Changing U.S. Output of Scientific Articles: 1988-2003
The U.S. Academic Sector
Overall Output Trends
The academic sector dominates U.S. S&E article production, accounting for nearly three-quarters of U.S. output. Because this sector plays such an important role, indicators of patterns and trends in its article output are worthy of special attention.
Similar to overall U.S. output, academic sector output grew considerably between 1988 and 1992 and flattened thereafter. Average annual growth in the U.S. academic sector was 3.0% between 1988 and 1992, but only 0.8% between 1992 and 2003. Field trends in output were generally consistent with the overall sector trend in both periods, and trends were similar whether measured in whole counts or fractional counts (table 6 and appendix table 12).
Academic sector performance compared favorably with performance in the nonacademic sectors, where growth in annual output was slower in the earlier period (2.4% versus 3.0% for the academic sector) and output declined in the later period�(-0.1% annual rate versus 0.8% for the academic sector) (table 7 ). Within the nonacademic sectors, output between 1988 and 2003 declined in the federal government, the state and local governments, and the private for-profit sectors and increased in the FFRDC and private nonprofit sectors (appendix table 13).
During the period of overall flattening in article output (1992–2003), in almost every ipIQ broad field, the academic sector's share of U.S. articles increased, while the private for-profit sector's and the federal government's shares declined (figure 10 ). The combined pattern of the academic sector's rising share of articles and the for-profit sector's falling share was especially dramatic in engineering/technology, physics, the social sciences, and the earth/space sciences. The fields of clinical medicine and biology (which has a large component of agricultural sciences) are the only two fields in which the academic sector's share did not increase and the only two in which the private for-profit sector's share did not decline.
Trends in the Top 200 Academic Research Universities
Within the academic sector, the top 200 R&D institutions (based on R&D expenditures during the 1988–2001 period) account for most of the article output. In 2001, the share of academic articles with at least one author from these institutions was 94%. Because the academic sector dominates U.S. S&E article production and the top 200 R&D institutions are at the core of the academic sector, the analyses that follow are structured around these institutions and how they operate.
Accordingly, unlike other analyses in this report, the following analyses use the S&E field classification of the Integrated Science and Engineering Resources Data System (WebCASPAR) when presenting data by field rather than the ipIQ field classification. (For a crosswalk between WebCASPAR and ipIQ fields, see appendix table 14.) In addition to aligning more closely with the way in which academic institutions are organized, WebCASPAR fields have a second important advantage: they enable analyses of the relationship between publication outputs and field-specific resource inputs (e.g., academic R&D expenditures, postdocs, S&E doctoral degree recipients).
The top 200 R&D institutions dominate S&E article output in all WebCASPAR fields, receiving authorship credit for between 82% and 83% (psychology, social sciences) and 98% (agricultural and medical sciences) of academic sector articles (figure 11 ). Output trends for these institutions basically mirror trends for the academic sector as a whole for both fractional and whole article counts (figure 12 ).
Article output trends varied somewhat for different types of academic institutions, even among the top 200 R&D performers. However, the overarching trend of stronger growth between 1988 and 1992 than in the subsequent period is manifested in practically every category of institution that the study examined.
The Carnegie Foundation for the Advancement of Teaching classification of academic institutions is a widely recognized system for distinguishing among institutions by the range and breadth of their research and education programs. The majority of the top 200 R&D institutions are classified as Carnegie research institutions (n = 125), followed by doctorate-granting institutions (n = 39) and freestanding medical institutions (n = 23) (table 8 ). Public institutions (n�= 145) constitute over 70% of the top 200 institutions.
For the entire 1988–2001 period, output growth rates were greater in private research I universities and both public and private freestanding medical institutions than in public research I universities and both public and private research II universities (table 9 and appendix table 15). There were also relatively large increases in top 200 institutions, both public and private, in the other category. For all of these Carnegie classes, growth rates for the 1988–92 period exceeded those for the 1992–2001 period.
Output growth appeared to be weakest in high-quality departments among the top 200 institutions. SRS used the 1994 National Research Council (NRC) ratings of the scholarly quality of graduate departments at U.S. universities as a measure of a university's quality in the corresponding WebCASPAR field. Figure 13 shows that a lower departmental rating for scholarly quality is generally associated with a higher rate of article growth, except for articles from departments that had a scholarly quality rating of 0 to 1. This pattern is not consistent across fields (table 10 and appendix table 16). Biology, chemistry, engineering, physics, and geosciences generally showed an increasing trend of publication with decreasing NRC ratings, and mathematics showed a reverse trend. Computer sciences, social sciences, and psychology did not demonstrate a clear trend.
Output growth trends are related to trends in institutions' R&D growth; on average, the group of institutions with the fastest R&D growth also had the fastest article growth (figure 14 and appendix table 17). This finding holds both for the entire 1988–2001 period and for the part of the period during which article growth flattened (1992–2001). (See figure 2 for an index of both R&D and article growth.)
SRS's international data indicate that during the period under study, collaboration across geographic boundaries became much more widespread. SRS's U.S. data contain information about patterns and trends in collaboration across sectoral and institutional boundaries that is not available internationally. If comparable data on other nations and regions become available, future research can examine how different national trends in collaboration across various kinds of boundaries (if such trends exist) relate to the different trends in overall national S&E publication output.
The U.S. academic sector participated in the worldwide trend toward increased collaboration across geographic boundaries. Between 1988 and 2001, the article output of the academic sector became more collaborative not only internationally, but also across U.S. sectoral and institutional boundaries. As a result, output trends for different types of articles varied substantially, depending on whether collaboration was involved and the type of collaboration that was involved. In general, article output for various types of collaborative articles increased, while single-author and single-institution output declined.
The ratio of fractional to whole counts is a measure of collaboration intensity (see sidebar "Effect of International Collaboration on Whole- and Fractional-Count Output"). For the top 200 institutions, the average ratio of fractional counts to whole counts declined between 1988 and 2001, indicating more collaboration over time. Increased collaboration occurred in each WebCASPAR field (figure 15 and appendix table 18); the fields that were more collaborative (astronomy, geosciences, physics) and less collaborative (chemistry, social sciences, psychology) in 1988 were generally the same in 2001.
For most of the top 200 institutions, the trend toward collaborative publications occurred not only on average, but also in each field. For all 200 institutions, the fractional-count to whole-count ratio for all fields combined was lower in 2001 than in 1988 (appendix table 19). At the individual field level, because some institutions produced a small number of articles in some fields in a given year, it is more appropriate to compare collaboration across two periods (1988–94 and 1995–2001) rather than for individual years throughout the 1988–2001 period. The percentage of the top 200 institutions in which collaboration was greater in the 1995–2001 period than in the 1988–94 period ranged from 72% in the agricultural and other life sciences to 98% in the biological sciences (figure 16 ).
The distribution of types of collaborative articles at the top 200 academic institutions changed dramatically between 1988 and 2001. Articles in which an author at a top 200 institution participates can be broken down into single-author articles and six types of multiple-author articles (figure 17 and appendix table 20):
Collaborative articles across institutional boundaries (the last four categories listed above) increased substantially during this period, whereas other types of articles did not (figure 17 ). In absolute numbers, the largest increase (about 19,000) occurred for articles in which authors from the top 200 institutions collaborated with authors at foreign institutions. In relative terms, the most dramatic increase (almost 300%) was for articles in which authors from the top 200 institutions collaborated with authors from both other U.S. institutions and foreign institutions, but this growth was from a small base. In contrast, articles with either a single author or multiple authors from a single department declined, and the number of articles with authors from multiple departments at a single institution remained fairly flat. The net result of these trends was a drop from 60% to 46% in the proportion of articles credited to a single top 200 academic institution and an increase from 10% to 23% in the proportion of publications with at least one author from a U.S. institution and one author from a foreign institution (figure 18 ). The trends in individual fields were similar to the overall change (figures 19 and 20 ). The extent of collaboration varied considerably across fields.
Collaborative articles also showed a trend toward involving more institutions and more individual authors. For articles with participation by a top 200 institution, the numbers of authors, top 200 institutional addresses, and non-top 200 institutional addresses increased between 1988 and 2001 (figure 21 ). The number of authors per article increased by nearly 50%, and the number of non-top 200 institutional addresses per article almost doubled during this period. The number of top 200 institutional addresses per article increased less rapidly, rising by less than 14%.
 Other sectors in which researchers produced articles are the federal government, the private for-profit sector, FFRDCs, the private nonprofit sector, and state and local governments. Articles that cannot be assigned to one of these sectors are categorized as "unknown."
 For technical reasons, the analysis of the top 200 research universities was confined to the 1988–2001 period. However, data for 2003 are generally consistent with the patterns and trends reported in this section.
 Such analyses will be presented in a future report, as discussed in the conclusion below.
 For the purposes of this study, the 1994 Carnegie classification was more useful than the 2000 version because it relies on criteria that are more relevant to the institutional characteristics associated with research and publication. The Carnegie Foundation recently completed a complete revision of its classification system.
 Research I and II universities offer a full range of baccalaureate programs, are committed to graduate education through the doctorate, and give high priority to research. Both classes of institutions award 50 or more doctoral degrees each year. Research I institutions receive $40 million or more in federal support annually and research II institutions receive between $15.5 million and $40 million annually.
 There were no ratings for the medical sciences.