Users of the survey results typically consider the information to be straightforward, accurate, and complete. However, none of these descriptors is quite the case. Federal agencies that report their R&D spending to NSF treat the surveys with differing degrees of attention to timeliness and accuracy. Some agencies periodically change their internal classifications and the ways in which they account for R&D spending, so the data have problems at their source.1 Even more problematically, the surveys ask for information in categories that are not used by all agencies for their own internal purposes, so the information provided to NSF is often a rough estimate, frequently based on unexamined assumptions that originated years earlier.
A key component of the reporting of federal R&D spending is the fields of S&E taxonomy, and Macro International recently conducted a study for NSF of its use by federal agencies (Macro International, 2008, p. 5). The study found that some of the major R&D agencies estimate spending by fields of science using staff judgment calls, rules of thumb, percentage distributions, or mapping of the codes to the agency’s plans or organizational structure. Some of the agency decisions made in these ways are rather arbitrary. For example, for purposes of federal funds survey reporting, the Centers for Disease Control and Prevention reports its entire research portfolio under one category, Life Sciences–Medical Sciences (Macro International, 2008, p. 7).
With these issues and others in mind, NSF asked the Committee on National Statistics of the National Research Council to review the two main surveys that are used to collect data on federal R&D spending and to consider ways to improve their accuracy and timeliness. Accordingly, the Panel on Modernizing the Infrastructure of the National Science Foundation Federal Funds Survey was established to consider the uses of the NSF federal R&D spending data and, in view of those uses, the quality of the data on federal funds for research and development and to recommend future directions for the program. The panel was asked to include the fields of science classification structure underlying the Survey of Federal Funds for Research and Development in its review.
In approaching this task, the panel has reached out to both senior officials of federal agencies that provide the federal funds data and key data users, solicited advice from providers of complementary and competing data sources, and reviewed past studies on federal funds data. This report, with recommendations on modernizing the infrastructure of the survey, is the primary product of the study.
For example, in fiscal year (FY) 2000, the National Aeronautics and Space Administration reclassified as research activities that had previously been classified as development, and in FY 2004 the agency implemented a new budget approach. Both actions introduced major discontinuities in the R&D data series (National Science Foundation, 2008a, p. 6).