by Francisco Moris [1]

Research and development in software contributes to emerging research, investment, and economic policy areas, such as artificial intelligence, information and communication technologies (ICT), and the digital economy.[2] Accordingly, software R&D is an increasingly large technology area of U.S. business R&D expenditures. For example, in 2016, software R&D accounted for $120.8 billion of the $374.7 billion in U.S. business R&D, or 32%, compared with 20% in 2006, as reported by the National Center for Science and Engineering Statistics (NCSES) within the National Science Foundation.

A recent change in the treatment of software R&D in the U.S. gross domestic product (GDP) and other national income and product accounts (NIPAs), published by the Bureau of Economic Analysis (BEA), underscores the importance of software for U.S. R&D statistics. R&D was capitalized, or recognized as investment (gross fixed capital formation [GFCF][3]), for the first time in the U.S. NIPAs in 2013 (Crawford et al. 2014; Moris et al. 2015). At that time, software R&D expenditures were capitalized under software investment—not under R&D—to avoid duplication, because software was already considered a fixed asset.[4] However, this also meant that for NIPA purposes, "R&D investment" was effectively non-software R&D investment, in contrast to NCSES's inclusion of both software and non-software R&D in total U.S. R&D expenditures.

The 2018 Comprehensive Update of the NIPAs reclassified software R&D as R&D investment to resolve the above inconsistency (Chute, McCulla, Smith 2018).[5] This InfoBrief presents updated BEA data on R&D investment and software investment and examines NCSES total R&D expenditures and business software R&D expenditures.

R&D Expenditures, R&D Investment, and Intellectual Property Products

After the 2018 NIPA update, R&D investment as reported by BEA matches all technology areas included in NCSES R&D expenditures statistics. Figure 1 compares total R&D expenditures (NCSES) and total R&D investment (BEA). R&D investment is generally larger than R&D expenditures because the former is designed to reflect the market value of R&D, not the current period's cost, as reflected in R&D expenditures data (Crawford et al. 2014).

FIGURE 1. NIPA R&D investment versus R&D expenditures: 1959–2017
FIGURE 1. NIPA R&D investment versus R&D expenditures: 1959–2017.

BEA = Bureau of Economic Analysis; NCSES = National Center for Science and Engineering Statistics; NIPA = national income and product account.

SOURCES: BEA R&D investment from NIPA tables are available at https://apps.bea.gov/iTable/index_nipa.cfm, accessed 25 August 2018. R&D expenditures are from the National Science Foundation, National Center for Science and Engineering Statistics, National Patterns of R&D Resources (annual series).

Figure 1 Source Data: Excel file

In national economic accounting, R&D investment along with software and entertainment, literary, and artistic originals are intangible assets under the intellectual property products (IPPs) account. In turn, IPPs are a component of fixed investment along with structures and equipment for both private and government sectors, as defined in the NIPA context (Moris et al. 2015). The shares of IPPs in total fixed investment have increased from around 15% in the 1960s, to around 20% in the 1990s, to over 25% in recent years (figure 2). Even though R&D was recognized by BEA as investment after software had already been recognized as investment, software is of course a more recent activity. Thus, based on the revised BEA investment data, over 80% of IPPs were categorized as R&D investment in the late 1950s. Since then, the share decreased to around 60% in the mid-1990s and has fluctuated narrowly around 55% since the late 1990s (figure 3). The share of software investment increased steadily to over 30% in the "dot com" era of late 1990s and early 2000s, reaching just below 40% in recent years. The balance is captured by investment in entertainment, literary, and artistic originals.

FIGURE 2. Types of fixed investment: 1959–2017
FIGURE 2. Types of fixed investment: 1959–2017.

BEA = Bureau of Economic Analysis; NIPA = national income and product account.

SOURCE: BEA NIPA tables available at https://apps.bea.gov/iTable/index_nipa.cfm, accessed 25 August 2018.

Figure 2 Source Data: Excel file

FIGURE 3. Types of intellectual property products: 1959–2017
FIGURE 3. Types of intellectual property products: 1959–2017.

BEA = Bureau of Economic Analysis; NIPA = national income and product account.

SOURCE: BEA NIPA tables available at https://apps.bea.gov/iTable/index_nipa.cfm, accessed 25 August 2018.

Figure 3 Source Data: Excel file

Between 2000 and 2002, the average annual total software R&D investment was on the order of $30 billion, most of it in the private sector,[6] compared with business R&D expenditures on software totalling $24 billion in 2002, according to NCSES data.

Trends in Business R&D Expenditures on Software

According to NCSES statistics, U.S.-located, for-profit businesses (regardless of ownership and industry classification) performed $120.8 billion in software R&D in 2016, or 32% of $374.7 billion in domestic business R&D, more than double its 13% share in 2002. Over the past 10 years, software R&D increased annually at a 9.6% compound average growth rate, compared with a rate of 4.2% for total business R&D (table 1). Since 2006, manufacturing industries exhibited larger growth rates of software R&D than did nonmanufacturing industries, although most R&D in this area is still performed by the nonmanufacturing industries. Figure 4 shows an increasing trend in software R&D relative to total business R&D.[7] And figure 5 shows that in contrast to overall U.S. business R&D, software R&D is concentrated in nonmanufacturing companies.

TABLE 1. Total domestic business R&D and software R&D expenditures: 2006 and 2016
(Millions of current U.S. dollars)
Year and Industry Total business R&D Software R&D

a Calculated as compound average growth rate.

NOTE: For relative standard errors and other information, see "Data Sources and Limitations."

SOURCES: National Science Foundation, National Center for Science and Engineering Statistics, and U.S. Census Bureau, Survey of Industrial R&D, 2006 (detailed statistical table 7) and Business R&D and Innovation Survey, 2016 (detailed statistical table 80).

Table 1 Source Data: Excel file

2006
All industries 247,669 48,299
Manufacturing 171,814 10,720
Nonmanufacturing 75,855 37,579
2016
All industries 374,685 120,824
Manufacturing 250,553 35,984
Nonmanufacturing 124,132 84,840
2006–16 annual growth rate (%)a
All industries 4.2 9.6
Manufacturing 3.8 12.9
Nonmanufacturing 5.0 8.5
FIGURE 4. Business R&D performance and share of software R&D: 2002–16
FIGURE 4. Business R&D performance and share of software R&D: 2002–16.

NOTES: Total software R&D from the Business R&D and Innovation Survey, and thus the corresponding shares relative to business R&D, are not available for 2008 to 2012. Software R&D includes R&D on software products and R&D on software embedded in other products, funded by the performer or by others. For more information, see "Data Sources and Limitations."

SOURCES: National Science Foundation, National Center for Science and Engineering Statistics, and U.S. Census Bureau, Survey of Industrial R&D (2002–07) and Business R&D and Innovation Survey (2008 forward).

Figure 4 Source Data: Excel file

FIGURE 5. Business R&D and software R&D, by major industrial category: 2016
FIGURE 5. Business R&D and software R&D, by major industrial category: 2016.

SOURCE: National Science Foundation, National Center for Science and Engineering Statistics, and U.S. Census Bureau, Business R&D and Innovation Survey, 2016.

Figure 5 Source Data: Excel file

Along with computer and communications equipment, software is a component of ICT (NSB 2018; OECD 2017). For purposes of U.S. business R&D statistics, ICT industries comprise computer and electronic products manufacturing (NAICS 334) and selected services industries within the information sector (NAICS 51) and the professional, scientific, and technical services sector (NAICS 54), which is consistent with international ICT statistics (Shackelford, Jankowski 2016; Mas et al. 2017). As discussed next, ICT industries are the top performers of software R&D.

Table 2 shows a profile of software R&D by industry and funding source. By industry of performance, the 30% share of software R&D in manufacturing is concentrated in computer and electronics products (NAICS 334), which includes semiconductor and other electronic components manufacturing as well as navigational, measuring, electromedical, and control instruments (the latter is included in the electronic products aggregate). Another notable industry is other transportation equipment, which includes the defense-related industry of aerospace products and parts (NAICS 3364).

TABLE 2. Total domestic business R&D and software R&D expenditures, by selected industry: 2016
(Millions of current U.S. dollars)
Software R&D expenditures
Industry NAICS code Business
R&D
expenditures
Total Paid by
company
Paid by
others
Federally
funded
Share of
software R&D
in business
R&D (%)
Share of
industry in
software R&D
(%)

* = amount < $500,000; ** = < 0.05%; D = data withheld to avoid disclosing operations of individual companies; i = > 50% of the estimate is a combination of imputation and reweighting to account for nonresponse.

NAICS = 2012 North American Industry Classification System.

NOTES: Detail may not add to total because of rounding or sub-categories that are not available. For relative standard errors and other information, see "Data Sources and Limitations."

SOURCE: National Science Foundation, National Center for Science and Engineering Statistics, and U.S. Census Bureau, Business R&D and Innovation Survey, 2016.

Table 2 Source Data: Excel file

All industries - 374,685 120,824 i 111,933 i 8,890 4,481 i 32.2 100.0
Manufacturing industries 31–33 250,553 35,984 i 31,527 i 4,457 3,057 i 14.4 29.8
Chemicals 325 73,575 575 i 430 i 145 i 1 i 0.8 0.5
Basic chemicals 3251 2,545 41 41 0 0 1.6 **
Pharmaceuticals and medicines 3254 64,628 302 i 165 137 i 1 i 0.5 0.2
Other chemicals other 325 6,402 232 i 224 i 8 0 3.6 0.2
Machinery 333 12,585 1,290 1,245 44 D 10.3 1.1
Agricultural implement 33311 1,565 187 185 2 2 11.9 0.2
Semiconductor machinery 333242 3,411 222 212 i 9 * 6.5 0.2
Engine, turbine, and power transmission equipment 3336 1,959 217 216 1 0 11.1 0.2
Other machinery other 333 5,650 664 i 632 i 32 D 11.8 0.5
Computer and electronic products 334 77,385 24,540 i 22,935 i 1,606 D 31.7 20.3
Semiconductor and other electronic components 3344 31,391 8,134 i 8,033 i 101 D 25.9 6.7
Other electronic products other 334 45,993 16,406 14,902 i 1,504 D 35.7 13.6
Electrical equipment, appliances, and components 335 4,771 855 i 616 i 238 i 1 i 17.9 0.7
Transportation equipment 336 51,275 6,301 3,903 2,398 D 12.3 5.2
Automobiles, bodies, trailers, and parts 3361–63 22,042 1,592 1,456 136 i 1 7.2 1.3
Other transportation other 336 29,234 4,709 2,446 i 2,262 D 16.1 3.9
Nonmanufacturing industries 21–23, 42–81 124,132 84,840 i 80,406 i 4,434 i 1,424 68.3 70.2
Information 51 70,748 63,213 i 62,870 i 343 11 89.3 52.3
Publishing 511 33,574 27,864 27,689 176 3 83.0 23.1
Telecommunications 517 D 2,996 i D D 0 D 2.5
Data processing, hosting, and related services 518 11,914 11,233 11,171 62 6 94.3 9.3
Other information other 51 D 21,120 i D D 2 D 17.5
Professional, scientific, and technical services 54 37,595 14,185 i 10,174 i 4,011 i 1,411 37.7 11.7
Architectural, engineering, and related services 5413 3,412 515 i 231 i 284 i 209 i 15.1 0.4
Scientific research and development services 5417 14,842 2,060 i 363 i 1,698 i 768 13.9 1.7
Other professional, scientific, and technical services other 54 19,341 11,610 i 9,581 i 2,029 i 435 60.0 9.6

Within information industries, the largest performer of software R&D is publishing (NAICS 511), which includes software publishers (NAICS 5112, not available separately). The largest performer of software R&D in professional, scientific, and technical services industries is other professional, scientific, and technical services, which includes computer systems design and related services (NAICS 5415, not available separately).

About 93% of software R&D is funded by the performing companies (table 2), compared with 85% for overall business R&D in 2016 (Wolfe 2018). However, of the amount paid by others, federal sources represent 50% for software R&D compared with 42% for overall business R&D paid by others. Further, for software R&D performed in the manufacturing sector, federal sources funded over two-thirds of software R&D paid by others.

Data Sources and Limitations

BEA's NIPA data in this InfoBrief were downloaded in early to mid-August 2018. For current data on GDP, fixed assets, and IPPs investment, see https://apps.bea.gov/iTable/index_nipa.cfm. In the 2008 System of National Accounts (SNA) (EC et al. 2009), the statistical manual that guides the economic accounts, the definition of R&D is consistent with the Organisation for Economic Co-operation and Development's Frascati Manual 2015 (paragraph 2.5). The SNA's definition of R&D investment is the following: "IPPs include the results of R&D... Research and [experimental] development consists of the value of expenditures on creative work undertaken on a systematic basis in order to increase the stock of knowledge...and use of this stock of knowledge to devise new applications... The value of R&D should be determined in terms of the economic benefits it is expected to provide in the future." (EC et al. 2009: paragraph 10.103). For BEA NIPA methodology, see https://www.bea.gov/resources/methodologies/nipa-handbook.

NCSES statistics on total U.S. R&D expenditures were obtained from National Patterns of R&D Resources reports, available at https://www.nsf.gov/statistics/natlpatterns/. These annual publications provide a comprehensive account of total U.S. R&D performance and funding across all sectors. NCSES estimates on U.S. business and software R&D expenditures were derived from the Survey of Industrial R&D [SIRD] (up to 2007) and Business R&D and Innovation Survey [BRDIS] (2008 forward). In SIRD, software development was defined as "the formulation of programs, applications, routines, etc., for computers, excluding those used exclusively for internal company operations."

Annual samples for BRDIS are selected to represent for-profit, nonfarm companies (publicly or privately held) with five or more U.S. employees. Company is defined as a U.S.-located business, either U.S. owned or a U.S. affiliate of a foreign parent company, that performs or funds R&D. Software R&D is defined by BRDIS as "R&D activity in software and Internet applications refers to activities with an element of uncertainty and that are intended to close knowledge gaps and meet scientific and technological needs. Software R&D expenditures are reported regardless of who is the eventual user of the R&D (internal or external)." Further, "R&D activity in software includes the following:

and excludes the following:

Software R&D refers to R&D performed by any company, not just companies classified in the software industry, on software products and on software embedded in other products. These components of software R&D are potentially difficult to separate out for reporting purposes. Totals for software R&D from BRDIS are not available for 2008 to 2012. For these years, the survey collected separately the two potentially overlapping software R&D components, so even though these components are available they should not be summed to obtain total domestic software R&D performed by the business sector. Further, for 2008 to 2010, software R&D data were collected only for company-funded R&D.

In BRDIS, software R&D activity is collected as percentage of specific expenditure totals in sections devoted to "technology focus." (Technology focus also includes items on biotechnology and nanotechnology. Data collection on artificial intelligence is being tested.) Further, these data are collected only for domestic R&D performed by large R&D companies. For 2016, the standard (longer) form, BRDI-1, was sent to companies having $1 million or more worldwide total R&D activity in 2014 or 2015. The screener form, BRDI-1(S), was sent to all other companies. Software R&D is included in R&D totals for both forms. However, technology focus detail was collected only in BRDI-1. (SIRD also used long- and short-survey forms for a given year, but software R&D detail was collected in both forms.) In the 2016 BRDIS, software R&D was collected in BRDI-1 items 4–12 (company-funded R&D) and 4–24 (R&D funded by others). For survey forms, see https://www.nsf.gov/statistics/srvyindustry/#qs.

In this InfoBrief, money amounts are expressed in current U.S. dollars. Statistics are subject to sampling and non-sampling errors. The full set of 2016 BRDIS detailed statistical tables, including imputation rates, relative standard errors, and technical notes will be available at https://www.nsf.gov/statistics/srvyindustry/. Historical SIRD data are available at https://www.nsf.gov/statistics/iris/. For further information on business R&D statistics, see Wolfe (2018).

References

Barefoot K, Curtis D, Jolliff W, Nicholson JR, Omohundro R. 2018. Defining and measuring the digital economy. BEA Working Paper. Suitland, MD: Bureau of Economic Analysis. Chute JW, McCulla SH, Smith S. 2018. Preview of the 2018 Comprehensive Update of the National Income and Product Accounts: Changes in methods, definitions, and presentations. Survey of Current Business April:1–21.

Crawford MJ, Lee J, Jankowski JE, Moris FA. 2014. Measuring R&D in the national economic accounting system. Survey of Current Business November:1–15.

European Commission (EC), International Monetary Fund, Organisation for Economic Co-operation and Development, United Nations, World Bank. 2009. System of National Accounts 2008. New York.

Grimm A. 2016. Trends in U.S. trade in information and communications technology (ICT) services and in ICT-enabled services. Survey of Current Business May:1–19.

Kelly PA, McCulla SH, Wasshausen DB. 2018. Improved estimates of the national income and product accounts: Results of the 2018 Comprehensive Update. Survey of Current Business September:1–54.

Mas M, Fernández de Guevara J, Robledo JC, López-Cobo M. 2017. The 2017 PREDICT key facts report: An analysis of ICT R&D in the EU and beyond. EUR 28594 EN. Luxembourg: Publications Office of the European Union.

Moris F, Jankowski J, Boroush M, Crawford M, Lee J. 2015. R&D Recognized as Investment in U.S. GDP Statistics: GDP Increase Slightly Lowers R&D-to-GDP Ratio. InfoBrief NSF 15-315. Arlington, VA: National Science Foundation, National Center for Science and Engineering Statistics. Available at https://www.nsf.gov/statistics/2015/nsf15315/.

National Science Board. 2018. Science and Engineering Indicators 2018. NSB-2018-1. Alexandria, VA: National Science Foundation. Available at https://www.nsf.gov/statistics/indicators/.

Organisation for Economic Co-operation and Development (OECD). 2010. Handbook on Deriving Capital Measures of Intellectual Property Products. Paris: OECD Publishing.

Organisation for Economic Co-operation and Development (OECD). 2015. Frascati Manual 2015. Paris: OECD Publishing.

Organisation for Economic Co-operation and Development (OECD). 2017. Digital Economy Outlook 2017. Paris: OECD Publishing.

Shackelford B, Jankowski J. 2016. Information and Communications Technology Industries Account for $133 Billion of Business R&D Performance in the United States in 2013. InfoBrief NSF 16-309. Arlington, VA: National Science Foundation, National Center for Science and Engineering Statistics. Available at https://www.nsf.gov/statistics/2016/nsf16309/.

Wolfe RM. 2018. Businesses Spent $375 Billion on R&D Performance in the United States in 2016. InfoBrief NSF 18-312. Arlington, VA: National Science Foundation, National Center for Science and Engineering Statistics. Available at https://www.nsf.gov/statistics/2018/nsf18312/.

Notes

[1] Francisco Moris, Research and Development Statistics Program, National Center for Science and Engineering Statistics, National Science Foundation, 2415 Eisenhower Avenue, Suite W14200, Alexandria, VA 22314 (fmorisor@nsf.gov; 703-292-4678).

[2] The digital economy includes Internet services and transactions, information and communications technologies and infrastructure, and databases. For example, ICT-enabled services trade is delivered over wireless or mobile networks. See Barefoot et al. (2018) and Grimm (2016).

[3] Fixed capital or fixed assets are produced assets that are used repeatedly, or continuously, in production for more than one year (EC et al. 2009: paragraph 10.11). GFCF (capital formation or investment for short) is the value of net acquisition of fixed assets, either purchased or produced on own account (internally developed) (EC et al. 2009: paragraphs 1.52, 10.32). Fixed assets consist of equipment, structures, and intellectual property products (IPPs). In the United States, IPPs include software; R&D; and entertainment, literary, and artistic originals.

[4] Computer software was capitalized in the U.S. NIPA accounts in 1999. Software investment consists of prepackaged software (excluding software embedded, or bundled, in computers and other equipment), custom software, and own account software. Software R&D was considered part of own account software (OECD 2010).

[5] There were two major definitional changes in the national accounts affecting software and R&D investment. In addition to the reclassification of software R&D, depreciation in own account software and in own account R&D was replaced with a measure of capital services (a measure reflecting both depreciation and the return to capital) (Chute, McCulla, Smith 2018). For data year 2016, private R&D investment was revised up by $78.2 billion because of definitional changes: $68.4 billion from the software R&D classification and $9.8 billion from the introduction of capital services in own-account R&D. Private software investment was revised down by $64.4 billion, which includes the $68.4 billion in software R&D reclassified into R&D investment and an offsetting increase of $4 billion from the capital services adjustment (see table 5 in Kelly, McCulla, Wasshausen 2018).

[6] Based on table 2 in Crawford et al. (2014) and on BEA's 2018 update in the treatment of software R&D. In the NIPAs, the private sector includes for profit-businesses, non-profit institutions, and private academic institutions.

[7] Totals for software R&D from BRDIS, and thus the corresponding shares, are not available for 2008 to 2012. For more information, see "Data Sources and Limitations."