The U.S. Corporate R&D data series supplements the NSF's SRS data series on U.S. industry R&D spending with timely information on both industrial sectors and individual firms. U.S. Corporate R&D is created under a joint agreement by U.S. Department of Commerce's OTP and the NSF's SRS. Based on Standard & Poor's Compustat database, the new data series features, for the years 1996 and 1997, the combined domestic and overseas R&D spending by the top 500 publicly held corporations headquartered in the United States. The data series provides not only individual firm and industry level R&D data, but also contextual data on net sales, capital spending, and employment.
The long-established SRS data series on U.S. industry R&D and the U.S. Corporate R&D data series are distinctly different, but serve complementary needs. The SRS data series on industrial R&D provides national estimates of the expenditures on R&D performed within the United States by industrial firms, whether U.S. or foreign-owned. Designed to capture all industrial R&D performed domestically, the data series includes, but does not distinguish between, privately held and publicly held firms. In providing the most comprehensive data on U.S. domestic R&D spending, the data series necessarily precludes information on individual firm activity, since it uses confidential firm data provided to the Bureau of Census. The data series selectively excludes the R&D spending of U.S. companies overseas, which is published in a separate SRS data series on U.S. foreign subsidiaries.
Among the SRS data series statistics are estimates of total R&D, the portion financed by the Federal Government, and the portion financed by the companies themselves or by other non-federal sources, such as state and local governments or other industrial firms under contracts or subcontracts. Total R&D is also separated into its character-of-work components: basic research, applied research, and development. Other statistics include R&D financed by a domestic firm but performed outside the United States, R&D contracted to organizations outside of the firm, and the funds spent to perform energy-related R&D. The series also provides statistics on domestic net sales, number of employees, number of R&D-performing scientists and engineers, and cost per R&D scientist and engineer.
These data are acquired through SRS's Survey of Industrial Research and Development. SRS has sponsored and managed surveys of industrial R&D since 1953. The content of these surveys has been expanded and refined over the years in response to an increasing need by policy makers for more detailed information on the nation's R&D effort. Beginning with the 1992 survey, the sample size was increased from approximately 14,000 to approximately 25,000 firms. This increase was made for several reasons: (1) to account better for births of R&D-performing firms in the survey universe; (2) to more fully and accurately survey R&D performed by nonmanufacturing firms, especially in the service sector; and (3) to gather more current information about potential R&D performers.
Because it is derived from Standard and Poor's Compustat, U.S. Corporate R&D meets a variety of analytic needs different from those provided by the Census-based SRS industry R&D data series. For example, tallies of latest year R&D spending data are available for firms very soon after the close of their fiscal year. By July of each year, Compustat compiles the latest R&D figures for a majority of the 9,800 active U.S. companies in its database. While U.S. Corporate R&D represents a smaller number of R&D firms than the SRS data series, it is sufficiently large and overlapping to help substantiate SRS R&D spending estimates issued earlier in the year for the most recently completed year. Furthermore, because Compustat provides data on other performance indicators for hundreds of individual firms (e.g., sales, employment, exports, foreign sales, and profits), the data series contains information that provides immediate context for analyses of industry R&D activity. Possible relationships between R&D spending and a wide variety of factors can be explored, the outcomes of which may be useful to researchers and policy makers. In addition, because it includes publicly held firms, U.S. Corporate R&D can provide R&D spending and other data for individual firms that often can be linked with other data series, thereby allowing for more detailed analysis of U.S. industrial R&D.
U.S. Corporate R&D is derived from Standard and Poor's Compustat database, which provides 20 years of annual and monthly data and 48 quarters of quarterly data for more than 18,500 U.S. and Canadian companies. Of these firms, more than 9,800 are active U.S. companies (of which more than 3,400 conduct R&D) and approximately 8,200 are no longer active U.S. companies (due to buyouts, bankruptcy, etc.). Compustat provides R&D data for corporations, and also provides other financial, statistical, and market data for corporations, banks, savings and loans, business segments, geographic areas, industry composites, aggregates, and indexes. It provides coverage of annual and quarterly Income Statement, Balance Sheet, Statement of Cash Flows, and supplemental data items on publicly-held companies.
Compustat company data are derived from publicly held companies, specifically those trading on the New York Stock Exchange, American Stock Exchange, National Association of Securities Dealers Automated Quotations (NASDAQ), Over-the Counter, three Canadian stock exchanges, and wholly owned subsidiaries of companies that are required to file with the SEC.
The convention used by Compustat to adjust for the fact that the fiscal years of individual U.S. firms vary, is to treat fiscal years ending January 1 through May 31 as ending in the prior calendar year. Thus, the data for a fiscal year beginning on June 1, 1997 and ending on May 31, 1998 are reported as the year 1997, whereas data for a fiscal year beginning on July 1, 1997 and ending on June 30, 1998, are reported as the year 1998.
Because Compustat draws its R&D data from the 10-K reports that corporations file with the SEC, it relies fundamentally on the SEC's definition of R&D. It defines R&D expenditure as all costs incurred during the year, by the company in question, that relate to the research and development of new products, processes, or services. Such expenditures generally include related software expenses and amortization of software costs. They generally exclude the following items: (1) customer or government-sponsored R&D (including reimbursable indirect costs); (2) extractive industry activities, such as prospecting, acquisition of mineral rights, drilling, mining, etc.; (3) those engineering expenses directed toward routine, ongoing efforts to define, enrich, or improve the qualities of existing products; (4) inventory royalties; and (5) marketing research and testing. R&D expense is not itemized with respect to banks, utilities, or property and casualty companies.
Compustat's SEC-based definition of R&D is the same as the SRS definition with two exceptions. First, unlike SRS, Compustat's SEC-based data does not exclude social science research. Second, following SEC rules, Compustat data does not include very small R&D amounts that are not material to a firm's decision-making. These differences in definition are unlikely to contribute to any significant differences.
Greater differences, however, may result from the fact that some firms consider certain routine engineering activities as qualified research for inclusion on SRS industry R&D questionnaires, but do not report such activity as R&D to the SEC, which disallows it.
The fact that Compustat only includes a given firm's own expenditure on R&D is significant. This means Compustat usually attributes R&D funds to the research-funding source, not the final performer of research when it is contracted out to another firm. As a result, reported levels of R&D by industry group are different from what they would be if they were based on the actual locations of final R&D activities. For example, if a manufacturing firm purchases the services of a research and testing firm, those expenses are generally reported under "manufacturing R&D" rather than "services-sector R&D," although the latter would better reflect where R&D is actually performed.
For each firm Compustat provides only a single annual R&D expense total. Thus, unlike some other data in Compustat, such as net sales, data on separate R&D expenses for a single firm are not available for R&D performed in different geographic regions, such as in the United States or performed abroad by U.S.-owned subsidiaries. Only a single, combined R&D total is provided for each firm.
Table 3 (page 25) and figure 14 compare 1997 R&D spending data from the U.S. Corporate R&D and SRS industry survey. Differences in the R&D spending of various industrial sectors are attributable to the smaller universe of firms in U.S. Corporate R&D, as well as, the structural differences between the two data series discussed in the text box on the previous page.
The U.S. Corporate R&D total of $111 billion for the top 500-R&D spenders in 1997 is 83.4 percent of the SRS industry survey estimate of $134 billion-about what might be expected given differences in the two series. According to the SRS industry survey, R&D increased by 10.4 percent over the survey's $121 billion total, in comparison to a 9.0-percent increase according to the U.S. Corporate R&D data series. U.S. Corporate R&D aggregates for the following four major industries also compare reasonably to corresponding SRS survey aggregates: aircraft, guided missiles, and space vehicles (82.3 percent of SRS data); chemicals (96.9 percent); information and electronics (89.9 percent); and machinery and electrical equipment (70.1 percent).
Nevertheless, within information and electronics, the U.S. Corporate R&D and the SRS industry survey assign quite different R&D spending amounts for various detailed industries. For example, according to U.S. Corporate R&D, the five computer-related hardware detailed sectors together spent a total of $16.9 billion, compared to the SRS survey's $12.8 billion. Conversely, U.S. Corporate R&D attributes $6.6 billion to electronic components, while SRS assigns $10.8 billion. Such differences between the two series within information and electronics may be due largely to differences in how Compustat and the U.S. Census Bureau assign Standard Industrial Codes to firms.
U.S. Corporate R&D attributes significantly higher amounts of R&D spending than SRS to two major sectors, motor vehicles and surface transportation and the large "substances" portion of medical substances and devices-respectively, 130.7 and 153.9 percent of the SRS amount. One possible explanation for these differences may be that U.S. firms in these industries spend greater amount for R&D abroad than do their foreign counterparts in the United States. Another explanation may obtain from the fact that SRS assigns R&D spending to the R&D performer (e.g. contractors), whereas, U.S. Corporate R&D attributes the R&D spending to the funding source. Accordingly, significant amounts of motor vehicle research may be under contract to some of the many firms in other sectors that provide motor vehicle subcomponents. In the case of medical substances, the SRS survey may view a sizable amount of that industry's R&D as occurring under contract in testing/research services. This "transfer" of R&D funds may also partly explain why U.S. Corporate R&D attributes a much smaller portion of R&D to testing/research services than does SRS (in addition to the fact that many testing and research service firms may be too small for inclusion in top 500 U.S. Corporate R&D firms).
The significantly smaller amount of R&D spending that U.S. Corporate R&D attributes to the basic industries and materials sector may be partly because SRS includes many more smaller and private firms. This same factor may contribute also to the very significant lower U.S. Corporate R&D amounts attributed to the finance, insurance, real estate, and general services sectors. Additionally, the amount of R&D attributed by U.S. Corporate R&D to these last two sectors is reduced by the fact that Compustat does not report the R&D expenditures of banks, utilities, and property and casualty companies.
Based on their latest annual R&D spending, U.S. Corporate R&D identifies the top 500 R&D-conducting, publicly held corporations that are headquartered in the states or territories of the United States. The data series excludes the R&D of U.S. subsidiaries owned by corporations headquartered in foreign countries. The top 500-R&D firms in U.S. Corporate R&D account for nearly 90 percent of all the R&D spending reported by Compustat from more than 3,000 firms. Because the list of top 500-R&D corporations changes from year to year, U.S. Corporate R&D uses two distinct 1996 and 1997 lists of top 500 firms for generating and comparing aggregate 1996 and 1997 activity.
Annual changes in the corporate composition and overall R&D content of the top 500 firms in U.S. Corporate R&D may be attributed to a variety of factors, such as: (1) firms increasing or decreasing their ordinary and usual R&D expenses as necessary; (2) firms merging or spinning off (such as Lucent from AT&T); (3) U.S.-headquartered firms acquiring other U.S.-headquartered firms, private firms, or foreign firms; (4) private U.S. firms becoming publicly held; and (5) foreign headquartered firms acquiring U.S.-headquartered firms, in which case the latter's R&D is no longer tracked.
U.S. Corporate R&D categorizes individual firms and relates R&D data according to eight "major" industrial sectors and 45 "detailed" industrial sectors. The detailed sectors are similar, but not identical to sectors featured by the SRS data series. The eight major sectors combine detailed sectors into some new, large industrial groups, one of which, information and electronics, includes both manufacturers and services (communications and software). By categorizing many multifaceted firms into broad industrial categories, these large sectors minimize the misleading effects of labeling firms by a narrower Standard Industrial Code (SIC) and the movement of R&D from one narrow category to another when companies acquire (or spin-off) companies into a different sector. Seven of these eight major sectors conduct significant amounts of R&D and/or include significant numbers of R&D firms among the top 500-R&D spending corporations. The remaining sector-general services; engineering, accounting, and research testing services; finance, insurance and real estate-contains only 10 of the top 500 1997 R&D firms in the Compustat, accounting for less than 0.5 percent R&D performed by the top 500.
While generally reflective of the activity or performance in a given sector, sectoral aggregates (including R&D, sales, and employment) should be used with caution. Because U.S. Corporate R&D data are tabulated at the enterprise level, all the R&D or other activity pertaining to a single firm are attributed to the major (most relevant) SIC of that firm. This means all the R&D of a given firm is classified under one SIC regardless of how many other SICs may apply to various firm activities and subdivisions. Industrial sector aggregates thus contain activities of many firms and subdivisions that logically should, but cannot be, attributed to other sectors.
Another reason for caution in using sectoral aggregates is that, from year to year, aggregates may rise or fall not only due to trends in the industries themselves, but also due to the acquisition of firms or subsidiaries by other firms. When such acquisitions occur within a single industrial sector, they do not affect year-to-year aggregates. However, aggregates may be significantly affected when major mergers or acquisitions cross industrial sectors. For example, the manufacturing sector was enlarged at the expense of the service sector when IBM acquired Lotus in 1995. Similarly, the spin-off of Lucent from AT&T in 1995 reassigned very significant amounts of R&D from the communications services sector to the communications equipment sector.
The U.S. Corporate R&D data series aims at using a firm's internal, current dollar expenditures to measure actual performance of R&D work (i.e., scientists and engineers doing work on research and development projects). However, when one firm acquires another, Compustat generally combines the acquiring firm's internal R&D expenditure with write-offs of purchased in-process R&D (IPR&D) and reports these as a single R&D expense. This practice presents difficulties for the U.S. Corporate R&D data series. This is because it mixes together acquired R&D, which is often measured according to some estimate of its future value, with R&D that is measured strictly by the current expenditure of dollars.
In order to eliminate potential overstatement of actual R&D expenses, the U.S. Corporate R&D series adjusts Compustat-reported R&D expenses when IPR&D write-offs do not represent actual dollars spent on R&D in the current year. Firms that may have such write-offs are identified by a combination of filters that look for unusual increases in R&D from year to year, or significant differences in annual percent changes between R&D and net sales, or R&D and employment. The SEC submissions of identified firms are then examined to determine whether an unusual R&D increase is due to either an exceptionally large increase in actual R&D performance or to the write-off of purchased IPR&D.
These screening and adjusting methodologies are not perfect. Some firms are examined based on past activity or other information. However, not all smaller firms are examined and the modest percentage increases of larger firms may contain relatively small IPR&D write-offs. Also, firms that may in the same year write off large amounts of IPR&D and significantly decrease their spending on actual R&D may not be identified, thereby leaving their R&D expenses subsequently unadjusted.
 While Compustat provides data on a limited number of foreign firms, another comparable Standard and Poor's database, Global Vantage, is the primary source of information for these firms.
 Given the latitude that exists in firms reporting such information, this may not always be the case, e.g., some firms probably do report federally supported R&D as their own, especially when such support is provided after the R&D was actually performed, as in the case of Federal support for independent R&D.
 Bronwyn Hall and William F. Long, "Difference in Reported R&D Data on the NSF/Census RD-1 Form and the SEC 10-K Form: A Micro-Data Investigation," unpublished report to the National Science Foundation, April 1998.
 The list of 500 companies is drawn from Compustat data reported on July 31, 1998. While Compustat reports latest year data for nearly all large firms by July 31, Compustat reports a significant number of smaller firms over the remainder of the calendar year. Some of these late firms may have R&D spending levels equal to or greater than some firms contained in the July 31 list of top 500 R&D spenders.
 While the detailed industrial sectors of U.S. Corporate R&D and the SRS data series are similar, differences exist in the method of classifying firms by category. Most importantly, in some cases the same firm may be grouped under different industrial categories by the two data series, thereby limiting the comparability of the two series in terms of industry aggregates.
 This distinction arises from existing accounting rules and related policy questions regarding the appropriate valuation of purchased in-process R&D (IPR&D), and the consistency of accounting treatment between IPR&D and an acquiring firm's own internal R&D expenditure. According to 1974 rules of the Financial Accounting Standards Board (FASB), R&D conducted internally should be expensed rather than capitalized (FASB Statement No. 2: Accounting for Research and Development Costs). In a later interpretation, the Board additionally determined that certain forms of R&D that might transfer in a corporate acquisition could also be expensed, including "even a specific research project in process." The extent of the write-off should be determined "from the amount paid by the acquiring enterprise and not from the original cost to the acquired enterprise." The amounts allocated to IPR&D are expensed at the date of consummation of the acquisition, unless the projects have an alternative future use.
In the last several years, amounts attributed to purchased IPR&D have increased, raising questions as to whether the valuations are measuring the fair value of the IPR&D. By increasing IPR&D, acquiring firms reduce their reported levels of assets attributable to goodwill and other intangibles, and thereby improve their returns on equity. While the write-offs reduce net earnings in the year of purchase, they also bolster future earnings, an important gauge to investors of a firm's health. Such write-offs, which are intended to represent the fair value of the acquired in-process technology, may be significantly larger than the amount originally spent on R&D by the acquired firm.
Under generally accepted accounting principles (GAAP), goodwill is the remaining value of a company's purchase price after accounting for acquired tangible and identified intangible assets such as IPR&D. Generally, firms wish to avoid recording large amounts of goodwill since it is amortized over its useful life, using periods as long as 40 years, during which earnings per share are reduced annually. Also, most securities analysts subtract goodwill from equity when examining a firm debt-to-equity position.
In a number of high-profile acquisitions, the purchasing firms have written off significant portions of acquisition cost as IPR&D. For example, in 1995 when IBM purchased Lotus Development Corporation it valued the acquired R&D as $1.800 billion, increasing IBM's total Compustat reported R&D from $3.382 billion in 1994 to $5.227 billion. Prior to its acquisition, Lotus reported an R&D expenditure of only $256 million. Similarly, in 1994 and 1995, acquisitions by Computer Associates resulted in the write off of significant portions of acquisition costs as IPR&D, with the result that its Compustat reported total R&D increased from $226 million in 1993 to $504 million in 1994, and to $1.607 billion in 1995. The practice is not limited to the purchase of software companies; any company with products under development may have IPR&D to be written off.
As more companies in recent years have written off acquired IPR&D, the Securities & Exchange Commission (SEC) has increased its scrutiny of company IPR&D valuations, and in several cases has forced companies to reduce their write-offs. In addition, the FASB has indicated that it will examine the appropriate accounting treatment for both in-house and acquired R&D.