Outputs of S&E Research: Articles and Patents
- S&E Article Output
- Coauthorship and Collaboration
- Trends in Output and Collaboration Among U.S. Sectors
- Trends in Citation of S&E Articles
- Academic Patents, Licenses, Royalties, and Startups
Chapter 2 of this volume and the previous section of this chapter discuss the outputs of S&E research and education in terms of human capital. This section examines additional indicators of the output of academic S&E research: articles published in the world’s S&E literature and patents received by U.S. academic institutions. In addition, licensing activities, royalties, and startups associated with university research are also discussed.
Published, peer-reviewed articles have traditionally been the means by which scientists and engineers report the results of their research and gain status in their fields. According to sociologist Robert K. Merton,
The institutional conception of science as part of the public domain is linked with the imperative for communication of findings. Secrecy is the antithesis of this norm; full and open communication its enactment. The pressure for diffusion of results is reinforced by the institutional goal of advancing the boundaries of knowledge and by the incentive of recognition which is, of course, contingent upon publication. (Merton, 1973, p. 274; see also de Solla Price 1978)
This section uses data on S&E articles to indicate world S&E knowledge production by country and by selected regions and/or groupings of countries related by geography, cultural ties, language, or political factors. Coauthorship of articles by researchers in different departments, different institutions, and different countries and regions illustrates the increasing trend of collaboration in research, both within and across countries and regions.
Citation of research articles indicates, albeit imperfectly, the relative importance of previously published research findings to future research; consequently, patterns in citation are also discussed in this section. Citation patterns, including trends in highly cited research articles, are contrasted with trends in total publication of articles.
The discussion of research outputs concludes with indicators of the flow of knowledge from academically based research to intellectual capital embodied in patents awarded to academic institutions, along with related other indicators.
S&E Article Output
The number of S&E articles in the dataset analyzed in this chapter totaled 10.6 million for the period 1988–2005. In the past 10 years, the total world S&E article output as contained in the Science Citation Index (SCI) and Social Sciences Citation Index (SSCI) (see sidebar, "Bibliometric Data and Terminology") grew at an average annual rate of 2.3%
Trends in Country and Regional Authorship
Although S&E authors from some 200 countries are represented among the articles discussed in this section, these authors are concentrated in a relatively small number of countries (see sidebar, "Distribution of Publication Data"). Authors from one country, the United States, dominated global article output in 2005 with 29% of the total, followed by Japan with 8% and the United Kingdom, Germany, and China with 6% each.
Previous editions of Indicators and other studies (e.g. NSF/SRS 2007a) reported steadily increasing investments in S&E education and research infrastructure, especially in Asia. As these investments matured and led to increased R&D in those countries, authorship by scientists and engineers in those countries also increased, as did their success in getting articles published in international peer-reviewed journals. Differences in recent rates of growth in article production are striking. Among Asian countries/economies that produce a major number of articles (defined here as more than 10,000 articles in 2005), average annual growth rates between 1995 and 2005 were highest in China, at 17%, and South Korea, at 16%
The 10-year change rate shown in
Even among nations with moderate S&E article production (defined as between 1,000 and 10,000 articles in 2005), a few stand out for increasing their publication over the past decade. In the Middle East, Iran’s article output grew at 25% a year, although its output was less than 3,000 in 2005
Trends in Country Rank by S&E Field
In a comparison of the top producers of S&E articles in 1995 and 2005, two patterns are evident: (1) U.S. scientists and engineers authored more S&E articles across all fields than authors in any other single country in both 1995 and 2005, and (2) overall, the top 20 article-producing countries were similar in both years
- China’s high rates of annual growth in S&E article production resulted in its movement from 14th to 5th place in overall S&E article authorship, to 2nd place in engineering and chemistry, and to 3rd place in physics and mathematics. China moved up in rank of authorships in other fields as well.
- South Korea improved its overall rank from 22nd in 1995 to 10th in 2005, with its highest rank (4th) in engineering. It made gains in other fields as well.
- Taiwan moved up in rank overall and in all fields shown except mathematics.
- India failed to demonstrate the fast growth of other Asia-10 countries and lost rank in some fields.
- Brazil and Turkey gained rank across all fields shown.
- Russia, whose growth rate was negative over the period, lost rank across all fields.
Coauthorship and Collaboration
In addition to the increasing volume of the world’s S&E published literature discussed in the previous section, another trend was an increase in the number of S&E articles with authors from different institutions. A related and even stronger trend, increases in the number of internationally coauthored S&E articles, was widely noted in previous editions of Indicators. The following discussion begins with consideration of broad trends for the world as a whole, moves to regional patterns, and ends with a discussion of country-level trends, including selected country-to-country coauthorship patterns and indexes of international collaboration. (Indicators of cross-sector coauthorship, available only for the United States, are examined below in the section "Trends in Output and Collaboration Among U.S. Sectors.")
Indicators of world S&E article output discussed in the previous section show a growing world article output, with just a few dozen countries producing the predominant proportion of all articles. Within that trend lie three additional patterns of interest: a growing tendency for articles to list multiple authors, authors from more than one institution, and authors from more than one country.
Previous editions of Indicators used coauthorship data as an indicator of collaboration among scientists and discussed possible underlying drivers for increased collaboration, including scientific advantages of knowledge and instrument sharing, decreasing costs of travel and communication, national policies, and so forth (NSB 2006). Katz and Martin (1997) and Bordons and Gómez (2000) analyze limitations of coauthorship as an indicator of research collaboration, but other researchers have continued to conduct studies of S&E research collaboration using such data (Adams et al. 2005; Gómez, Fernández, and Sebastián 1999; Lundberg et al. 2006; Wuchty, Jones, and Uzzi 2007; Zitt, Bassecoulard, and Okubo 2000). The coauthorship data used in this section as indicators of collaboration in S&E research are presented with knowledge of neither the motive(s) underlying the collaboration nor the nature of the collaboration that actually occurred. They should be seen as broad indicators of a secular trend in the S&E publishing record that reflects changes in the way S&E research is conducted and reported in today’s world.
Article Author Names and Institutions
Indicators of the extent of these changes are shown in
A slightly different indicator, coauthored articles, has also increased steadily. Coauthored articles are defined as S&E articles with more than one institutional address in the byline. ("Institution" here may refer to different departments or units within the same institution; multiple listings of the same department or unit are counted as one institutional author.) Adams and colleagues (2005) offer several hypotheses that might explain growing collaboration, including specialization by researchers and a consequent increase in division of labor; decreases over time in the cost of collaboration (and of international collaboration) due to the Internet; and increases in the sharing of large research resources like instruments and large datasets. They also argue that increases in the division of labor of scientists on a team lead to increases in scientific productivity. On the other hand, Cummings and Kiesler (2005, 2007) report high coordination costs in studies of two large U.S. government programs that sought to foster collaboration.
Coauthored articles grew from 40% of the world’s S&E articles in 1988 to 61% in 2005
Coauthorship From a Regional Perspective
Use of the same region/country categories as in "S&E Article Output" above shows changes in the patterns of interregional coauthorship. Over the period 1995–2005, interregional coauthorship increased as a percentage of total article output for the United States (from 17% to 27%), the European Union (from 18% to 26%), and the Asia-10 (from 16% to 19%)
Coauthorship Patterns From an International Perspective
When the region-level data discussed in the previous section are disaggregated to the country level, a richer picture of international S&E article coauthorship emerges.
Narin and colleagues (1991) concluded that "the direction of international coauthorship is heavily dependent on linguistic and historical factors." Coauthorship data suggest intriguing "preferences" at the national level (Glänzel and Schubert 2005; Schubert and Glänzel 2006) based on the geography, cultural relations, and language of particular pairs or sets of countries, and these preferences have been evolving over time (Glänzel 2001). Some researchers have focused on the growing S&E article output and international coauthorship of particular countries mentioned in the previous section, for example, Korea (Kim 2005), China (Zhou and Leydesdorff 2006), and Turkey (Uzun 2006).
International Coauthorship With the United States
When authors of S&E articles from U.S. institutions collaborate with authors from abroad, in which countries are these authors likely to be located?
Readers may note the asymmetry between the columns of data in
International Collaboration in S&E
In developing indicators of international collaboration between countries and across regions, researchers have developed statistical techniques that account for unequal sizes in countries’ S&E article output and coauthorship patterns (Glänzel and Schubert 2004). One of the simplest of these techniques is used in calculating the index of international collaboration shown in
None of the collaboration indexes between countries on opposite sides of the North Atlantic was as high as expected based on their total international collaboration. In Europe, collaboration patterns were mixed. Among the large publishing countries of Germany, the United Kingdom, and France, collaboration was less than expected. The indexes for France-Spain and Italy-Switzerland were somewhat higher than expected, and very strong rates of collaboration were evident throughout Scandinavia.
Cross-Pacific collaboration was rather weak between the United States and both China and Japan, but somewhat stronger than expected between the United States and both South Korea and Taiwan. Canada showed a lower tendency than the United States to coauthor with other Pacific Rim countries.
Collaboration indexes between the large article producers within the Asia-10 were generally higher than expected. Indexes for Japan-China and for Japan-South Korea were strong. Australia’s collaboration with Singapore (1.72) and New Zealand (4.23) was particularly strong. India collaborated more than would be expected with Japan (1.31) and South Korea (1.84).
Trends in Output and Collaboration Among U.S. Sectors
S&E articles authored at academic institutions have traditionally accounted for just under three-fourths of all U.S. articles
Article Output by Sector
Total annual publications by authors in U.S. nonacademic sectors changed little over the past decade
Trends in Sector Coauthorship
The previous section on "Coauthorship and Collaboration" presented coauthorship data as an indicator of collaboration between and among U.S. and foreign scientists and engineers. This section considers coauthorship data as an indicator of collaboration at the sectoral level between U.S. institutional authors and between U.S. sectors and foreign institutions. These data show that the growing integration of R&D activities, as measured by coauthorship, is occurring across the full range of R&D-performing institutions.
Between 1995 and 2005, coauthorship increased in all U.S. sectors and, most notably, between U.S. institutional authors in all sectors and non-U.S. authors. Authors in FFRDCs, industry, and private nonprofit institutions increased their coauthorship with foreign authors by 10 percentage points between 1995 and 2005
The extent of coauthorship between U.S. sectors and authors from another country varied by broad field of science. Astronomy had the highest rate of international coauthorship in 2005, at 58%, well above the U.S. national average of 27% across all fields and all sectors
U.S. cross-sectoral coauthorship increased between all sectors except FFRDCs and industry. The largest gains in all sectors were with coauthors in academia (by far the largest sector with the largest pool of potential S&E coauthors). State/local government, the sector with the highest percentage of articles with coauthors from academia in 1995, at 63%, also had the highest percentage in 2005, at 71%, followed by private nonprofit institutions at 62% and the federal government at 59%
Within-sector coauthorship (e.g., FFRDC authors with authors from other FFRDCs) increased as well. Starting from the highest base of within-sector coauthorship in 1995, at 36%, academic authors increased their coauthorship with authors from other academic institutions to 43% in 2005. FFRDC-FFRDC coauthorship, and private nonprofit/private nonprofit coauthorship both increased by more than 4 percentage points over the decade.
Except for the decline in coauthorship between FFRDCs and industry, the indicators presented in this section show steadily increasing integration between and among the different types of U.S. institutions that publish the results of R&D in the scientific and technical literature. The data in
Trends in Citation of S&E Articles
When scientists and engineers cite the published results of previous research, they are formally crediting the influence of that research on their own work. Previous editions of Indicators presented data on the growing number of worldwide citations to foreign S&E literature. Like the indicators of international coauthorship discussed above, cross-national citations are evidence that S&E research is increasingly international in scope.
The indicators discussed here present a coherent picture of a world S&E literature dominated by the United States. At the same time, a decade of increases in the publication of research articles by a few dozen countries in Asia and Europe has chipped away at the U.S. share on a number of publication indicators. The following sections continue to explore this theme by contrasting worldwide research output trends with worldwide trends in highly cited S&E literature by field.
Citation Trends in a Global Context
Much of the world’s S&E research literature is never cited in another article, although citation rates vary by field
Trends in Highly Cited S&E Literature
Another indicator of performance of a national or regional S&E system is the share of its articles that are highly cited. High citation rates can indicate that an article has a greater impact on subsequent research than articles with lower citation rates.
Citation percentiles for 1995, 2000, and 2005 are shown by field and region/country in
This is the case in every field for U.S. articles. Across the 11 years displayed in
These data are summarized in
However, when citation rates are normalized by the share of articles during the citation period to produce an index of highly cited articles, the influence of U.S. articles is shown to increase. Between 1995 and 2005, the U.S. index of highly cited articles increased from 1.73 to 1.83
The United States experienced notable gains on the index of highly cited articles in engineering, mathematics, and computer sciences (although with relatively low counts in the latter) and declines in chemistry and geosciences
Academic Patents, Licenses, Royalties, and Startups
Other indicators of academic R&D outputs reflect universities’ efforts to capitalize on their intellectual property in the form of patents and associated activities. Although some U.S. universities were granted patents much earlier, the majority did not become actively involved in the management of their own intellectual property until late in the 20th century. The Bayh-Dole Act of 1980 gave colleges and universities ownership of income streams from patented discoveries that resulted from their federally funded research. To facilitate the conversion of new knowledge produced in their laboratories to patent-protected public knowledge that can be potentially licensed by others or form the basis for a startup firm, more and more research institutions established technology management/transfer offices.
Efforts to encourage links between university-based research and commercial exploitation of the results of that research have been widely studied by researchers. Mowery (2002) notes the strong growth in funding by NIH and the predominance of biomedical-related patenting by universities in the 1990s. Branstetter and Ogura (2005) identify a "bio-nexus" in patent-to-paper citations, and Owen-Smith and Powell (2003) explore the effects of an academic medical center as part of the "scientific capacity" of a research university. In a qualitative study of two research universities that would appear to have similar capacities, Owen-Smith and Powell (2001) examine the very different rates of invention disclosure of the two campuses. Stephan and colleagues (2007) found strong differences in patenting activity among university scientists by field of science; a strong relationship between publication activity and patenting by individual researchers; and patenting among university researchers restricted to a small set of the potential population.
The following sections discuss overall trends in university patenting through 2005 and related indicators.
University Patenting Trends
U.S. Patent and Trademark Office (USPTO) data show that patent grants to universities and colleges increased sharply from 1995 to about 2002, when they peaked at just under 3,300 patents per year, and then fell to about 2,700 in 2005
The previous edition of Indicators noted that three biomedically related utility classes dominated university patenting in the 1980s and 1990s (NSB 2006, pp. 5-54 and 5-55). In 2005, these same three classes together accounted for more than one-third of all utility patents awarded to U.S. academic institutions: drug, bio-affecting and body treating compositions (15.4%); chemistry: molecular biology and microbiology (13.8%); and organic compounds (5.6%)
Patent-Related Activities and Income
In contrast to the USPTO-reported decline in the total number of patents awarded to U.S. universities and colleges in 2004 and 2005
Most royalties from licensing agreements accrue to relatively few patents and relatively few of the universities that hold them, and many of the AUTM respondent offices report negative income. (Thursby and colleagues  note that the objectives of university technology management offices include more than royalty income.) At the same time, one-time payments to one university can complicate analysis of the overall trend in university income due to patenting. The median net royalty per university respondent to the AUTM surveys has both risen and fallen since 1996 but overall climbed from $440,000 in 1996 to $950,000 in 2005
During the same period, the inventory of revenue-generating licenses and options across all AUTM respondent institutions increased, from 5,000 in 1996 to more than 10,200 in 2005
 The data in this edition of Indicators do not include articles from journals in professional fields. Thus the article counts reported here for past years will be slightly lower than counts reported in previous editions. See sidebar, "Bibliometric Data and Terminology."
 European Union (EU) data include all member states as of 2007 (see appendix table 5-33 for a list of member countries); previous editions of Indicators considered a smaller set. Thus the larger world share of S&E articles accounted for by the European Union is in no small part a result of the expanded EU membership. However, see the discussion of growth rates by region and country later in this section.
 The Asia-10 includes China (including Hong Kong), Japan, India, Indonesia, Malaysia, Philippines, Singapore, South Korea, Thailand, and Taiwan.
 Uzun (2006) describes 20 years of Turkish science and technology policies that underlie the expansion of its article output.
 Another use of these data, showing within-country/within-region S&E article field distributions as an indicator of the region/country portfolio of S&E research, has been discussed in past editions of Indicators. Although countries and regions display somewhat different emphases in their research portfolios, these patterns are stable and change only slowly over time. See, for example, Science and Engineering Indicators 2006, figure 5-38 and appendix tables 5-44 and 5-45 (NSB 2006).
 The reader is reminded that the data on which these indicators are based give the nationality of the institutional addresses listed on the article. Authors are not associated with a particular institution and may be of any nationality. Therefore the discussion in this section is based on the nationality of the institutions, not authors themselves and, for practical purposes, makes no distinction between nationality of institutions and nationality of authors.
 Merton (1973, p. 409) points out the tension between the norms of priority and of allocating credit in science: "Although the facts are far from conclusive, this continuing change in the social structure of research, as registered by publications, seems to make for a greater concern among scientists with the question of 'how will my contribution be identified' in collaborative work than with the historically dominant pattern of wanting to ensure their priority over others in the field…It may be that institutionally induced concern with priority is becoming overshadowed by the structurally induced concern with the allocation of credit among collaborators."
 In this section only, author names refer to counts of individually listed authors of articles, not institutional authors. Since authors may appear on more than one article per year, they may be counted more than once. However, because NSF does not analyze individual author names, the extent of such multiple counting is unknown.
 The coauthorship data discussed in this paragraph are restricted to coauthorship across the regions/countries identified in table 5-23; i.e., collaboration between or among countries of the European Union, for example, is ignored. Intraregional coauthorship is discussed in the following sections.
 Readers are reminded that each country participating in an international coauthorship receives one full count for the article; i.e., for an article coauthored by the United States and Canada, both the United States and Canada receive a count of one. In the percentages discussed in this paragraph, the numerators for the country pairs are the same. The denominators vary, accounting for the different rates of coauthorship.
 Readers are reminded that the number of coauthored articles between any pair of countries is the same; each country is counted once per article in these data. However, countries other than the pairs discussed here may also appear on the article.
 Identification of the sector of the non-U.S. institution is not possible with the current data set.
 Readers are reminded that coauthors from different departments in an institution are coded as different institutions.
 See note 42.
 This chapter uses the convention of a 3-year citation window with a 2-year lag, e.g., 2005 citation rates are from references in articles in the 2005 tape year to articles on the 2001, 2002, and 2003 tapes of the Thomson Scientific Science Citation Index and Social Sciences Citation Index databases. Analysis of the citation data shows that, in general, the 2-year citing lag captures the 3 peak cited years for most fields, with the following exceptions: in astronomy and physics the peak cited years are generally captured with a 1-year lag, and in computer sciences, psychology, and social sciences with a 3-year lag.
 Percentiles are specified percentages below which the remainder of the articles falls, for example, the 99th percentile identifies the number of citations 99% of the articles failed to receive. Across all fields of science, 99% of articles failed to receive at least 21 citations. Matching numbers of citations with a citation percentile is not precise because all articles with a specified number of citations must be counted the same. Therefore, the citation percentiles discussed in this section and used in appendix table 5-38 have all been conservatively counted, and the identified percentile is in every case higher than specified, i.e., the 99th percentile is always >99%, the 95th percentile is always >95%, etc. Actual citations/percentiles per field vary widely because counts were cut off to remain in the identified percentile. Using this method of counting, for example, the 75th percentile for engineering contained articles with two citations, whereas the 75th percentile for biological sciences contained articles with 5–8 citations.
 This pattern holds for even lower citation percentiles (e.g., the 95th or 90th).
 The previous edition of Indicators discussed various factors that may have contributed to the rise in university patenting, including federal statutes and court decisions (see NSB 2006, p 5-51 through 5-53).
 For an overview of these developments in the 20th century, see Mowery (2002).
 It is unclear whether the recent downturn in patents granted to universities/colleges is a result of changes in processing at the U.S. Patent and Trademark Office (USPTO). For example, in its Performance and Accountability Report Fiscal Year 2006, USPTO reported an increase in overall applications from 2002 to 2006; a decrease in "allowed" patent applications; and an increase in average processing time from 24 to 31 months (USPTO 2006).
 The institutions listed in appendix table 5-40 have been reported consistently by USPTO since 1982. Nevertheless some imprecision is present in the data. Several university systems are counted as one institution, medical schools may be counted with their home institution, and universities are credited for patents only if they are the first-name assignee on a patent; other assignees are not counted. Universities also vary in how they assign patents, e.g., to boards of regents, individual campuses, or entities with or without affiliation with the university.