater scientific collaboration affects all article fields, and a steadily growing fraction of most nations' papers involves international coauthorship. This section examines these trends, the U.S. position in international collaboration, who collaborates with whom, how developing and developed nations compare, and what collaboration patterns exist for and among Asian nations.
Trends in Scientific Cooperation. A pronounced worldwide tendency exists toward greater scientific collaboration, as evidenced by patterns of corporate coauthorship of scientific and technical articles written by authors located in two or more different institutions. This phenomenon can be observed in every field, every sector, and most countries. Moreover, such collaboration is increasingly international, involving researchers from different nations. In 1995, the proportion of the world's papers that were coauthored (in this multi-institution sense) was 50 percent; almost 30 percent of these involved international collaboration. (See appendix table 5-52.) The number of coauthored articles increased from 121,000 in 1981 (33 percent of the total) to 219,400 in 1995 (50 percent). Over this period, the number of internationally coauthored articles worldwide increased by 200 percent-from 21,000 to 63,800while the total number of articles rose by about one-fifth. This increase in turn caused a rise in the proportion of all papers published worldwide involving some degree of international coauthorshipfrom 6 percent in 1981 to 15 percent in 1995.
Corporate coauthorship varies by field. For example, in the 1991-95 period, the U.S. average of coauthored articles was 56 percent, but clinical medicine was well above that with 64 percent of its articles coauthored. Chemistry, engineering and technology, biology, and mathematics had lower rates of corporate collaboration, at 39, 43, 46, and 47 percent, respectively; the other fields were close to the mean. (See appendix table 5-53.) Wider variation exists in rates of international collaboration. Measured against all coauthored articles, the U.S. average was 29 percent for 1991-95, but this was heavily influenced by a 19 percent rate of clinical medicine articles. On the other hand, 51 percent of coauthored mathematics articles involved international collaboration, as did 46 percent of physics and 42 percent of earth and space sciences articles.
The position of the United States in international collaboration (as evidenced by coauthorship) is characterized by two complementary trends. For almost every nation with strong international coauthorship ties, the number of articles involving a U.S. author rose strongly between 1981 and 1995. During this period, however, many nations broadened the reach of their international collaborations, causing a gradual diminution of the U.S. share of the world's internationally coauthored articles. (See appendix table 5-54.)
The United States has one of the highest coauthorship rates in the world: 58 percent of U.S. articles published in the ISI journal set involved corporate coauthorship in 1995, up from 43 percent in 1981. U.S. authors contributed 42 percent of all coauthored articles and participated in 45 percent of all internationally coauthored articleswell in excess of the 33 percent U.S. article share. But of all U.S. articles published in 1995, only 18 percent involved international coauthors, a smaller percentage than that of most other nations. These numbers reflect the sheer size of the domestic U.S. science base. Worldwide, domestic and international coauthorships have also risen, often more steeply (in terms of the proportion of a country's papers involved) and to higher levels than in the United States. For most countries, the share of internationally coauthored articles ranges from 25 to 40 percent of their output; but Japan and India (15 percent each), Russia (21 percent), and other former Soviet countries (13 percent) are well below this range. (See appendix table 5-52.)
Who Collaborates With Whom? International scientific collaboration, as measured by the percentage of a country's multi-author articles involving international coauthorship, centers to a considerable degree on the United States. (See figure 5-28.) In the first half of the 1990s, about one in five internationally coauthored papers published in major European industrial nations involved collaboration with the United States; for many other nations, the rate was much higher. For example, Japan and India, with low rates of international collaboration, shared 40 and 28 percent of their international coauthorships with the United States, respectively; China, 28 percent; Taiwan, 62 percent; and South Korea, 50 percent. (See appendix table 5-54.) Rates of collaboration with the United States ranged from 25 to 35 percent for major South and Central American countries, 45 percent for Israel, and near 30 percent for Australia and New Zealand. Countries of the former Soviet Union collaborated relatively less frequently with U.S. partners, as did all Central European nations except Hungary.
Examination of this same indicator for an earlier period1981-85suggests that the scientific world is witnessing the development of new centers of activity, probably reflecting continuing political and economic developments in the wake of the end of the Cold War. Comparison of 1981-85 and 1991-95 data shows strong growth in the number of articles with authors from more than one nation, andat the same timea broadening of international collaborative ties. (See appendix table 5-54.) While coauthorship with the United States continued to rise in terms of number of publications, it declined with many countries in terms of the share of all their internationally coauthored articles. (See figure 5-28.) The share drop (but not a decline in the number of articles) in collaboration with the United States was most striking for Chinaroughly 20 percentage pointsbut is evident for most other countries as well. A similar pattern, though much attenuated, is evident for the major European industrial nations.
In the Asian region, the trends are somewhat erratic, but generally indicate the development of regional cooperative patterns involvingespeciallyChina and the newly industrialized economies. Regional collaboration, as measured by the proportion of coauthored articles with an author from another Asian country, is almost 25 percent for South Korea, in excess of 30 percent for Singapore and Hong Kong, and around 15 to 20 percent for most other countries; India and Japan have lower rates of coauthorship. The degree of collaboration with Japan has increased for some but not all of these nations, and the absolute number of papers with Japanese coauthors has risen. Collaboration with the United States is high for these economies: Taiwan, 62 percent; South Korea, 50 percent; Japan, 40 percent; China and India, 28 percent each; and the other Asian nations about one-fifth. Collaboration with Europe is less prominent, ranging from 10 to 25 percent for the entire continent.
The Central European states have fairly strong regional collaborative ties, given the relatively small volume of their collective publications output. They share 10 to 15 percent of their internationally coauthored articles. From roughly half to 60 percent of these articles are shared with the rest of Europemost strongly with Germany (around 20 percent); and the United Kingdom, France, and Italy combined (15 to 20 percent). These figures have increased over their levels in the 1980s, as ties to the countries of the former USSR have generally attenuated in the 1990s. International collaboration involves U.S. scientists in about 10 percent of the cases in Czech Republic, Slovakia, and Bulgaria; in excess of 15 percent for Poland; and over 20 percent for Hungary.
Russia's collaborative ties are mainly with the United States (roughly 15 percent); Germany (15 percent); and the United Kingdom, France, and Italy combined (20 percent). The rest of Europe represents 20 percent; collaboration with other former member states represents 10 percent. As a group, the countries of the former Soviet Union (except the Baltic states) have much the same pattern, though with weaker cooperative links to the United States and Germany, and stronger links to other European nations. Scientists in the Baltic states who collaborate internationally tend to do so with colleagues in the Scandinavian countries (25 percent), attesting to strong cultural and regional ties among these nations.
The U.S. pattern of international coauthorship stands in sharp contrast to those just described (as it must, given the high percentages of U.S. involvement in most other nations' international collaborations). No one country contributes more than 10 percent to the U.S. articles with multinational authors. Canada, the United Kingdom, Germany, all of Southern Europe, the Northern European countries, and all other Western European nations combined contribute between 7 and 10 percent each; the Eastern European and former Soviet states combined contribute another 7 percent; Japan and the other combined Asian nations contribute about 8 percent each. This is a much more even distribution of international collaborative ties than is seen for the other countries.
Countries with small indigenous science establishments tend to have higher levels of international coauthorship as a percentage of their total article output than do those with larger, more mature systems. Rather than collaborating regionally, scientists from developing nations tend to work with those from major science-producing nations. In the case of small but mature nations (e.g., the Northern or smaller Western European countries), this pattern is augmented by regional collaboration. Political isolation, economic disruption (as in the case of the states of the former Soviet Union), and cultural or language barriers (as in the case of Japan) can influence these patterns and result in unusually low degrees of international collaboration.
The global dimensions of the conduct of scientific activity, discussed above in terms of international research collaboration, are also reflected in the patterns of citations to the literature. Scientists and engineers around the world cite prior work done elsewhere to a considerable extent, thus demonstrating the usefulness of this output in their own work. Citations to one's own country's work are generally prominent and show less of a time lag than citations to foreign outputs. Regional citation patterns are evident as well, but citations to research outputs from around the world are extensive.
U.S. scientific and technical articles are cited by virtually all mature scientific nations in excess of the U.S. output's world share. (See appendix table 5-55.) This broad finding needs to be qualified, however, since citation patterns and practices vary by field. More specifically, the finding holds for chemistry, physics, biomedical research, and clinical medicine. U.S. articles in the remaining fields tend to be cited at or slightly below their world output share.
Not surprisingly, all countries cite their domestic literature well in excess of their respective world shares, but no other country cites its domestic literature as heavily as does the United States67 percent. Another 14 percent of U.S. citations are to British, French, German, and Italian articles; 7 percent each to the articles of other European nations and Asia and the Pacific (4 percent for Japan); and 3 percent to Canadian articles. The high U.S. self-citation rate might conceivably reflect insularity, but the high proportion of involvement of U.S. scientists in internationally coauthored articles casts doubt on this interpretation.
A comparison of citations to the U.S. literature (the leftmost column of appendix table 5-55) with those to a nation's domestic output (diagonal values) shows a generally larger share of total citations to U.S. than to domestic articles. (See figure 5-29.) In part, of course, this reflects the scale and breadth of the U.S. scientific and technical establishment. Yet there is no compelling reason why one country's literature should be cited in proportion to its world output share by any other country. For example, no European country cites another European country's output at the rate of the cited country's article share, despite the many arrangements that foster collaboration and knowledge flows among the European nations. It appears reasonable to conclude that scientists elsewhere find the outputs from U.S. research quite useful in the conduct of their own work, as evidenced by the volume of references to the U.S. literature in other countries' scientific and technical articles.
The citations in articles from the four largest European industrial nationsthe United Kingdom, France, Germany, and Italyrefer to their respective domestic outputs 21 to 30 percent of the time, to articles of the other countries in the set 11 to 18 percent of the time, and to U.S. articles between 36 and 38 percent of the time. Output from the rest of Europe receives 10 to 12 percent of citations; Canada, 3 percent; and Asia and the Pacific, 7 percent (4 to 5 percent to Japanese articles).
The citations from other Western, Southern, and Northern European nations refer to their own domestic literature 10 to 23 percent of the timereflecting their generally smaller domestic science baseand the four large European industrial nations 18 to 28 percent. The United States receives 32 to 42 percent of the citations; and other European nations combined, 10 to 17 percent. Asia and the Pacific receive 7 to 9 percent of these nations' citations.
The pattern of citations among Central European nations is similar, with a regional component of 3 percent, and an additional 1 to 3 percent referring to the literature of the former Soviet states. A stronger orientation than for most other countries is evident toward Asia and the Pacific, which receive a combined 9 to 11 percent of the citations.
Somewhat less reliance on European science output, somewhat greater reliance on that of the United States, and more of a regional Asian/Pacific focus mark the citation ties of the Asian nations. China and the newly industrialized economies cite their own articles only 10 to 20 percent of the time, but cite each others' articles 12 to 16 percent of the time-high relative to the size of their science bases. Japan's pattern is different (37 percent self-citation and only 2 percent of citations to articles from other states in the region); as is India's (29 percent self-citation, 6 percent citation to Japan's output, and 2 percent to the rest of the region).
Governments assign property rights to inventors in the form of patents to foster inventive activity that may have important economic benefits. The U.S. Patent and Trademark Office grants such government-sanctioned property rights in the form of patents for inventions deemed to be new, useful, and non-obvious. This section discusses recent evidence about strengthening ties between scientific and technical research and patenting activity, trends in academic patenting, and income from these activities flowing to universities and colleges.
Patent applications cite "prior art," including scientific and technical articles, that contributes materially to the product or process to be patented and upon which it improves. These citations provide some indication of the potential contributions of published research results to patentable U.S. inventions. A number of caveats apply. The use of patenting varies by industry segment, and many citations on patent applications are to prior patents. Industrial patenting is only one way of seeking to ensure firms' ability to appropriate returns to innovation and thus reflects, in part, strategic and tactical decisions. Such patenting can be defensive or forward-looking, or can lay the groundwork for cross-licensing arrangements. Most patents do not cover specific marketable products but might conceivably contribute in some fashion to one or more such products in the future. These caveats notwithstanding, citations to the scientific and technical literature give one indication of the linkage between research outputs and innovative applications, as judged by the patent applicant.
The scientific and technical literature is increasingly likely to be cited on U.S. patents. The percentage of U.S. patents citing at least one scientific or technical article increased from 11 percent in 1985 to 14 percent in 1990 and 23 percent in 1995. To further explore this trend, citations to U.S. research articles included in the SCI set of journals were identified and classified by field and performer sector for all U.S. patents issued from 1987 through 1996. The number of such citations increased from 8,600 in 1987 to 47,000 in 1996 (see figure 5-30 and text table 5-13), and their field distribution shifted dramatically toward the life sciences. The rise in the number of citations held for all fields and for papers from all sectors. (See appendix table 5-56.) The fastest growth, however, occurred in the life sciences. The biomedical research share rose from 28 to 44 percent, and that of clinical medicine rose from 26 to 29 percent. The combined share of physics, chemistry, and engineering and technology citations dropped from 43 to 24 percent of these patent citations-but not their absolute number, which rose from 4,018 in 1988 to 11,246 in 1995.
Citations to academic articles rose faster than to those from industry or government authors, pushing the academic share of the total from 49 to 55 percent. The increase was driven by strong gains in chemistry (where the academic share rose from 58 percent in 1988 to 65 percent in 1995), physics (from 29 to 40 percent), and engineering and technology (from 31 to 44 percent).
A recent study examined all citations on the front page of all 397,660 U.S. patents awarded in 1987-88 and 1993-94 (Narin, Hamilton, and Olivastro 1997). Many of these citations are to other patents, but among all citations, 430,226 referred to nonpatent materials; of these, 242,000 were judged to be science references, of which 175,000 were to materials in SCI journals. Among the study's findings are a rapid increase in the number of citations to scientific and technical articles on U.S. patent applications; a shortening of the time elapsed between publication and citation on patents; and a large proportion of such citations to publicly funded science (defined by the authors to include articles by academic, nonprofit, and government authors). References tended to be to articles appearing in nationally and internationally recognized, peer-reviewed journals, including journals publishing basic research results, and to be field- and technology-specific. The authors note both national (U.S. patents citing U.S. authors with greater than expected frequency) and regional components in the patterns of citations.
Patents may be awarded on the results of academic R&D deemed to have potential utility for the development of new or improved products or processes. While the bulk of academic R&D is basic research (i.e., research that is not undertaken to yield or contribute to immediate practical applications), data on the patenting activities of universities and colleges suggest that academic institutions are giving increased attention to the potential economic benefits that may be inherent in their R&D results. A growing number of universities and colleges are applying for, and receiving, protection for results of work conducted under their auspices, even though the returns on such patents remain low, on average, when measured against their R&D expenditures. (See "Income From Patenting and Licensing Arrangements," below.) The number of patents and institutions involved is small when viewed against the backdrop of all U.S. patenting activity, but the increases are of interest.
After slow growth in the 1970s, the number of academic institutions receiving patents increased rapidly in the 1980s from about 73 early in the decade to more than double that by 1989 and 168 by the mid-1990s. This development, pronounced during the 1980s and more muted in this decade, affected the number of both public and private institutions receiving patent awards. (See figure 5-31.) Starting in the early 1980s, the number of institutions outside the ranks of the largest research universities (defined here as the top 100 in total 1995 R&D expenditures) with patent awards increased at a rapid pace. While the largest research universities had constituted 70 percent of all academic institutions receiving patents in 1982, their share of all academic institutions had fallen to just half in 1995-signaling a broadening of the institutional base, especially among public universities and colleges. (See appendix table 5-57.) Nevertheless, by 1995, 86 of the top 100 universities in total R&D expenditures received one or more patents.
This expansion of the number of institutions receiving patents coincided with rapid growth in the number of patent awards; this latter rose from 458 in 1982 to 1,860 in 1995. Public institutions expanded their patenting activity somewhat more rapidly than did their private counterparts: the former received 64 percent of newly issued academic patents in 1995, up from 53 percent in 1982. At the same time, the top 100 R&D-performing institutions increased their share of the expanding academic patent base from about three-quarters to over 80 percent of the total, where it has leveled off. (See appendix table 5-57.)
The number of academic patents rose more than sevenfold in just over two decades, from about 250 annually in the early 1970s to more than 1,800 in 1995. (See figure 5-32.) This is a far more rapid increase than for all annual U.S. patent awards, which roughly doubled over the period. As a result, academic patents now constitute about 3 percent of all new awards, up from less than one-half of 1 percent two decades ago. A change in U.S. patent law may have contributed to the strong rise in the 1980s; the law now allows academic institutions and small businesses to retain title to inventions resulting from federally supported R&D. Other contributing factors may be the creation of specialized university technology transfer and patenting units, an increased focus on commercially relevant technologies, and closer ties between scientific and engineering research and technological development (see Henderson, Jaffe, and Trajtenberg 1995).
Patents are assigned to utility classes according to their likely areas of application. The distribution of all patents over these areas has evolved slowly, but for academic patents, two pronounced changes have taken place. The growth in the number of academic patents was accompanied by a decrease in the number of utility classes in which they fall. In addition, academic patents are more heavily concentrated in relatively few application areas than are all U.S. patents. This is not surprising, since many patents in many application areas are not science-based at all. Nevertheless, the concentration is remarkable. Over the entire period covered by the database, 1969-95, utility classes in which universities were at least twice as likely as others to be awarded patents accounted for 12 percent of all patents, but half of all academic patents. (See appendix table 5-58.) A heavy concentration is evident in areas connected with the life sciences, along with some areas of physics and chemistry. (See appendix table 5-59.) In fact, the fraction of academic patents in just three utility classesall with presumed biomedical relevancejumped from 8 percent of the total in the early 1970s to more than a quarter in the mid-1990s. (See figure 5-33.)
Valuation of patentsespecially of science-based onesis difficult. Actual use is uncertain, there is generally no direct connection between an individual patent and an economically valuable product or process, and acquisition of licensing rights may be motivated by protection rather than by intent to use. Nevertheless, universities increasingly are negotiating royalty and licensing arrangements based on their patents. While total reported revenue flows from such licensing arrangements remain low, a strong upward trend points to the confluence of two developments: a growing eagerness of universities to exploit the economic potential of research activities conducted under their auspices, and readiness of entrepreneurs and companies to recognize and invest in the market potential of this research.
A 1992 survey by the U.S. General Accounting Office based on 35 universities found that they had substantially expanded their technology transfer programs during the 1980s. Typical licensees were small U.S. pharmaceutical, biotechnology, and medical
businesses. During 1989-90, the reported income flows based on these licenses were modest: a mere $82 million. A more extensive survey conducted periodically since 1991 (AUTM 1996) reported gross revenue receipts of $299
million in 1995, compared with $130 million in 1991. (See text table 5-14.) The surveywhile extensiveis not nationally representative; thus, these estimates must be seen as lower bound numbers. Moreover,
a portion of these reported revenue increases reflects expanded coverage.