Trends in Federal Support of Research and Graduate Education (2001)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Appendix Note on Sources of Data This report relies primarily on a series of annual sur- for “development” by research field because of the diffi- veys conducted by the National Science Foundation’s culty of categorizing highly applied work that may draw Division of Science Resources Studies (NSF/SRS). The on many disciplines (for example, in the development of a following descriptions of these surveys emphasize the military aircraft). information collected that is relevant to the subject of this Uncharacterized research. The agencies also report report. Privately collected data on charitable foundation funding of research “not elsewhere classified” or “n.e.c.” giving supplement these surveys. for each broad field (i.e., life sciences, n.e.c.; engineering, n.e.c.) and for research that cannot be attributed to any broad field (i.e., other sciences, n.e.c.). These may be SURVEY OF FEDERAL FUNDS FOR RESEARCH AND multidisciplinary or interdisciplinary projects that do not DEVELOPMENT fall within any of the broad fields or subfields. The n.e.c. categories represent the majority of research in the social (http://www.nsf.gov/sbe/srs/sffrd/start.htm) sciences and psychology (60.4 percent and 86.6 percent, The Federal Funds Survey is conducted annually and respectively) and are quite large in engineering (28.3 includes retrospective reports of 32 federal agencies and percent) and environmental sciences (19.4 percent), but their major subdivisions on obligations actually incurred they are much smaller in other major fields. They have during the past fiscal year and those expected to be in- been growing in most major fields but not very rapidly. All curred during the current and following fiscal years. The n.e.c. research combined was 15.5 percent in 1993 and information is collected in a number of categories (i.e., 16.8 percent in 1999. research field, character of research, and performer) that Two-year projections. The survey asks agencies to can be cross-tabulated in useful ways for an analysis of estimate future research allocations for the current and trends in federal research spending. Obligations are com- next fiscal years for the seven broad fields but not for the mitments to spend money, regardless of when the funds 26 subfields. For the fiscal year in which the survey is were appropriated and of whether actual payment may be administered agencies’ estimates reflect enacted appropria- made later, for example, under multiyear contracts. In most tions levels and are presumably fairly realistic. For ex- cases respondents are agency budget analysts who consult ample, the latest survey was conducted between February with R&D program managers in their agencies. and September 2000, when agencies were in the middle or Character and field of research. Obligations for re- late stages of obligating the FY 2000 budget and had a search are classified as basic research or applied research good idea of what the final amounts would be. For the in seven broad research fields and in 26 narrower natural subsequent fiscal year (FY 2001 in the case of the latest and social science and engineering fields, called subfields survey reported), the estimates reflect probable administra- in this report. Although the nomenclature corresponds to tion budget request levels, which frequently differ from academic disciplines and departments, in the Federal actual appropriations. In the case of the National Institutes Funds Survey it also is applied to research performed in of Health, for example, congressional appropriations have industry and government laboratories and by other non- consistently exceeded executive branch requests. profit institutions. The field categorization used by the Performers. Agencies report how much they obligate Federal Funds Survey is further discussed in Chapter II. for basic and applied research by performer (i.e., industry, NSF/SRS does not try to collect information on obligations universities, nonprofit institutions, and in-house and 93 

94 APPENDIX contract laboratories) but with one exception do not sepa- affects of the 1996 changes in NSF classification criteria rately report these figures by field of science and engineer- affecting reported obligations in certain fields. ing. The six largest R&D supporting agencies report research performed by universities and colleges by field and also by whether it is basic or applied research. How- SURVEY OF RESEARCH AND DEVELOPMENT ever, research performed in industry or government labora- EXPENDITURES AT UNIVERSITIES AND COLLEGES tories is not reported by field. Anomalies. Unexplained discontinuities in the data (http://www.nsf.gov/sbe/srs/sseeuc/start.htm) usually prompt NSF/SRS or its contractor to inquire about The university R&D expenditures survey has collected the source of anomaly, which may be a genuine change in data annually from a sample of education institutions that priorities or a reclassification. Beginning with FY 1996, grant science and engineering degrees and perform a the reporting unit for NSF, for example, changed its minimum level of separately budgeted R&D. Three years procedures for classifying research obligations by field. ago the survey was expanded to a census of all research The change most affected engineering, where the amount universities. Respondents are usually central administrative classified as “n.e.c.” by NSF jumped from 20 percent to 40 staff who consult with their academic departments that are percent of engineering research, while mechanical engi- receiving support. neering dropped from 13 percent to 2 percent of NSF Character and field of research. The major fields and engineering research funding. Because NSF is the second subfield classifications used in this survey are virtually the largest funder of mechanical engineering (after the Depart- same as those used in the Federal Funds Survey, but ment of Defense), the total amount of funds for mechanical institutions are not asked to disaggregate R&D into basic, engineering appeared to drop by 30.1 percent from FY applied, and development categories. Thus, funds for 1995 to FY 1996, although almost certainly any shift in research and funds for development cannot be distin- NSF’s portfolio was not as significant. A similar shift into guished as they are in the Federal Funds Survey. the n.e.c. category occurred in the classification of physical Unclassified research. Academic respondents may science research, which especially affected physics. The avail themselves of the same “not elsewhere classified” NSF-reported increase in oceanography research and drop categories in each major field and overall. “N.e.c” research in geology research from 1995 to 1996 may also have is highest in engineering (22 percent in a recent survey), involved reclassification of some activities from one lowest in the life sciences (4 percent), and about 10 per- category to another rather than a real change. cent overall, or lower than reported in the Federal Funds Features of our analysis. In this report, the current Survey. dollar expenditure data from the Federal Funds Survey are Sources of support. NSF asks for total expenditures by converted to constant 1999 dollars, using the official field and federal expenditures by field. This enables one to deflators issued by the Office of Management and Budget derive data on nonfederal sources of support but not to (OMB) in February 2001.1 Subfields of the social sciences disaggregate these data further, for example, to determine and psychology are not considered in this report because of how state government or industry or philanthropic support the high proportion of research in both major fields re- varies by field and over time. ported as “n.e.c.” “Engineering, n.e.c.,” called other engi- Discrepancies between the surveys. The university neering in this report is included in the analysis because it survey results show higher levels of support for academic encompasses a series of discrete fields (e.g. biomedical, research overall and in most fields than do the Federal nuclear, etc.) constituting a significant part of the disci- Funds Survey results. For example, federal obligations for pline rather than a residual category for unclassifiable or academic electrical engineering research declined in multidisciplinary research. Accordingly, our analysis is constant dollars by more than 30 percent between 1993 based on trends in 22 subfields, counting social sciences and 1997. But according to the academic R&D survey, and psychology as subfields. Most of the analysis is based federally funded R&D expenditures in that field increased on actual obligations through FY 1999, the last year for which subfield data are available. In some instances, trends in broad fields are reported through FY 2000, because of 2National Research Council. 2000. Measuring the Science and the reliability of the estimations for that year in contrast to Engineering Enterprise: Priorities for the Division of Science Resources the estimations for FY 2001. In describing trends in the Studies, pp. 94-95. Washington, D.C.: National Academy Press. The affected fields the report tries to take into account the National Academies’ report observes that the gap between federal obligations to universities and the level of federally funded R&D expenditures by universities opened in 1992 and has been growing, to about $1.9 billion in 1997. A much larger discrepancy exists between the 1Office of Management and Budget. 2001. Historical Tables, Budget results of the Federal Funds Survey and NSF’s Survey of Industrial R&D of the United States Government: Fiscal Year 2002, Table 10.1. Washing- with respect to the size of federally funded R&D expenditures in industry ton, D.C.: U.S. Government Printing Office. (p. 47). 

APPENDIX 95 by 27.2 percent.2 The omission of development funds from completed for each department in each responding unit at the Federal Funds data on university research and the an institution either centrally or by departmental staff. inclusion of development funds in the data reported by Variables. Data variables drawn on for our analysis academic institutions account for some of the difference include enrollment status, postdoctorate status, primary between the results of the two surveys, especially in fields source of support, mechanism of support, and field of such as engineering where development comprises about study. 60 percent of academic R&D compared with 13 percent of Field of study. The major fields and subfield classifica- academic R&D in science. A second and possibly larger tions are in some respects more detailed than those in the contributor to the difference is that the academic R&D Federal Funds and Academic R&D Expenditures Surveys. survey counts twice some funds that derive from the For example, there are 14 instead of 7 engineering fields, federal government but are transferred from one institution 17 biological sciences fields, and 23 “health” fields. In two to another.3 Third, obligations (reported in the Federal cases, subfields in the Federal Funds Survey are combined Funds Survey) differ from actual expenditures (reported in in the GSPSE data. “Aerospace engineering” indicates the academic R&D survey) from year to year. Fourth, both aeronautical and astronautical engineering. “Biologi- Federal Funds Survey respondents tend to assign more cal sciences” incorporates environmental biology. research to n.e.c. categories than do academic respondents, Unclassified students. In GSPSE there are departments perhaps in part because it is easier for institutions to that are classifiable by broad, but not fine, field. Thus there identify departments receiving the funds than for federal are residual categories such as “physical sciences (and officials to assign a field of research. biosciences, psychology, engineering, etc.), other.” These Features of our analysis. Because of double counting residual categories comprise from less than 1 percent to of federal dollars, we rely on the Federal Funds Survey more than 25 percent of their broad field category. data to analyze the field distribution of government sup- Primary source and mechanism of support. Respon- port. Nevertheless, the Academic R&D Expenditures dents are asked to identify the primary source of support Survey provides the only data on university funds by for each graduate student and postdoctorate. For graduate research field from other sources (e.g., state government students, sources of support include federal agencies (e.g., and industry), and the latter are cited in Chapter 5. Department of Defense, National Institutes of Health, other HHS, National Science Foundation, Department of Agri- culture, NASA, etc.) and nonfederal sources (institutional SURVEY OF GRADUATE STUDENTS AND support, self support, other U.S. and other foreign). GSPSE POSTDOCTORATES IN SCIENCE AND ENGINEERING also asks respondents to cross-tabulate source of support (GSPSE) against the following mechanisms of support: graduate fellowships, graduate traineeships, graduate research (http://www.nsf.gov/sbe/srs/sgss/start.htm) assistantships, graduate teaching assistantships, and other types of support. The graduate student survey collects data on the number Features of our analysis. In this report we have aggre- and characteristics of graduate science and engineering gated subfields in the GSPSE to arrive at a classification students and postdoctorates enrolled in U.S. institutions closely corresponding to that of the R&D data. We use the offering postbaccalaureate programs in science and engi- GSPSE category of “Health Fields” as most comparable to neering. The final 1998 survey universe consisted of 722 the Federal Funds category of “Medical Sciences.” responding units at 601 master’s- and doctorate-granting institutions in the United States. Survey questionnaires are SURVEY OF EARNED DOCTORATES (SED) 3Reporting on a workshop requested by the Senate Commerce, Science, and Transportation Committee, the Congressional Research (http://www.nsf.gov/sbe/srs/ssed/start.htm) Service of the Library of Congress concluded: There is general agreement among the workshop participants that the The doctoral survey is a census of individuals receiving source of the discrepancy is a result of a changing research environ- research doctorate degrees from U.S. institutions since ment—such as more cooperative research ventures under which a significant fraction of the funds received by universities can be counted 1958. Graduate schools are responsible for collecting twice. As funds get transferred between different academic institutions questionnaires from doctoral recipients and submitting involved in joint research efforts or between different parts of the same them to be compiled in the Doctorate Records File (DRF), institution, tracking the funds becomes more difficult and can lead to which maintains data on the number and characteristics of double counting….The (federal) agencies believe that they can accurately all recipients since 1958. The population for the 1999 account for R&D funds obligated to universities. Michael E. Davey and Richard E. Rowberg. January 31, 2000. survey consisted of all individuals receiving a first re- Challenges in Collecting and Reporting Federal Research and Develop- search doctorate from a U.S. academic institution in the ment Data, p. 17. Washington, D.C.: Congressional Research Service. 12-month period ending on June 30, 1999. The total 

96 APPENDIX universe consisted of 41,140 persons in approximately 392 but not subfields, to characterize research as basic, applied, institutions. or development, and to further distinguish research by Variables. The DRF contains a wide range of data on budget function such as health, agriculture, economic the demographic characteristics, citizenship status, educa- development, and support of the science and technology tional history, field of study, financial support, and planned infrastructure. Thus, respondents reported the share of employment of doctorate recipients. engineering research directed at economic development Field of study. Major fields and subfields in the SED (20 percent) and the share of biological research directed at are roughly comparable to those of the Federal Funds agricultural production (35 percent). Survey with one exception. The SED collects data on fine Uncharacterized research. State respondents were able fields within the major field of earth, atmospheric, and to employ a “not elsewhere classified” category, overall ocean sciences. However, published tables on citizenship representing about 21.1 percent of research reported in the status do not disaggregate this major field. 1995 survey. Citizenship status. The DRF provides five citizenship categories: native-born U.S. citizen, naturalized U.S. FOUNDATION GIVING TRENDS citizen, permanent resident, temporary visa holder, and unknown citizenship. Trends in permanent and temporary The private nonprofit Foundation Center compiles visa holders can be divergent, as they were in the 1990s, reports from a sample of approximately 1,000 large inde- for idiosyncratic political reasons. The Chinese Student pendent, corporate, and community foundations, most Protection Act of 1992, which allowed Chinese students in recently in 2000. The sample was expanded in 1991 to the U.S. at the time of the Tianamen Massacre to become include more smaller and corporate foundations, introduc- permanent residents, generated a dramatic shift in Chinese ing a discontinuity in the data series. Excluded from the Ph.D.’s away from temporary to permanent visa status. sample are grants by wealthy individuals and expenditures The overall trend in foreign students’ enrollment in doctor- of private research institutes (e.g., Howard Hughes Medi- ate programs is best observed by looking at the larger cal Institute) whether or not established by philanthropic category of non-U.S. citizens. The citizenship variable has bequests. also been clouded in recent years by a significant increase Classification of grants. Expenditures are classified by in the “citizenship unknown” category.4 function such as art and culture, education, and health. Features of our analysis. For cross-survey comparison, “Science and technology” is further distinguished as the component fields of geology, atmospheric sciences, general science, physical science, technology, life science, and oceanography in data from the Federal Funds Survey and “other;” but general science includes grants to improve and the Survey of Graduate Students and Postdoctorates science education below the college level. “Medical re- have been aggregated to match the DRF data. We also search” is a subcategory of health rather than science and aggregate the native born and naturalized citizens into one technology. In general, interpretation of the data is compli- U.S. citizen category and temporary visa holders and cated by the fact the same expenditures may be reported permanent residents into one non-U.S. citizen category. under two or several categories. In the case of the physical sciences, year to year fluctuations are common and attrib- utable to the fact a number of large foundation donors SURVEY OF STATE RESEARCH AND DEVELOPMENT (e.g., Keck, Kellogg) primarily fund large capital projects. EXPENDITURES: FY 1995 The state R&D survey was a one-time NSF/SRS- SURVEY OF INDUSTRIAL RESEARCH AND sponsored survey conducted and published by the State DEVELOPMENT Science and Technology Institute of the Battelle Memorial Institute in Columbus, Ohio (STTI/BMI). The survey (http://www.nsf.gov/sbe/srs/srs01410/start.htm) identified only directly targeted R&D funds, not general education purpose funds used by universities for research NSF/SRS has sponsored and the Bureau of the Census expenses or to cover the indirect costs of research. Previ- has conducted an annual industry survey since 1953. The ous intermittent NSF surveys of state agency R&D ex- survey is directed to the central corporate headquarters of cluded funds directly appropriated to academic institutions, U.S.-based public and privately held, U.S.- and foreign- making comparisons over time difficult if not impossible. owned corporations and asks them to report separately Character and field of research. The SSTI/BMI survey annual corporate domestic spending on R&D regardless of asked respondents to categorize research into major fields business unit or product or service lines, together with sales, employment, numbers of employed scientists and 4A. Sanderson and B. Dugoni. Summary Report 1997: Doctorate engineers, and cost of R&D per scientist/engineer. Respon- Recipients from United States Universities, p. 18. Chicago: National dents are asked to identify funds from the federal govern- Opinion Research Center. ment, the principal source of non-self-financed R&D. But 

APPENDIX 97 R&D performed in-house is not distinguished from con- COMPUSTAT R&D DATABASE tract R&D or grants to non-profit institutions. As a Census An alternative source of industrial spending data is the Bureau survey the identity of respondents is strictly confi- the Standard and Poor’s Compustat database, which con- dential. Data are not reported where their publication tains information required to be reported annually to the might reveal the identity of the single or a small number of U.S. Securities and Exchange Commission (SEC) on so- respondents. called l0K forms by publicly held firms headquartered in Character of research. Activity is classified as basic the United States. Banks, utilities, and insurance compa- research, applied research, or development according to nies are not required to report R&D expenditures. In standard definitions that are uniform across NSF/SRS addition to domestic and foreign R&D spending, the R&D surveys. information includes other characteristics such as corporate Classification. Respondents and responses are classified sales, employment, exports, foreign sales, profits, and according to the U.S. Standard Industry Classification capital investment. Because the data are mandatory and (SIC) code at the three- and four-digit levels. From one public, they are available more quickly than NSF survey survey to another, a firm may shift from one classification results and the respondents can be identified. On the other to another either because of acquisition or because of a hand, basic and applied research and development are not shift in business emphasis (e.g., computer manufacturing distinguished and therefore the data are not used in this to computer services), but all of its R&D falls in a single analysis.5 classification regardless of how diversified its business operations. Because of the confidentiality requirement, reclassifications of firms are not made public. In 1999 firms were classified under a new scheme, the North American Industry Classification System (NICS). To avoid 5For a Compustat-based analysis of the distribution of R&D by field a sharp discontinuity in the data series, NSF/SRS reclassi- see Carl Shepherd and Steven Payson. 1999. U.S. Corporate R&D Vol. I: fied the data from the previous two calendar years, 1997 Top 500 Firms in R&D by Industry Category. Washington, D.C.: U.S. and 1998, but cannot assure users that the adjusted data are Department of Commerce and National Science Foundation. This strictly comparable to the 1999 results. analysis is being updated.