Research and development expenditure data, in spite of their limitations, have long been the principal innovation input measure because they are available for an extended period for firms and industries. In the absence of direct output measures, these data frequently serve as a proxy for innovation output as well. The workshop discussions focused on problems in the collection and coverage of the R&D data and their integration with other databases. Some strengths and limitations of R&D expenditure data as an innovation input indicator are given in the text box.
R&D Expenditure Data as an Innovation Input Indicator
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 18
--> 5 Improving Information on Industrial R&D Research and development expenditure data, in spite of their limitations, have long been the principal innovation input measure because they are available for an extended period for firms and industries. In the absence of direct output measures, these data frequently serve as a proxy for innovation output as well. The workshop discussions focused on problems in the collection and coverage of the R&D data and their integration with other databases. Some strengths and limitations of R&D expenditure data as an innovation input indicator are given in the text box. R&D Expenditure Data as an Innovation Input Indicator Strengths Are publicly available from several sources and readily understood, Have been collected over time, enabling time series analysis, and Provide a consistent basis for comparisons across projects and industries. Limitations Do not capture design, product engineering, and technology service expenditures that are closely related innovation inputs, and Poorly represent technology development investments in small firms lacking distinct R&D laboratories or other units with separate functional and accounting identity.
OCR for page 18
--> Current Sources of Data on Industrial R&D The two principal public sources of data on U.S. industrial R&D expenditures are the National Science Foundation's Survey of Industrial Research and Development (Census form RD-1) and the 10-K financial reports that publicly traded U.S. firms file with the Securities and Exchange Commission (SEC). A database of the SEC filings by companies conducting R&D is maintained by Standard and Poor's CompuStat service. The public data sets differ in several respects. For example, the NSF survey but not CompuStat data includes privately held companies and foreign-based companies conducting R&D in the United States. CompuStat includes U.S.-based company R&D performed abroad, while the NSF survey does not.2 Other data on industrial R&D are collected by private firms and associations, research centers, and universities. Several examples were mentioned at the workshop. An annual R&D survey of Industrial Research Institute member companies has been carried out for four years in cooperation with the Center for Innovation Management Studies at Lehigh University under an NSF grant. The survey collects financial data about the source and allocation of R&D funds at the levels of the firm, the business segment, and the laboratory. The survey also requests information about the number and use of R&D personnel and measures of R&D performance at the firm and business unit levels. Wesley Cohen directed a onetime Carnegie Mellon University survey of manufacturing innovation in 1994. The Carnegie Mellon survey was administered in the United States and Japan and is designed to be compatible with parts of the European innovation survey. Cohen and his associates are in the process of analyzing the results. Until 1995, Business Week reported the results of its survey of 900 large companies in the "R&D Scoreboard." From 1982 to 1995, the Battelle Memorial Institute collected petroleum industry R&D information with detail by line of business and technology. Suggestions to Improve R&D Data Suggestions to improve national R&D survey data focused on business unit-level reporting, the integration of R&D data with other economic databases, and industry coverage.3 Targeting Surveys to the Appropriate Level Several participants suggested that NSF conduct its R&D survey at the level of the line of business of the firm rather than at the level of the enterprise as a whole. Especially in large, multifaceted, multiproduct firms, investments in technology development and commercialization correspond more closely to particular lines of business than to companywide operations. Furthermore, corporate managers at the unit level are more familiar with the level of spending and char-
OCR for page 18
--> acter of R&D and thus are likely to provide more accurate survey responses than are corporate headquarters officials. Types of data that should be collected at the business unit level include R&D expenditures, composition of R&D (process vs. product, basic vs. applied research vs. development), and share of R&D that is self-financed or supported by government or other contract, as well as contextual information on business unit sales, domestic and foreign, and growth history of the business unit. As a model for data collection at the business unit level, some participants suggested the Line of Business Reporting program instituted by the Federal Trade Commission in the late 1970s but discontinued after collecting data for the years 1974–1977. Although it yielded information at a level of detail appropriate for the analysis of research and innovation, the program attracted controversy as a function of a federal regulatory agency at a time when concerns about federal reporting requirements and the confidentiality of proprietary information were high. Expanding Industry Coverage In the opinion of several workshop participants, the collection and reporting of R&D data for emerging, rapidly technologically evolving industries is a priority. They endorsed recent efforts by the NSF and other agencies to extend their surveys to the service sector, biotechnology industry, and small research-intensive firms and urged further improvements. However, in the workshop session on information technology and the service sector, it was often noted that survey instruments designed for more traditional manufacturing industries may not be revealing of the character and processes of innovation in these industries and that further research on these processes might yield better questionnaires and thus better survey responses. Improving the Collection of R&D Data Workshop participants considered various steps to improve responsiveness or expand the information provided by responses to the NSF R&D survey. It was observed, for example, that the number of questions for which responses are mandatory has been limited and the number of voluntary questions reduced, partly in the interest of minimizing the reporting burden on companies. Currently, answers are required only to questions about sales, employment, total R&D spending, and government-supported versus self-financed R&D. Answers to questions about the character and orientation of R&D, a matter of concern to some science and technology analysts, are voluntary. One suggested compromise was to reduce the overall number of data items requested on NSF's RD-1 surveys while making more or all questions mandatory, although it was not mentioned what questions could be omitted. Discussion of the SEC 10-K filings elicited similar suggestions, for example, that reporting of R&D expenditures by business seg-
OCR for page 18
--> ment, which is currently voluntary, be made mandatory, although it was not mentioned what questions could be omitted. One participant recommended a thorough review of 10-K information requirements to identify ways of generating more and more useful information on innovative activity. Aggregating R&D Expenditures to the National Accounts: The Bureau of Economic Analysis R&D Satellite Account Total U.S. R&D spending, compared with other countries' expenditures, is often cited as an indicator of the nation's investment in new technologies and future competitiveness of its industries. To show the R&D share of gross domestic product, a measure of the intensity of the national innovative effort, and to incorporate the investment in R&D into the national income and product accounts (NIPAs), the Commerce Department's Bureau of Economic Analysis (BEA) in 1993–1994 developed a satellite R&D account.4 Using NSF's RD-1 and other data to construct this account, BEA (1) adjusted the R&D expenditures to make them consistent with the NIPAs, (2) treated R&D expenditures as a form of investment rather than only as an expense, (3) estimated real R&D investment, and (4) provided estimates of the stock of knowledge (R&D) capital. Steven Landefeld, BEA director, explained to the workshop how the R&D investment and capital stock time series could be used in studies of economic growth and productivity in lieu of other indirect measures of technological change. He noted, however, that the R&D satellite account was an exploratory, one-time effort and that there are several further steps that would be required to develop a complete satellite account consistent with the national accounts. He said that BEA would consider extending the project if there are sufficient policy-level and research interests, and he listed several possible next steps: Extending the account by changing national investment and capital stocks, reducing intermediate consumption by the amount of R&D investment, and including depreciation of R&D; Going beyond the using-up of capital as a measure of its output by developing estimates of returns to R&D capital over and above depreciation; Using new depreciation schedules comparable to those soon to be adopted for tangible capital; Using chain indices rather than fixed-weight indices to calculate R&D prices; Including R&D services in the calculation of input-output tables; Applying quality adjustments to the relevant prices for investment, stocks, and output, as BEA is doing in other areas;
OCR for page 18
--> Capitalizing computer software for inclusion in the national accounts and moving from there to other intangibles such as human capital, including R&D embodied in education; and Developing a model of the national accounts that incorporates nonmarket values, including household production. Any of these steps would require the commitment of additional resources, which in turn would depend on whether researchers and policy makers value the results relative to other data needs. The workshop did not explore the utility of extending the satellite account, but some participants suggested that BEA and NSF seek the opinions of analysts and public officials. Other Issues Raised It was noted that processing RD-1 data could be improved by updating hardware, software, and methodological practices, including automating sample selection, submitting forms to respondents, collecting forms from respondents, editing and auditing reported data, and applying data-nondisclosure procedures. Inconsistencies between RD-1 survey responses and 10-K information from the same firm are troublesome to many analysts. Although the sources of these discrepancies are not all known, it does not help that the reports are prepared by different units within firms. As a possible remedy it was suggested that the RD-1 survey present companies with the R&D numbers they reported in their 10-Ks and ask respondents to ensure that their disaggregated data are consistent with their companywide 10-K report. Other workshop participants expressed concern that in reporting data to the SEC, companies have incentives to enhance the value of their stock, and therefore the data may not be as reliable as the responses to Census surveys. Other issues regarding the NSF RD-1 survey were raised but not discussed in detail during the workshop. These included whether questions about the composition of R&D that have been removed from the survey should be reinstated or are too ambiguous to be useful (e.g., process vs. product and short-term vs. long-term R&D). It was also questioned whether R&D expenditures should be collected by product class and, if so, what product classification should be used. Some participants noted that the product field classification of industry detail in the NSF R&D survey is not very useful, as it corresponds neither to Standard Industrial Classification (SIC) codes nor to what the Federal Trade Commission (FTC) considers to be distinct markets. Furthermore, the level of detail of the product fields has been reduced. This is an important loss of information in, for example, the capital goods sector. It was suggested that these be reviewed.