Since the early 1980s, R&D performance in some sectors has grown much faster than in others. The industrial sector in particular has grown increasingly dominant (figure 1). In 1980, industry performed 68.4 percent of the nation's R&D, the academic sector performed 10.2 percent, laboratories within Federal agencies (Federal intramural R&D) performed 12.4 percent, and the nonprofit sector performed 2.6 percent. All FFRDCs combined performed 6.4 percent of the nation's R&D. Industry's defense-related R&D efforts accelerated in the early 1980s, and its share of R&D performance rose to 71.8 percent in 1985.
From 1985 to 1994, R&D performance grew by only 1.4 percent per year in real terms for all sectors combined. This growth was not evenly balanced across performing sectors, however. R&D performance at universities and colleges grew by 5.4 percent per year in real terms, compared with only 1.0 percent for industry, 0.5 percent for Federal intramural performance, 5.0 percent for nonprofit organizations, and 0.3 percent for all FFRDCs combined.
These growth rates changed dramatically from 1994 to 2000. Total R&D performance in real terms averaged 5.8 percent growth per year, which was substantially higher than in the earlier sluggish period. Yet, R&D performance at universities and colleges grew at a slower rate of 4.1 percent per year in real terms. Industrial R&D expanded at a rate of 7.1 percent in real terms (despite a decline in company-reported Federal financing of R&D). Federal intramural performance decreased by 0.3 percent per year in real terms. Nonprofit organizations, according to current estimates, increased their R&D performance by 7.1 percent per year in real terms over the same 6-year period. Finally, R&D performance at all FFRDCs experienced essentially no change in real terms over this period.
Industry is projected to have performed 68.3 percent of the nation's total R&D in 2003 (table 1). The estimated $193.7 billion in industrial R&D performance represents a 2.2 percent average annual decrease in real terms from the 2000 level. Universities and colleges are projected to have performed 14.2 percent ($40.3 billion) of national R&D in 2003, representing an average annual increase of 7.9 percent in real terms over their R&D in 2000. The Federal Government is projected to have performed 8.8 percent ($25.0 billion) of U.S. R&D in 2003, an average annual increase in real terms of 9.9 percent over the 20002003 period. All FFRDCs combined performed a projected $12.2 billion of R&D in 2003, or 4.3 percent of the U.S. total. The nonprofit sector performed a projected $12.7 billion in 2003, or 4.5 percent of the U.S. total.
As previously described, R&D performed by private industry reached a projected $193.7 billion in 2003 (table 1). This total represents a 2.2 percent average annual decline in real terms from the 2000 level of $197.5 billion. Most of this decline was in industry-financed R&D. Companies funded 91.1 percent ($176.4 billion) of their 2003 R&D performance, with the Federal Government funding nearly all the rest ($17.3 billion, or 8.9 percent, of the total). The federally funded share of industry's R&D performance total has fallen considerably from its most recent peak of 31.9 percent in 1987. For more than a decade the largest component of R&D in the United States has been performed by private industry through private industry's own funds. This component of U.S. R&D grew from 43 percent of total U.S. R&D in 1953 to a projected 62 percent in 2003.
Until the 1980s, little attention was paid to R&D performed by nonmanufacturing companies largely because R&D activity in this sector was relatively small compared with the R&D operations of manufacturing companies. Before 1983, nonmanufacturing industries accounted for less than 5 percent of total industrial R&D performance (including industry-administered FFRDCs), but by 2001 (the most current year for detailed data on industrial R&D), they accounted for 39.2 percent. In 2001, firms classified in nonmanufacturing industries performed $77.8 billion of R&D ($72.4 billion in funds provided by companies and other non-Federal sources and $5.4 billion in Federal support) (table 5). Of this amount, 79 percent ($56.9 billion) can be attributed to the following three groups of nonmanufacturing industries: trade, software and computer-related services, and scientific R&D services. An examination of these three groups of industries helps explain the dramatic growth in nonmanufacturing R&D over the past 2 decades.
R&D performance attributed to the trade industry reached $24.4 billion in 2001. Recent analysis of survey microdata revealed that only a fraction of this R&D was performed by companies whose primary business was wholesale or retail trade. As a consequence of assigning firms to one industry based on payroll datathe classification method used for the NSF Survey of Industrial Research and Developmentsome large companies were classified into an industry that was not closely related to its reported R&D activities. Although imperfect, this classification scheme reasonably categorizes most companies into industries closely aligned with their primary business activities. Unfortunately, because the sale and marketing of goods and services is a trade activity, a large pharmaceutical firm or electronics manufacturer would be classified in the trade industry if the payroll associated with its sales and marketing efforts outweighed that of any other industrial activity in the company. (See sidebar, "Redistributing Trade R&D" for further discussion of this issue.)
Nonmanufacturing industries associated with software and computer-related services such as data processing and systems design performed approximately $24.0 billion of company-funded R&D in 2001. As computing and IT became more powerful and ubiquitous over the past 2 decades, the demand for services associated with these technologies boomed. The R&D of companies providing these services also grew dramatically during this period. In 1987, when an upper bound estimate of software and other computer-related services R&D first became available, companies classified in the industry group "computer programming, data processing, other computer-related, engineering, architectural, and surveying services" performed $2.4 billion of company-funded R&D, or 3.8 percent of all company-funded industrial R&D. In 2001 the company-funded R&D of a comparable group of industries (excluding engineering and architectural services) was greater by a factor of 10 and accounted for 13.2 percent of all company-funded industrial R&D (table 6). This trend in the growth of software and computer-related services R&D shows no sign of slowing. Despite essentially no growth in total company-funded, industry-performed R&D between 2000 and 2001, the company-funded R&D for this group of industries grew by 10 percent.
The R&D performed by companies in the scientific R&D services industry more than doubled in the 4 years between 1997 and 2001 from $7.0 to $14.2 billion. The portion of this industry's R&D that was company-funded increased at an even faster pace, from $4.7 billion in 1997 to $10.9 billion in 2001. The scientific R&D services industry comprises companies that specialize in conducting R&D for other organizations, such as many biotechnology companies. (See sidebar, "Biotechnology R&D in Industry.") Although these companies and their R&D activities are classified as nonmanufacturing because they provide business services, many of the industries they serve are manufacturing industries. This implies that the R&D activities of a research firm that services a manufacturer would have been classified as R&D in manufacturing if the same research firm were a subsidiary of the manufacturer. Consequently, a growth in measured R&D in services may, in part, "reflect a more general pattern of industry's increasing reliance on outsourcing and contract R&D" (Jankowski 2001).
Although a great deal of R&D in the United States is related in some way to health care, companies specifically categorized in the health care services sector accounted for only 0.4 percent of all industrial R&D. In many industries, innovation often results from R&D performed in other industries. This is especially true with respect to health care where R&D in the pharmaceutical, scientific instrument, and software industries all play a role. This illustrates how R&D data disaggregated according to industrial categories (including the distinction between manufacturing and nonmanufacturing industries) may not always reflect the relative proportions of R&D devoted to particular types of scientific or engineering objectives or to particular fields of science or engineering.
Within the manufacturing industries, three groups dominate: computer and electronic products, transportation equipment, and chemicals (table 5). In 2001 computer and electronic products accounted for the largest amount of R&D performed among all industries at $47.1 billion, or 23.7 percent of all industrial R&D and 39.0 percent of all manufacturing R&D. For this subsector, industrial firms provided $41.2 billion in R&D support and the Federal Government funded the remainder. Reclassifying the R&D of wholesalers of computer and electronic products into manufacturing increases the R&D of this group of industries to $57.9 billion, or 29.2 percent of all industrial R&D. (See sidebar, "Redistributing Trade R&D.")
In 2001 transportation equipment accounted for the second most R&D performed in the manufacturing sector at $26.0 billion, or 13.1 percent of all industrial R&D. Of these expenditures, 19.1 percent was federally funded, primarily for R&D on aerospace products (planes, missiles, and space vehicles). In addition to aerospace products, this subsector includes a variety of other forms of transportation equipment, such as motor vehicles, ships, military armored vehicles, locomotives, and smaller vehicles such as motorcycles, bicycles, and snowmobiles.
In 2001 chemicals ranked third in R&D performed in the manufacturing subsector at $17.9 billion, approximately 1 percent of which was federally funded. In terms of R&D performance, the largest industry within the chemicals subsector is pharmaceuticals and medicines. In 2001 R&D performed by these companies accounted for 57 percent of non-Federal R&D funding in the chemicals subsector ($10.1 billion). Reclassifying the R&D of wholesalers of drugs and druggists' sundries into manufacturing increases the R&D of pharmaceuticals and medicines to $18.1 billion and the R&D of chemicals to $25.9 billion, or 13.0 percent of all industrial R&D. (See sidebar "Redistributing Trade R&D.")
Manufacturing R&D performers tend to be larger firms that perform more R&D on average than nonmanufacturing firms (table 8). As a share of the nation's GDP, manufacturing contributes less than 20 percent, but manufacturing industries account for 61 percent of total industrial R&D performance. Of the approximately 33,000 firms in the United States that performed R&D in 2001, 51 percent were in the manufacturing sector. Manufacturers dominate in terms of R&D performance largely because of the activities of the largest manufacturing firms. In 2001 the largest manufacturing firms (those with 25,000 or more employees) accounted for 49 percent of the R&D in the manufacturing sector, whereas nonmanufacturing firms in the same size category accounted for only 25 percent of total nonmanufacturing R&D.
Among smaller R&D-performing firms (those with fewer than 500 employees), those in the nonmanufacturing sector conduct significantly more R&D than those in the manufacturing sector, both in aggregate and on a per-firm basis. These firms accounted for 12 percent of manufacturing R&D, 31 percent of nonmanufacturing R&D, and 19 percent of all industrial R&D in 2001.
Although R&D tends to be performed by large firms in the manufacturing sector and smaller firms in the nonmanufacturing sector, considerable variation can be found within each sector, depending on the type of industry. R&D tends to be conducted primarily by large firms in several industrial subsectors: aircraft and missiles; electrical equipment; professional and scientific instruments; transportation equipment (not including aircraft and missiles); and transportation and utilities, which are in the nonmanufacturing sector. In these same sectors, however, much of the economic activity occurs in large firms to begin with, so the observation that most of the R&D in these sectors is also conducted by large firms is not surprising.
In addition to absolute levels of and changes in R&D expenditures, another key indicator of industrial commitment to science and technology (S&T) is R&D intensity, a measure of R&D relative to production in a company, industry, or sector. For most firms, R&D is a discretionary expense in the sense that it is not directly related to short-term revenues. Since R&D does not directly generate revenue in the same way that production expenses do, companies can trim their R&D budgets when profits fall. Evidence suggests, however, that R&D enjoys some degree of immunity from belt-tightening endeavors, even when the economy is faltering, because of its crucial role in laying the foundation for future growth and competitiveness.
Many ways exist to measure R&D intensity; the one used most frequently is the ratio of company-funded R&D to net sales. This statistic provides a way to gauge the relative importance of R&D across industries and among firms in the same industry. The industrial subsectors with the highest R&D intensities in 2001 were scientific R&D services (36.5 percent), software (19.3 percent), communications equipment (16.6 percent), and computer systems design and related services (16.5 percent). The R&D intensities of the professional, scientific, and technical services industries are particularly high because, as previously explained, much of the R&D reported by these companies also appears in their reported sales figures because the R&D activity is the product being sold. Industries with the lowest R&D intensities (0.5 percent or less) were food, broadcasting and telecommunications, and utilities (table 9). A decrease in the net sales of R&D-performing companies between 2000 and 2001 resulted in the ratio of R&D to sales for all industries increasing to 3.8 percent in 2001, up from 3.4 percent in 2000.
Although overall industrial R&D intensity increased between 2000 and 2001, the R&D intensity of very small companies (less than 100 employees) declined. These companies, on average, have much higher R&D-to-sales ratios than larger companies (table 9) because they include a large number of startups and young companies with less established revenue streams. Large, well-established companies often have reserves of cash and other liquid assets that allow them to maintain their R&D activities amid short-term economic downturns. Less mature companies, however, tend to be more reliant on outside investment and thus their expenditures on R&D are more likely to be cut in the event of a contraction in the economy or capital markets. This is one explanation for the divergence in the R&D intensities of very small companies and all other companies between 2000 and 2001.
Based on data from R&D performers, Federal agencies and FFRDCs performed a projected $37.1 billion of total U.S. R&D in 2003 (table 1), an average annual increase in real terms of 9.3 percent from the 2000 level of $27.1 billion. Among individual agencies, DOD continued to perform the most intramural R&D and is expected to account for more than half of all Federal obligations for intramural R&D in the future. In FY 2003, DOD is expected to perform more than twice the R&D of the second largest R&D-performing agency, HHS, which performs most of its intramural R&D at NIH (table 10).
DOE sponsors the most FFRDCs of any agency16 of the 36. These 16 FFRDCs performed a total of $8.7 billion of R&D in FY 2002, approximately three-fourths of all the R&D performed by FFRDCs (appendix table B-15). First established during World War II, FFRDCs are unique organizations that help the U.S. government meet special long-term research or development goals that cannot be met as effectively by in-house or contractor resources. The Federal Register states that an FFRDC is required "to operate in the public interest with objectivity and independence, to be free from organizational conflicts of interest, and to have full disclosure of its affairs to the sponsoring agency" (National Archives and Records Administration 1990). Total R&D performed by all FFRDCs (projected to be $12.2 billion in 2003) has grown at a real annual rate of 8.1 percent from its level of $9.2 billion in 2000 (appendix table B-1).
Besides performing R&D directly and funding R&D within other sectors as discussed earlier, the Federal Government also encourages R&D activity indirectly in the form of tax incentives. (See sidebar, "Federal R&D Tax Credit.")
Universities and colleges performed a projected $40.3 billion of R&D in 2003, an average annual increase in real terms of 7.9 percent from the 2000 level of $30.6 billion. The Federal Government is the largest source of support for academic research in the United States, funding an estimated 60.8 percent ($24.5 billion) of academic R&D in 2003. The next largest source of support for academic R&D is university-own funds (19.7 percent in 2003) followed by nonprofit institutions (7.4 percent), state and local governments (6.7 percent), and industry (5.3 percent).
Although industrial firms provide only a small portion of the R&D funding at U.S. universities and colleges, their funding of academic research has grown faster than any other sector over the past 2 decades. Between 1980 and 2000, industry's funding of academic R&D grew at an average annual rate of 7.7 percent after adjusting for inflation, outpacing total academic R&D, which grew at an average annual rate of 4.7 percent over the same period. After adjusting for inflation, industry's funding of academic R&D declined at an average annual rate of 2.2 percent between 2000 and 2003, lessening at about the same rate as industry's R&D funding of industrial R&D.
Although both public and private colleges and universities rely on the same funding sources for their R&D, the relative importance of those sources differs substantially for these two types of institutions (figure 15). For all public academic institutions combined, just over 9 percent of R&D funding in 2001 came from state and local governments, about 25 percent came from institutions, and about 52 percent came from the Federal Government. Private academic institutions received a much smaller portion of their R&D funding from state and local governments (about 2 percent) and institutional sources (about 10 percent) and a much larger share from the Federal Government (72 percent). The large difference in the role of institutional funds at public and private institutions is most likely due to a substantial amount of general-purpose state and local government funds that public institutions receive and decide to use for R&D (although data on such breakdowns are not collected). Both public and private institutions received approximately 7 percent of their respective R&D support from industry in 2001. Over the past 2 decades, the Federal share of support has declined, and the industry and institutional shares have increased for both public and private institutions.
Most academic R&D is now, and has been historically, concentrated in relatively few of the 3,600 U.S. institutions of higher education. The top 200 institutions ranked by total R&D expenditures accounted for about 96 percent of 2001 R&D expenditures. In 2001:
The historic concentration of academic R&D funds diminished somewhat between the mid-1980s and mid-1990s but has remained relatively steady since then (figure 16). In 1985, the top 10 institutions received about 20 percent of the nation's total academic R&D expenditures and the top 1120 institutions received 14 percent, compared with 17 and 13 percent, respectively, in 2001. The composition of the universities in the top 20 also fluctuated slightly from 1985 to 2001. There was almost no change in the share of the group of institutions ranked 21100 during this period. The decline in the top 20 institutions' share was matched by an increase in the share of those institutions in the group that were not in the top 100. This group's share increased from 17 to 20 percent of total academic R&D funds, signifying a broadening of the base.
Many universities and colleges are able to report their total and federally financed R&D expenditures for separate fields of science and engineering. The majority of academic R&D expenditures are devoted to the life sciences (figure 17)approximately four times the expenditures for the next largest field, engineering. Despite dramatic growth in Federal funding of R&D in the life sciences (particularly from NIH), Federal funding represents a smaller share of total R&D in this field than in many others. R&D in the physical sciences, mathematical sciences, and psychology, for example, show a higher concentration of Federal funding. The social sciences is the only group of fields that receives less than half of its funding from Federal sources (NSF, Division of Science Resources Statistics 2003).
Although there is no detailed information about the various non-Federal sources of academic R&D funding broken out by S&E field, it is possible to make some general observations. Agricultural sciences (a subset of life sciences) R&D, for example, is concentrated in public universities and colleges and receives most of its funding from non-Federal sources. It is fair to assume that most of this non-Federal funding comes from state and local governments and university-own funds. There also appears to be a correlation between industry funding and both medical sciences (a subset of life sciences) and engineering. Conversations with representatives of several large research universities confirmed that, at least for these schools, industry-funded R&D is concentrated in the fields of medical sciences and engineering.
 Recent methodological improvements in the estimation of total academic R&D have resulted in a break in the time series. Data for years before 1998 are slightly overstated compared with the data for later years. Had the same methodology been used for all years in the series, the average annual growth rate would have been closer to 4.3 percent per year in real terms from 1994 to 2000. See sidebar, "Academic Passed-Through Funds."
 Beginning with the 2001 survey cycle, industry-administered FFRDCs were removed from the industrial R&D statistics. They are now separately surveyed and reported in the national statistics. This resulted in a relative increase in the share of R&D performed by nonmanufacturing industries. In 2000, when these FFRDCs were included in the industrial R&D totals, R&D performed by nonmanufacturing industries accounted for 37.8 percent of total industrial R&D.
 "Trade" refers to both the wholesale and retail trade industries and is a distinct entry in the NSF industry R&D statistics, as is scientific R&D services. Software and computer-related services, however, is the sum of three distinct industries from the NSF statistics: software, other information, and computer systems design and related services.
 Details on how companies are assigned industry codes in the NSF Survey of Industrial Research and Development can be found on the NSF website (http://www.nsf.gov/statistics/nsf02312/sectb.htm#frame).
 Although disclosure of Federal R&D funding prohibited the precise tabulation of total R&D performance for this industry, total R&D was between $24.5 billion and $24.6 billion.
 The introduction of a more refined industry classification scheme in 1999 allowed more detailed reporting in nonmanufacturing industries. For the cited 2001 statistic, the R&D of companies in software, other information, and computer systems design and related services industries were combined. These three industries provided the closest approximation to the broader category cited for earlier years without exceeding the coverage of the broader category.
 Despite the change of industry classification schemes in 1999, analysts have verified that data for the scientific R&D services industry are comparable over the period under discussion.
 R&D performance is even more skewed toward companies with large R&D programs (total R&D of $100 million or more). The 243 firms in this category accounted for 73 percent of manufacturing R&D, 56 percent of nonmanufacturing R&D, and 67 percent of all industrial R&D in 2001.
 A similar measure of R&D intensity is the ratio of R&D to value added (sales minus the cost of materials). Value added is often used in studies of productivity because it allows analysts to focus on the economic output attributable to the specific industrial sector in question by subtracting materials produced in other sectors. For a discussion of the connection between R&D intensity and technological progress, see, for example, R. Nelson, "Modeling the connections in the cross section between technical progress and R&D intensity," RAND Journal of Economics 19(3) (Autumn 1988): 478485.
 Federal intramural R&D obligations are interpreted as R&D performance expenditures for the purpose of this analysis.
 The Carnegie Foundation for the Advancement of Teaching classified about 3,600 degree-granting institutions as higher education institutions in 1994.