Chapter 5 | Academic Research and Development

Outputs of S&E Research: Publications

Related Content

• Related Content-
• Publication Output, by Country
• Publication Output, by U.S. Sector

In 2016, developed economies produced nearly 1.4 million S&E publications, whereas developing economies produced just over 900,000 S&E publications. However, over the last decade, publications from developing economies grew faster than those from developed economies (8.9% versus 1.7%). U.S. S&E publication production grew from just over 383,000 in 2006 to almost 410,000 in 2016, growing merely 0.7% (Appendix Table 5-27). As U.S. publication volume leveled off and developing economies’ publication volume grew more rapidly, the U.S. global share fell from 24.4% in 2006 to 17.8% in 2016 (Figure 5-22).

The top five countries producing S&E publications in 2016 are China (18.6% of global output volume), the United States (17.8%), India (4.8%), Germany (4.5%), and the United Kingdom (4.3%). When treated as one entity, the European Union (EU) accounts for 26.7% of the world’s S&E publications in 2016 (Table 5-22; Figure 5-22).^​ Although Japan has been a major producer for several decades, Japan’s output has trended downward since 2013. In 2016, Japan was the sixth largest global producer of S&E publications. Together, the United States, China, and the EU accounted for almost two-thirds of the world’s S&E publications in 2016.

China has continued its steady growth since the mid-2000s and is now the largest global producer (with an 18.6% global share). India’s publication volume has grown rapidly, from 2.5% of global output in 2006 to 4.8% in 2016. Overall, 50 countries—a quarter of those that produced S&E publications in 2016—account for 96.9% of global output.

Between 2006 and 2016, total world S&E publication output grew at an average annual compound rate of 3.9%; the total for developing countries grew more than twice as fast (about 8.6%).^​ China’s 8.4% growth rate led the developing countries, resulting in China’s collective global share climbing from 12.1% in 2006 to 18.6% in 2016.^​ The strong growth in the developing world points to rapidly increasing science and technology capabilities.

Among other large emerging economies, the 2006–16 average publication growth rate in India was 11.1%; Brazil averaged 6.6%, but this was from a much lower base of total publications. India’s and Brazil’s 2016 global shares increased to 4.8% and 2.3%, respectively, with India becoming the third largest producer of S&E publications (Table 5-22). The change in the absolute number of publications seen during the last 10 years provides context to the growth rate. The absolute increase in the number of publications between 2006 and 2016 is much larger for China (236,406) and India (71,729) than for Brazil (25,447). Rapid growth of S&E publications in Brazil, India, and China coincided with increased R&D expenditures and growth in S&E degrees awarded at the bachelor’s- and doctoral-degree levels (see Chapter 2 section International S&E Higher Education). Smaller developing countries with more than 5,000 publications in 2016 and over 15% growth rate from 2006 to 2016 included Iran, Malaysia, Saudi Arabia, Indonesia, and Colombia.

The output of the EU, the world’s largest producer, grew 2.5% from 2006 to 2016, faster than the average for developed countries (1.7%). Among EU member countries, growth rates were lower for the three largest producers—France (1.1%), Germany (2.0%), and the United Kingdom (1.0%)—and were generally much higher in smaller member countries. Several former Eastern Bloc countries, including the Czech Republic, Romania, and Slovakia, had publication growth rates above 6.0% for 2006–16. Like that of the United States, the EU’s global share fell, from 30.7% in 2006 to 26.7% in 2016.

In Japan, absolute numbers of S&E publication output declined at a 1.3% average rate over 2006–16, decreasing Japan’s global share from 7.0% to 4.2% during this period. Publication output from other developed economies outside of the EU and the United States grew, particularly in Australia, Norway, South Korea, and Singapore.

The distribution of S&E publication output by field indicates the priorities of scientific research in different locations. The S&E publication portfolios of five major producers—the United States, the EU, China, Japan, and India—have distinct differences by field (Table 5-23; Appendix Table 5-28 through Appendix Table 5-40). Nearly half of U.S. publications are focused on biological sciences, medical sciences, or other life sciences, compared with 38.6% for the world at large in 2016. The United States also produces a higher proportion of S&E publications than the rest of the world in psychology and social sciences. In this context, it is useful to acknowledge that publications in the Scopus database must have an abstract in the English language to be included in the publication counts (Archambault et al. 2009), meaning that publication counts in the social sciences, where publications are more likely in the national language, may be underestimated where English is not the country’s national language.

Like the United States, the EU is more focused on biological sciences, medical sciences, and other life sciences; these three fields together account for 40.7% of the EU’s publications, compared with the 38.6% of world publications. Relative to the United States, the EU has higher shares of publications in physics, chemistry, and engineering. Relative to the world total, China’s S&E publications are more heavily focused on engineering, chemistry, physics, and geosciences. Engineering publications made up 28.9%, and chemistry publications made up another 12.3%, of China’s publication output in 2016.

Engineering publications as a share of total publication output volume for India in 2016 (24.2%) are also above the world proportion (18.4%). India’s portfolio has the highest concentration in computer sciences of the regions, countries, and economies discussed here, with a 14.1% share, and is above world average concentration in chemistry.

Recent bibliometric research has focused on merging administrative data sets to explore publication data by gender, potentially revealing differences between countries and research fields (see sidebar S&E Publication Patterns, by Gender).

This report divides the U.S. institutional landscape into six sectors, each of which produced S&E publications: the federal government, industry, academia, federally funded research and development centers (FFRDCs), private nonprofit organizations, and state and local governments.^​

In the United States, the academic sector is the largest producer of S&E publications, accounting for three-fourths of U.S. S&E publication output. This sector was largely responsible for the growth of U.S. S&E publication output between 2006 and 2016. The number of academic S&E publications increased from 271,502 to 307,413 between these years, rising from 70.9% to 75.2% as a share of all U.S. publications (Figure 5-23). Public universities accounted for 44.2% of all U.S. publications, and private universities accounted for 25.3% (Appendix Table 5-41).

S&E publications in U.S. non-academic institutions fell from 87,513 in 2006 to 76,012 in 2016. Trends in non-academic sectors include the following (Appendix Table 5-41):

• Industry publications reached a high of 33,498 in 2005 and then declined to 24,565, or 6.0% of the U.S. total, in 2016.
• Federal government publications grew in the early 2000s, peaking at 22,580 in 2012 and declining to 19,556 in 2016, accounting for 4.8% of the U.S. total in 2016.
• Publications from FFRDCs grew to a peak of 10,927 in 2012 and have declined to 9,107 in 2016.
• Publications with institutional addresses in the private nonprofit sector decreased slightly (from 21,555 in 2006 to 21,248 in 2016), accounting for 5.2% of U.S. publications in 2016.

As noted previously, life sciences (biological sciences, medical sciences, and other life sciences) dominate the research portfolios of U.S. sectors, accounting for nearly half or more of all publications produced in the federal government, academic, private nonprofit, and state and local government sectors (Appendix Table 5-41). The dominance of life sciences is especially pronounced in the nonprofit sector, with 86.9% of publications in these fields: 63.1% in medical sciences, 19.8% in biological sciences, and 4.1% in other life sciences. With a much larger number of total publications, academia has 49.2% of its S&E literature in life sciences. The exception to the life sciences focus is the research portfolio of industry (28.2% engineering) and FFRDCs (29.6% physics); most of the FFRDCs are controlled by DOE or DOD.^​ The largest science fields in the FFRDC portfolio are physics (within physical sciences) (29.6%), chemistry (14.1%), and engineering (24.6%).

Coauthorship on S&E research publications is based on multiple institutional addresses associated with the same publication. Such interconnections among researchers in different institutional settings may indicate researchers’ growing capacity to address complex problems by drawing on diverse skills and perspectives. Collaborative S&E research facilitates knowledge transfer and sharing among individuals, institutions, and nations. Between 2006 and 2016, international collaboration increased; domestic-only collaboration held steady as a share of the total, and single-institution authorship declined (Figure 5-24).

U.S. coauthorship data at the sector level—academic, nonprofit, industry, FFRDCs, and federal and state government—indicate collaboration among U.S. sectors and between U.S. sectors and foreign institutions. Over the period 2006–16, the share of publications produced in collaboration with other U.S. sectors or with foreign institutions increased in all sectors (Table 5-25).^​ The proportion of academic publications coauthored with other U.S. sectors and foreign institutions increased from 41.4% in 2006 to 51.0% in 2016. The share of academic publications coauthored with foreign institutions increased from 24.9% to 37.2% over this period. FFRDCs, where the research conducted focuses primarily on the physical sciences, have the highest percentage of international coauthorship of U.S. sectors, at 45.6% in 2016.

The percentage of worldwide publications produced with international collaboration—that is, by authors with institutional addresses from at least two countries—rose from 16.7% to 21.7% between 2006 and 2016 (Figure 5-24). This increase in part reflects increasing global capabilities in R&D and an expanding pool of trained researchers, as well as improvements in communications technology. These collaborations may also reflect the strengthening of a network of international scholars who increasingly collaborate with each other (Wagner, Park, and Leydesdorff 2015). Finally, the research challenges of climate change, food, water, and energy security are fundamentally global, rather than national, in scope, thereby calling for international research collaboration (Royal Society 2011). Although these factors affect the overall trend, the patterns of international scientific collaboration also reflect wider relationships among countries, including linguistic and historical factors, as well as geographic, economic, and cultural relations (Glänzel and Schubert 2005; Narin, Stevens, and Whitlow 1991).

The increase in international coauthorship occurs in every broad field of science. Astronomy is the most international field, with more than half of its publications internationally coauthored (54.0%) in 2016 (Figure 5-25). Geosciences, mathematics, biological sciences, and physics also have percentages of international collaboration above the average of 24.2% across all fields. Factors influencing variations among fields include the existence of formal international collaborative programs and the use of costly research equipment (e.g., atomic colliders, telescopes), which result in cost sharing and collaboration among countries. However, even fields with relatively low percentages of international collaboration have experienced increases in collaboration between 2006 and 2016. For example, over the time period, social sciences grew from 11.4% to 15.4%, and engineering grew from 13.7% to 17.7%.

Countries vary widely in the proportion of their internationally coauthored S&E publications. Scale effects play a role in this. Countries with large populations or communities of researchers may have high rates of domestic coauthorship because of the large pool of potential domestic coauthors in their field. Researchers in smaller countries have a lower chance of finding a potential partner within national borders, so collaborators are more likely beyond their national borders. The EU program Horizon 2020 (like its predecessor, the 7th Framework Programme for Research and Technological Development) actively promotes and funds international collaboration within the EU.^​

The aforementioned publication output data in Figure 5-22 show the 28 nations of the EU as one region.^​ By individual country, Figure 5-26 shows the percentages of international collaboration for the largest producers of S&E publications in 2016. The nations within this group that had the highest percentages of international collaboration in 2016 were three EU nations, the United Kingdom, France, and Germany, which are also the three largest European producers of S&E publications. International collaboration increased for these European countries between 2006 and 2016.

China increased its collaboration percentage across the same period but was slightly below the world average for 2016 (21.7%). The share of Indian publications that are coauthored with another country declined from 18.5% in 2006 to 17.4% in 2016.

This section describes global partnership patterns, with special focus on patterns of U.S. involvement in international collaboration.

U.S. institutional authors collaborate most frequently with authors from China, currently the largest producer of S&E publications. China accounted for 22.9% of U.S. internationally coauthored publications in 2016 (Table 5-26). Other substantial partners for the United States include the United Kingdom (13.4%), Germany (11.2%), and Canada (10.2%).

China, South Korea, and Canada are notable among these countries for having unusually high percentages of U.S. participation in their own internationally coauthored publications (46.1%, 47.6%, and 43.5%, respectively). For the other 12 countries in Table 5-26, the shares range from 25.3% to 36.1%.

A measure of the relative strength of collaborative ties between two countries can be obtained by dividing a country’s share of collaboration with a specific country by its overall share of international collaborations. This index is 1.00 (unity) when coauthorship between two countries is exactly proportional to their overall shares of international collaborations (Table 5-27). Index values above 1.0 indicate stronger ties, while scores below 1.0 indicate weaker collaborative ties (see sidebar Bibliometric Data and Terminology).

Geographical regional collaboration, as measured by this index of international collaboration, shows trends that reflect geographic proximity and other historical factors (Table 5-27; Appendix Table 5-43 and Appendix Table 5-44). In North America, the Canada-U.S. index shows a percentage of collaboration that is 13% (1.13) greater than would be expected by size of overall international collaboration alone and that has not changed much between 2006 and 2016.

Proximity alone does not explain these relationships. Language may be a factor. The U.S.-Mexico index is relatively stable and is just what would be expected by overall shares of international coauthorship alone—near unity. Mexico has very strong collaboration with the Spanish-speaking South American nations of Argentina and Chile (4.64 and 4.09, respectively, for 2016). In turn, Argentina is likely to collaborate with regional neighbors Brazil and Chile. Collaboration between the United Kingdom and Ireland is more than twice what would be expected: 2.16 in 2016.

In addition to the above-average relationships that reflect geographic proximity, Appendix Table 5-43 shows other strong collaboration relationships that reflect historical, linguistic, and educational ties between nations. For example, Spain had a collaboration index measure in 2016 that was two to three times higher than expected with Mexico, Argentina, and Chile. Despite the substantial geographic distances, the United Kingdom has a higher-than-expected collaboration index with Australia and New Zealand. Malaysia has higher-than-expected collaboration ties with the Middle East nations Iran and Saudi Arabia in 2016.

Strong collaboration relationships also evolve over time among countries with strong educational ties, such as the United States and China, where the collaboration index has increased from 0.88 to 1.19 from 2006 to 2016. China is the largest foreign country of origin for international recipients of U.S. S&E doctorates. China accounted for more than one-quarter of all international S&E doctorate recipients from 1995 to 2015 (see Chapter 2 section International S&E Doctorate Recipients).

The publication counts and collaboration rates described previously provide partial indicators of the quantity of S&E research output and the ties between researchers. Citations provide an additional indicator of the impact of research on subsequent work (Martin and Irvine 1980). This section provides indicators of S&E publications that are cited in other S&E publications. Citations indicate impact and are increasingly international in character, with publications authored in one country citing those authored in other countries. An increase in citation trends potentially indicates stronger international ties, or it can indicate changes in quality of research outside the country or region. Measured by average citation rates and by the shares of the most highly cited publications, the developed world continues to maintain a substantial impact advantage over the developing world. Nevertheless, the developing world is making rapid gains.

The next sections examine two aspects of publication citations in a global context: the overall citation rate of a country’s scientific publications, and the share of the world’s most highly cited literature authored by different countries. The discussion of publication citations will conclude with an examination of citations to publications authored by researchers at U.S. academic institutions and in other U.S. sectors.

The rate of citations to S&E literature vary across fields of science and frequently peak within a few years after publication. However, even old publications can “awaken” to receive citations many years after publication (Ke et al. 2015). The average of relative citations (ARC) presented in this chapter is an index designed to allow for lags of varying length and to normalize across fields of research (see sidebar Bibliometric Data and Terminology). In contrast, when looking at the share of a country’s citations that come from an international source, data presented in this report are calculated based on only the subsequent 24–36 months after publication, depending on the month in which a publication initially appears.

Between 2004 and 2014, the share of citations to U.S. publications that come from abroad increased from 47.0% to 55.7% (Figure 5-27). The relative shares of foreign citations increased in most countries of the world over that same period (Appendix Table 5-47). By contrast, China’s international share of citations decreased, from 42.2% in 2004 to 37.7% in 2014. However, the international share of a publication’s use in citations tends to decrease as a country’s domestic publication volume grows. Thus, changes in international citations for China were to be expected; China dramatically increased its output in recent years, so the associated decrease in its international share of citations supports the overall trend. The inverse relationship also holds because the U.S. share of total world output has decreased and the international share of U.S. citations has grown. Similarly, between 2004 and 2014, almost three-quarters (21) of the 28 countries in the EU increased their international share of citations (Appendix Table 5-47), and the EU as a unit increased its external share of citations from 43.7% to 48.1%, all while Europe’s share of global output has gradually decreased. One country that does not follow this pattern is India, whose output volume and international share of citations have been increasing (Figure 5-22 and Figure 5-27); additionally, India’s international share of citations fluctuates year over year. Further investigation may determine whether these observations result from changes in the database coverage over time or whether underlying changes in India’s research ecosystem explain these phenomena.

The patterns of international citations between country pairs shows the impact of one country’s S&E publications on another country’s S&E researchers. The relative citation index normalizes cross-national citation data for variations in relative size of publication output, much like the collaboration index (see sidebar Bibliometric Data and Terminology). The expected value is 1.00, but unlike the collaboration index, citation index scores are not symmetric. For example, if country A cites publications by country B 15% more often than expected, that does not mean that country B also cites publications by country A 15% more than expected. Table 5-28 shows the relative citation index for 2014 for major publishing locations in four regions: North America, South America, the EU, and Asia. These data show the following:

• From among the major producers of S&E publications, U.S. publications cite publications from Canada (1.15) and the United Kingdom (1.13) more than expected, based on size.
• U.S.-based authors cite Chinese (0.31), Indian (0.20), and other Asian S&E publications less than expected.
• Publications by Mexican researchers are heavily cited in publications from Argentina and Chile. Likewise, Mexican researchers cite publications by South American authors more than they cite publications from other areas of the world.
• Inter-European influence is strong, with France, Germany, and United Kingdom country pairs exhibiting index values greater than 1.0 (with the exception of the United Kingdom citing France, which occurs at the expected rate).

Similar to the coauthorship patterns, these data indicate the strong influence that geographic, cultural, and linguistic ties have on citation patterns.

The ARC scores are calculated to allow for citation lags of varying lengths and to normalize for field of research (see sidebar Bibliometric Data and Terminology). Appendix Table 5-50 provides the ARC scores for 1996–2014 for countries and regions with enough publications to compute robust scores. Through 2014, the U.S. ARC score held steady around 1.40, or 40% more citations than would be expected, based on the number of peer-reviewed publications and representation by field. China’s ARC score rapidly increased across 2004–14, from 0.62 to 0.96, improving from 38% fewer citations than would be expected, based on size, to just reaching the expected level of citations.

When viewed as a group, the countries of the EU increased from slightly more citations than would be expected based on size (1.06) in 2004, to nearly 10% more (1.16) in 2014, based on ARC scores (Figure 5-28). Appendix Table 5-50 provides country-level measures for the EU that show that Belgium, Cyprus, Denmark, Estonia, Finland, Ireland, the Netherlands, Sweden, and the United Kingdom had the highest ARC scores in 2014 (all at or above 1.50). In East and Southeast Asia, Singapore had the highest ARC score, reaching 1.83 in 2014.

At the field level, the average impact of U.S. publications is also higher than would be expected. U.S. citation impacts for computer sciences publications are especially high, at 69% higher than the world average value. Although the 2014 U.S. citation impacts remain above the world average for all fields combined—and individually for each of the 13 broad fields of science—average U.S. impact has been decreasing between 2004 and 2014 in engineering, mathematics, chemistry, social sciences, and psychology, while U.S. physics ARC remained unchanged during that period (Figure 5-29).

Among all publications, only a small share receives more than a few citations. Publications that are in the top 1% of total global citations, adjusted for field and year, can be considered to have the highest impact. Citations are indexed to show a country’s share of publications among the world’s top 1% most-cited literature. This top 1% of publications can be segmented by authors’ institutional addresses to show which countries and regions are producing these high-impact S&E publications. Similar to the ARC, country and region contribution rates for highly cited publications need to be normalized for the share of total publications produced by each country. This indicator is based on the expectation that if highly cited publications were evenly distributed, every country would have 1% of its publications among the top 1% of publications ranked on the basis of citation. If 2% of a country’s publications are among the top 1% of all cited publications, the country has twice as many highly cited publications than would otherwise be expected, based on its number of publications.

World citations to U.S. research publications show that, in all broad fields of S&E, U.S.-based researchers continue to produce a larger-than-expected percentage of the top 1% most highly cited publications. Even when normalized for U.S. overall publication output, the U.S. share is one of the highest among major S&E producers. Of publications that appeared in 2014, almost 2% of U.S.-authored publications are among the top 1% of publications, ranked by citation. This pattern of U.S. publications receiving more citations than expected holds throughout the top half of the percentile distribution; U.S. publications are more likely to be in the top 5%, 10%, and 20% and are less likely to be in the bottom 50% of the distribution of articles, based on relative citations (Appendix Table 5-48).

U.S. publications in the fields of agricultural sciences, astronomy, biological sciences, computer sciences, geosciences, medical sciences, and physics are a growing share of the top 1% most-cited articles worldwide, with about twice as many U.S. publications in this elite set in 2014 than would be expected, based on output volume. In three fields—chemistry, engineering, and mathematics—the U.S. relative share of the top 1% of articles declined between 2004 and 2014 but held constant for social sciences (Appendix Table 5-48).

Between 2004 and 2014, China and the EU experienced more rapid growth than the United States in their share of the world’s most highly cited publications (Figure 5-30). The share of China’s publications among the top 1% of publications ranked by citation increased from 0.49 to 1.01. China’s scores on the top 1% most-cited scholarly literature by research field are highest in chemistry, computer sciences, mathematics, other life sciences, and social sciences (Appendix Table 5-48).

During this same period, several of the smaller research-intensive nations of the EU have made gains in their relative share of the top 1% of highly cited publications—notably, Austria, Belgium, Cyprus, Denmark, Estonia, Finland, Greece, Ireland, Luxembourg, the Netherlands, and Sweden (Appendix Table 5-51). Each of these nations had a larger share than that of the United States in 2014. Figure 5-31 shows the top 1% shares for the United States, the EU, the Netherlands, Sweden, and Switzerland. Among other European nations with over 15,000 publications, Belgium, Czech Republic, France, Germany, Italy, Spain, and the United Kingdom were at or above 1% shares in recent years.

Relative citations can also be used to examine the citation impact of publications by each U.S. sector. Figure 5-32 shows the index values for each of the six sectors of U.S. institutions relative to world output, normalized by field, and how they have changed between 2004 and 2014. U.S. academic publications, which make up the vast majority of U.S. publications, held constant at about 50% more citations on average than would be expected. Publications authored at FFRDCs have shown a marked improvement since 2004; in 2014, they received the highest index value of all U.S. sectors—almost 90% more citations on average than would have been expected.