U.S. Academic Scientific Publishing
5.0 Trends in Publications, Citations, and Resources
This section describes trends in publications, citations, and resources (personnel and funding) of the top 200 R&D performing academic institutions from 1988 to 2001.
Section 5.1 discusses trends in publications and citations at the institution level. Graphs are presented which show the publication counts for the top 200 R&D performing academic institutions by year using both fractional and whole counting in both the fixed and expanding journal set. Depending on which measure of publications is used, the trend in publications since the mid 1990s is either flat or declining. The gap between fractional and whole counts has widened reflecting increased collaboration. Other graphs are presented that show trends in citation counts (which demonstrate similar trends to those in publications) and the ratio of citations to publications (which is generally rising, although this may be due to factors other than increased influence).
Section 5.2 discusses publication trends across years for various field groups. For three of these groups (engineering, math and physical sciences; medical sciences; and biology, life and agricultural sciences) the trends are reasonably consistent; for the social sciences and psychology group the trend is similar in the early and later years, but more erratic in the intermediate years.
Section 5.3 discusses changes from 1988 to 2001 in the primary personnel and monetary resources for publications (i.e., faculty, non-faculty doctoral research staff, postdoctoral researchers, S&E Ph.D. recipients, and total academic R&D expenditures), and compares those trends to those in publication counts. Three of these resources (total academic R&D expenditures, the number of postdoctoral researchers and the number of non-faculty doctoral research staff) increase much more rapidly than publication counts. S&E Ph.D. recipients increase at approximately the same rate, and the total number of faculty (including non-S&E faculty) increase at a much lower rate. However, limitations of the available measures of faculty and other faculty data suggest that the increase in S&E faculty with primary research responsibilities may be close to or only slightly less than the rate of increase of publication counts. Thus, the data suggest that total resources per publication have increased over time, which may be attributable to a variety of factors (including changes in regulatory and administrative burden, need for more extensive or complex collaborations, decreased cross-subsidization, etc.). It also suggests that the ratio of faculty to other resources (i.e., postdoctoral researchers, non-faculty doctoral research staff, etc.) may have declined.
5.1 Publication and Citation Trends Across All Fields
During the period from 1988 to 2001, the annual number of publications as measured by fractional counts in the expanding journal set for the U.S. as a whole (including academic, federal, state and local, private, and other sectors) increased from 177,662 to 200,870. Academic institutions accounted for a rising percentage of this total, from 71.7% in 1988 to 73.6% in 2001. The top 200 R&D performing academic institutions accounted for 87.7% of the academic total from 1988 to 2001 and this percentage has varied little from year to year.
Figure 5 displays publications as measured by fractional counts aggregated over the top 200 R&D performing academic institutions for the period 1988 to 2001 and publications as measured by whole counts for these institutions considered as a single entity (i.e., the observed number of publications in which one or more authors from any of the top 200 R&D performing academic institutions participated). Publications as measured by whole counts are larger than publications as measured by fractional counts because non-top 200 R&D performing academic institutions may have participated in authorship. Note that if the sum of the whole counts for individual institutions had been graphed, their values would have been considerably larger than either the sum of the fractional counts or the number of whole count publications for all top 200 R&D performing institutions considered as a single entity.
As seen in figure 5, publications as measured by fractional counts in the expanding journal set increased from 1988 to 1995, and have been approximately constant since that time, although there appears to be a slight increase in 2001. A very similar pattern is evident in publications as measured by whole counts in the expanding journal set, which increased from 1988 to 1996, was almost constant during the period from 1996 to 2000 inclusive and began to rise again in 2001. While the resources (R&D funding, S&E postdoctorates, and S&E graduate students) available to the top 200 R&D performing academic institutions increased during the period from 1996 to 2000, publications as measured by fractional and whole counts in the expanding journal set have remained roughly constant. The increasing gap between fractional and whole count publications in the expanding journal set is due to increasing amount of collaboration between top 200 R&D performing academic institutions and other institutions (both foreign and domestic) over this period of time.
Publications as measured by fractional counts using the fixed journal set increased from 1988 to 1992 and then decreased from 1992 through 1996 at which time it was approximately at the same level as 1988. After that point in time it continued to decline until stabilizing in 2001. Publications as measured by whole counts using the fixed journal set increased from 1988 through 1992, and then started to decrease, and has been approximately constant since 1997, at or slightly above the 1988 level. The widening gap between fractional and whole count publications in the fixed journal set is due to increased collaboration of top 200 R&D performing academic institutions with other institutions. The pattern for counts in the fixed journal set may differ from the pattern of counts in the expanding journal set for a number of reasons, among which may be a shift away from the fixed set to other journals such as in the expanding set.
Figure 6 shows the relative changes in these various publication counts, normalized so that the counts in 1988 equal 100. The similarity in the pattern of change is quite evident. Publications as measured by fractional and whole counts in the expanding journals set exhibited faster growth between 1988 and 1992 than fractional and whole count publications in the fixed journal set. Growth of fractional and whole count publications in the expanding journal set, however, slowed starting in 1992. Publications as measured by fractional and whole counts in the fixed journal set showed a general downward trend after 1992. The gap between whole and fractional count publications in both the expanding and fixed journal set widened during this period due to increased collaboration of top 200 R&D performing academic institutions with other institutions.
Figure 7 displays citation counts aggregated over the top 200 R&D performing academic institutions for the period 1992 to 2001. Figure 8 displays the citation counts normalized to 1.0 in 1988. Citation counts are based on a 3-year window with a 2-year lag. For example, the whole citation count for 1992 for an institution is equal to the number of citations in 1992 to any articles produced by that institution from 1988 through 1990. Because article data are not available before 1988, the first available year of citation data is 1992.
Citations as measured both by fractional counts and by whole counts in the expanding journal set increased from 1992 through 1999. Citations as measured by fractional counts in the expanding journal set appeared to have leveled off from 1999 to 2001, while citations as measured by whole counts continued to increase during this time period. The rate of growth of citations as measured by whole counts for the expanding journal set is somewhat higher than the rate of growth of citations as measured by fractional counts, possibly as a result of more collaboration among institutions and multiple whole counting of citations when there are authors from more than one top 200 R&D performing academic institution.
For the fixed journal set, the number of citations as measured by whole counts increased from 1992 to 1996 and then appeared to be approximately constant through 2001. Citations as measured by fractional counts in the fixed journal set increase slightly through 1996, and then decrease through 2001. Some of this pattern may be attributable to the similar pattern in the number of publications in the fixed journal set (that is, if there is a constant or slightly decreasing number of publications, then one would expect a constant or slightly decreasing number of citations in the absence of change in the citation to publication ratio).
Figure 9 shows the citation to publication count ratio for the top 200 R&D performing academic institutions. For example, the 1992 value of this ratio for whole counts in the fixed journal set is the number of citations as measured by whole counts in the fixed journal set in 1992 for journal articles published in 1988 through 1990 divided by the number of publications as measured by whole count in the fixed journal set from those same years. For the fractional count fixed journal set the pattern is an increasing ratio from 1992 through 1996, a constant ratio from 1996 through 1999, and a slight decline thereafter. For the whole count fixed journal set the pattern is a increasing ratio from 1992 through 1999, and a slight decline thereafter. For the expanding journal set the pattern is an increasing ratio from 1992 through 1999, and a decrease in the growth rate or a leveling off thereafter. This figure should not be interpreted to mean that U.S. publications have become more influential over time. Some of the increase in citation counts in the expanding journal set is due to there being more articles being counted each year and therefore more opportunities for citation; some of the increase in whole counts may be due to there being more authors per paper, each of whom may cite his or her own prior work.
5.2 Publication Trends by Field Groups
Each publication was assigned by ipIQ, Inc. (formerly CHI Research, Inc.) to a single field based on the journal in which the publication appeared. NSF specified how these fields were to be aggregated into "field groups." The field groups were labeled computer sciences (field 42); medical sciences (field 53); biology, life and agricultural sciences (bio-life-ag) (fields 51, 52, and 59); engineering, math and physical sciences (eng-math-physical sciences) (fields 11 to 41); and social sciences and psychology (soc-sciences) (fields 60 to 79, and 99). A listing of the constituent fields can be found in appendix C. In the "eng-math-physical sciences" field group the distribution of publication fractional counts in the expanding journal set is as follows: 1) 31.7% physics, 2) 24.9% chemistry, 3) 17.7% engineering, 4) 14.5% geosciences, 5) 7.5% math and statistics, and 6) 3.6% miscellaneous (astronomy and other physical sciences).
Over the period from 1988 to 2001, the average number of publications as measured by fractional counts in the expanding journal set was as follows: 1) bio-life-ag (39,430), 2) computer sciences (1,221), 3) eng-math-physical sciences (37,873), 4) medical sciences (34,207), and soc-sciences (11,958). Publication counts for the computer sciences field are smaller than other fields and exhibit more variability. Publication counts in the ISI database may have been a poorer measure of total publication output for computer sciences due to more acceptance in that field of publication in media not counted in the ISI database. Consequently, in this report, we sometimes exclude computer sciences from field-specific analyses.
Figure 10 shows the publication counts for the different field groups (excluding computer sciences) as measured by fractional counts in the expanding journal set. Three of the field groups (bio-life-ag, eng-math-physical sciences, and medical sciences) appear to demonstrate approximately the same behavior, generally increasing to about 1996 and then leveling off. The soc-sciences field group demonstrates a more erratic path, with increases until 1990, decreases from 1991 to 1994, increases in 1995 and 1996, decreases in 1997 and 1998, and then increases to 2001. Data for computer sciences are not shown.
Figure 11 shows the publication counts for the different field groups as measured by whole counts in the expanding journal set. While all field groups show growth from 1988 to 2001, the growth for the soc-sciences group is less and slightly more erratic. Data for computer sciences are not shown.
5.3 Trends in Resources
Resources that are primary drivers of publications include academic R&D expenditures, number of faculty, and number of S&E postdoctorates. Figure 12 shows the change over time in these resources in the top 200 R&D performing academic institutions, scaled so that the 1988 value equals 1.00. This figure also shows the number of whole count publications in the expanding journal set, number of S&E Ph.D. recipients, and number of non-research faculty.
As shown in figure 12, the fastest rising resource is total academic R&D expenditures (adjusted using the GDP implicit price deflator), which increased by 81% between 1988 and 2001. The next largest increases occurred in S&E postdoctorates (66%), non-faculty research staff with doctorates (52%), and science and engineering (S&E) Ph.D. recipients (23%). The number of whole count publications increased by 27%. All of the trends are generally increasing, except for number of total faculty.
The faculty counts shown in figure 12 came from the NCES IPEDS Salaries, Tenure and Fringe Benefits of Full-time Instructional Faculty Survey and have three limitations. First, faculty counts only include faculty whose major (more than 50%) regular assignment is instruction (including those with released time for research) so that any growth in faculty members whose regular time allocation is 50% or greater in research is not captured. Second, faculty counts are not available by discipline, so the faculty counts include faculty in non-S&E disciplines. Consequently, the rate of increase of faculty in S&E disciplines may be misestimated. Third, the counts are of the total number of full-time faculty (as defined by the institution), not the number of full-time equivalents. Consequently, if the growth rate of part-time faculty were less than the growth rate of full-time faculty, the growth rate of full-time faculty would overestimate the growth rate of FTE faculty; similarly if the opposite were true, the growth rate of full-time faculty would underestimate the growth rate of FTE faculty. The use of the NCES faculty counts in modeling S&E publications is dependent upon the assumption that the rate of growth of S&E faculty in the top 200 R&D performing academic institutions is the same as the rate of growth for all full-time faculty in those institutions. All full-time faculty increased by 5.4% from 1988 to 1999.
An alternative source of data on faculty counts is the NSF Survey of Doctorate Recipients (SDR), which can be used to estimate the number of faculty members with S&E doctoral degrees granted by U.S. institutions at the national level. This survey does not include S&E faculty who do not have at least one doctoral degree granted by a U.S. institution. It would underestimate the growth in S&E faculty if faculty with only foreign doctorates had increased faster than the growth of S&E faculty with at least one U.S. granted doctoral degree. For the SDR survey, full time faculty consist of those ranked as instructor, assistant, associate, or full professor who are employed at universities, colleges, and 2-year postsecondary institutions and who have full-time status in the labor force. The NSF SDR shows that for all academic institutions the number of full-time faculty with teaching or research positions and doctoral degrees in S&E increased by 10.5% from 1988 to 1999 (using interpolation to obtain the 1988 estimate) and by 11.7% from 1988 to 2001. However, for research universities, the number decreased by 0.05% between 1988 and 1999 and by 0.08% from 1988 and 2000. If one expects the change in S&E faculty for the top 200 R&D performing academic institutions to be somewhat between the change in S&E faculty for research universities and all academic institutions, this suggests that the assumption underlying the use of the NCES faculty counts may be reasonable.
The NSF SDR data also reveal that the change in S&E faculty counts is not uniform across principal work activity. For all academic institutions, the increase from 1988 to 2001 in S&E faculty with primarily teaching responsibilities was 4.3%, while for faculty with primarily research responsibilities the increase was 21.7%. In research universities, faculty with primarily teaching responsibilities decreased by 14.9% from 1988 to 2001, while those with research responsibilities increased by 8.6%. Thus, while the NCES faculty counts may reflect changes in total S&E faculty in the top 200 R&D performing academic institutions, those counts may not reflect the growth rates of faculty with primarily research responsibilities.
Based on the available data it seems likely that the number of S&E faculty with primary or secondary research and teaching responsibilities either increased at about the same rate as publication counts or somewhat more slowly. As noted previously, however, resources other than faculty required per publication (which we shall term the publication "efficiency") increased from 1988 to 2001 resulting in the ratio of faculty to other resources declining. Our database does not address the reasons for increased resources per publication (such as increased regulatory and administrative burden, need for more extensive or complex collaborations with the associated communication overhead, decreased cross-subsidization of medical research from clinical revenues, etc.).
Averaged over the period from 1988 to 1999, based on the NCES IPEDS data, the distribution of full-time instructional faculty was approximately 39% full professor, 28% associate professor, 24% assistant professor, 8% instructors and lecturers, and 1% other ranks. Figure 13 shows relative trends in number of faculty by rank (except for the category of instructors, lecturers, and other ranks, which behaved erratically over the time period and would not be expected to have a major impact on publications). Overall faculty and associate professor counts increased by 5% and full professor by 9%. Assistant professors increased until 1991, remained constant until 1993, decreased until 1998, and then increased in 1999, ending up about 2% lower than 1988. The number of full and associate professors combined increased by 7.2% from 1988 to 1999. This is substantially less than the increase in publication counts. Consequently, if they were the faculty most responsible for publications, their publication output per capita would also have increased substantially during this time period. Of course, these figures are subject to the same limitation as existed for total faculty counts (i.e., the assumption that S&E faculty counts behave the same as total faculty counts), and therefore must be regarded with some degree of skepticism.
 The sum of the whole counts over the 200 institutions would be larger than the number of publications because publications that have two or more institutional authors among the top 200 R&D performing academic institutions would be counted more than once.
 The total number of citations with the top 200 R&D performing academic institutions considered as a single entity was not available, so we have used the summation of citations as measured by whole counts over the top 200 R&D performing academic institutions. The latter quantity includes multiple counting of citations when there are two or more institutional authors among the top 200 R&D performing academic institutions.
 A decrease in efficiency (i.e., an increase in resources being used per publication) should not be interpreted as implying that the use of additional resources is unjustified. The resources required (by an "efficient" research team) to produce a publication may have changed over time.