National Academies Press: OpenBook
« Previous: 1 Personal Computing
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 53
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 54
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 55
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 56
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 57
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 58
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 59
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 60
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 61
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 62
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 63
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 64
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 65
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 66
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 67
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 68
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 69
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 70
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 71
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 72
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 73
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 74
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 75
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 76
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 77
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 78
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 79
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 80
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 81
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 82
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 83
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 84
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 85
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 86
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 87
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 88
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 89
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 90
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 91
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 92
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 93
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 94
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 95
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 96
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 97
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 98
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 99
Suggested Citation:"2 Software." National Research Council. 2008. Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies). Washington, DC: The National Academies Press. doi: 10.17226/12112.
×
Page 100

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

2 Software Ashish Arora Carnegie Mellon University Chris Forman Georgia Institute of Technology JiWoong Yoon Kyung Hee University, Seoul, South Korea INTRODUCTION The global movement of software services activities (defined to include soft- ware engineering services and research and development [R&D] as well as the development of software products) to locations outside of the United States is an important and growing phenomenon that has recently attracted widespread atten- tion. Over the period 1995-2002, exports of business services and computer and information services grew at an average annual rate of over 40 percent in India and at a rate of 20 percent in Ireland. These changes have received widespread attention within the United States and have led to concerns of a “hollowing out” of the American information technology (IT) sector and about the potential loss of American technological leadership. However, despite these changes in the location of production of IT services, there is relatively little evidence of global changes in the location of new software product development. U.S. companies have historically been and continue to be the leading exporters of software products. Moreover, evidence from software patents suggests that inventive activity in software continues to be concentrated in the United States. In the short run, the United States will continue to enjoy a significant lead over other countries in the stock of highly skilled programmers and software designers that provide it with an advantage in the production of new software products. Moreover, proximity to the largest source of IT demand and potential agglomeration economies arising from proximity to competitors and complementors provide software product companies located in the United States with a significant advantage. 53

54 INNOVATION IN GLOBAL INDUSTRIES DISPERSION OF INVENTIVE ACTIVITY IN SOFTWARE In this chapter we provide evidence on the geographic distribution of inven- tive activity in software. Economists have long made a distinction between in- novation and invention in the study of technological change. Schumpeter (1934) defined innovations as new, creative combinations that upset the equilibrium state of the economy. Mokyr (2002) defines invention as an increment in the set of technological knowledge in a society. Schumpeter pointed out that invention does not imply innovation, and that it is innovation that provides capitalism with its dynamic elements. Because it is more easily measured, in this chapter we will focus on the geographic dispersion of inventive activity. However, we adopt the position of Mokyr (2002), who argues that in the long run invention is a neces- sary precursor to innovation. Unlike some of the other industries studied in this volume, one feature of software development is that it is frequently performed both by suppliers of software packages and services and by users. As a result, software development occurs throughout all industries in the economy, and so to understand the location of inventive activity in software it is insufficient to examine where one or two industries are located. To understand this point further, it is helpful to gain a better understanding of the types of software development activity. The design, installation, implementa- tion, and use of software consist of several phases. Messerschmitt and Szyperski (2002) identify two distinct value chains in software development. First, there is a supply value chain in which software creators develop software artifacts that provide value for the end user. This part of the software value chain consists primarily of design and development activities that can be thought of as software “production.” In the past this role had been played primarily by independent us- ers, third-party programmers, or independent software vendors creating custom software, but over the past 20 years this role has passed increasingly to indepen- dent software vendors creating software products. The output of this value chain contains all of what we would traditionally de- fine as software products, such as word processors, operating systems, enterprise software such as enterprise resource planning (ERP) and business intelligence software, as well as middleware software, such as some transaction processing middleware and enterprise application integration. The total value of production in the software product industry was $61,376.9 million in 1997, and 195,200 persons were employed in this industry in the same year. Firms that operate in   Data from the U.S. Bureau of Economic Analysis input-output tables. This figure includes the total value of products made in NIPA industry 511200 (Software Publishers); 1997 is the latest benchmark year for the input-output tables. More recent years do not separate software producers from other information publishers.   Data from the Bureau of Labor Statistics (BLS) on the number of employees in the software publishing industry (NAICS 5112), available at http://www.bls.gov/ces/home.htm.

SOFTWARE 55 this value chain include all of the well-recognized names that are traditionally regarded as “software” firms, including Microsoft, Adobe, Oracle, and the SAS Institute, as well as smaller firms such as Oblix and Primatech. This value chain also includes the activity of third-party firms involved in custom programming and software analysis and design. Such firms create custom software products for their customers and include firms like CIBER, Inc., Intergraph Corp., and xwave Solutions. The total value created in custom programming and design services was $115,834.6 million in 1997 while total employment was 675,000 in 1997, indicating that both revenue and employment in this sector are greater than that in the packaged software industry.  Moreover, custom programming and design services are also growing faster than is the soft- ware publishing industry. Though 1997 is the last year for which we have data on revenues by industry, we can compare employment growth across these two industries. Employment in custom programming and design services has grown from 675,000 in 1997 to 1,025,300 in 2005, for an average annual growth rate of 5.8 percent. In contrast, employment in software publishing has grown from 195,200 in 1997 to 238,700 in 2005, for an average annual growth rate of 2.5 percent. Second, there is a software requirements value chain in which users add functionality to software to meet their own needs. Users engage in co-inventive activity (Bresnahan and Greenstein, 1996) to translate general-purpose software into a specific application. Such co-inventive activity may include modifications to packaged software applications or development of new applications. However, in business software it also involves changes to business processes or organiza- tion design. Activity in this value chain includes both programming by professional pro- grammers and software designers employed by IT-using firms and programming activities performed by users themselves. The activity of both groups is difficult to measure but represents a major share of value created. Scaffidi, Shaw, and Myers (2005) estimate that there were approximately 80 million end-user pro- grammers in 2005, compared to 3 million professional programmers. Moreover, occupation data from the United States indicate that over two-thirds of software professionals do not work for IT firms but rather work for IT-using industries. Neither this software development activity performed by users nor the work performed by software professionals working for IT users is measured in any systematic statistics.  These calculations are based on total sales in custom computer programming services (NAICS 541511) and computer systems design services (NAICS 541512). This latter category may include activities outside of programming, such as IT systems design and integration. A conservative estimate of the value and employment of third-party custom programming services uses only NAICS 541511 and yields estimates of $86,326.8 million and 522,300, respectively.  This estimate includes those who create user-developed software that is not sold in markets.  Data from BLS Occupational Employment Statistics.

56 INNOVATION IN GLOBAL INDUSTRIES Though systematic evidence is rare, what we do know suggests that eco- nomic activity in this value chain is likely to be far greater than that in the supply value chain. According to Gormely et al. (1998), though the typical cost of imple- menting an ERP application suite is $20.5 million, only $4 million of this cost is related to hardware and software; the rest is due to the costs of implementing and deploying the software within the business. Using data on sales of software products and services in several Western European countries, Steinmueller (2004) estimates that for every €1 spent on software there is an additional €2.36 spent on IT-related business services. However, this estimate is likely a lower bound, because it includes only software services conducted through market transactions and excludes software development activities within IT-using firms themselves. The importance of the software requirements value chain has two implica- tions for the measurement of where inventive activity in software takes place. First, a large part of value creation in software takes place outside of firms that reside in what is considered the software product industry. The value of this activ- ity goes largely unmeasured in traditional government statistics, as it often occurs as a labor expense within firms developing or implementing packaged software. Second, it is very difficult to place a precise definition of what exactly con- stitutes inventive activity in software. Creation and modification of source code is of course one major component, but so are user modification and business process change. Should these latter activities be included as well? Moreover, how should we treat changes to software code that are embedded in IT hardware? Are these hardware or software inventions? As we will discuss next, given available data, a precise estimate of inventive activity in software is probably not feasible. Instead, we provide a variety of metrics that enable us to estimate broad trends and orders of magnitude in economic and inventive activity in software. In the section “Trends in the Location of Value Creation” we provide evi- dence of recent trends in globalization of software services. These data provide evidence on globalization of activity in the software requirements value chain and some inventive activity conducted by services firms in the supply value chain, though they will largely miss changes in cross-country software service activi- ties that are undertaken by firms outside of the software services industry. In the section “Empirical Evidence on the Location of Inventive Activity” we use U.S. software patent data to examine changes in the global dispersion of inventive activity in software product development. TRENDS IN THE LOCATION OF VALUE CREATION In this section we investigate broad trends in the location of value creation activities in software. We begin with some statistics describing global variation in   Itis interesting to note that the U.S. Patent Office has struggled with similar definitional issues, within the context of so-called business method patents (Allison and Tiller, 2003).

SOFTWARE 57 the exports and imports of software products and services, followed by a qualita- tive description of recent trends in countries that have been known to be active producers in the market for software products and services. Statistical Trends Software Products Figure 1 shows the percentage of total 2002 software product exports and imports by selected Organisation for Economic Co-operation and Development (OECD) countries. The figure shows that among OECD countries the United States continues to be the leader by a wide margin in the export of software prod- ucts, accounting for 21.7 percent of total software exports. The next closest coun- try is Ireland, which accounts for 16 percent of software exports. However, as we will discuss in further detail, most of Ireland’s software exports arise from U.S. multinational companies that utilize Ireland as a base of operations to localize All Others United States United Kingdom Switzerland Sweden Netherlands Korea Japan Italy Ireland Germany France Canada Austria 0% 5% 10% 15% 20% 25% Exports % Total Imports % Total FIGURE 1  Percentage of total 2002 software product exports and imports by OECD country. SOURCE: OECD (2004, Table C.1.8; OECD trade in software goods, 1996- software-1.eps 2002). Compiled from International Trade Statistics database.

58 INNOVATION IN GLOBAL INDUSTRIES U.S. software products to be shipped to countries in the European Union.  Since the bulk of software product exports from Ireland are due to U.S. multinationals in Ireland—Sands (2005) shows that over 92 percent of Irish software exports are from foreign firms—this suggests that the share of U.S. software exports in global trade flows is probably closer to one-third rather than the one-fifth that the OECD statistics indicate. Following that, the next largest exporters are Germany (due in part to software exports from ERP giant SAP) and the United Kingdom. No other country accounts for more than 10 percent of software exports. Most notably, Japan accounts for only 2.5 percent of total software exports. Figure 2 presents total packaged software product sales by region. The story here remains the same: North America represents the largest share of packaged software sales, and this percentage has been increasing over time from 47 percent in 1990 to 54 percent in 2001. We explore why other countries have not been more successful in developing software products in further detail in the next section. Software Services Figure 3 shows data from the OECD Economic Outlook (2006) and reports the global share of 1995 and 2004 exports in IT services, obtained by summing the categories “computer and information services” and “other business services” from the IMF Balance of Payments data. Though subject to a variety of caveats about measurement and coverage, Figure 3 suggests that the distribution of IT service exports is more evenly distributed across countries than is the distribu- tion of software product exports. Many smaller countries are experiencing rapid growth in their exports of IT services, though some are starting from a very small base. To explore trends in imports, we use data from the U.S. Bureau of Economic Analysis (BEA) on International Trade in Services. Table 1 provides data on in- terfirm trade in exports and imports of IT services in 1998 and 2004, calculated by summing the categories “Computer and Information Services” and “Royal- ties and License Fees.” Exports of these services grew from $6,900 million to $10,862 million from 1998 to 2004, while imports grew from $1,992 to $2,591 million from 1998 to 2004. Cross-border exports to and imports from unaffiliated foreign firms of com-   Localization activities include activities such as manual translation or adapting software products to local markets.  The columns labeled “Computer and Information Services” provide data on exports and imports of private services among unaffiliated firms. The columns, “Royalties and License Fees” in the same table include computer-related services that were delivered to foreign markets through cross-border software licensing agreements. These data do not include intrafirm exports of computer services because BEA does not in general release statistics on many of the countries in Table 1. They also do not include wages of U.S. residents who provide computer services to nonresidents.

SOFTWARE 59 200 200 180 180 9% 160 160 140 140 OECD: Asia 30% 120 120 Other OECD 100 100 Rest of the world 80 OECD: EU-14 80 60 60 11% 54% 40 40 35% OECD: North America 20 20 47% 0 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 FIGURE 2  Packaged software sales by region, 1990-2001 (U.S. dollars). SOURCE: OECD (2002) using International Data Corporation data. Reported in Thoma and Torrisi (2006). with type replaced 1995 2004 14 12 Exports 10 Percent 8 6 4 2 0 US UK Germany Netherlands Ireland France Italy Japan China Hong Kong (1) Belgium Spain Austria Singapore Sweden India (1) Canada Denmark Switzerland Korea Indonesia Israel Norway Saudi Arabia Luxembourg Brazil Russia Thailand Australia Lebanon FIGURE 3  Top 30 country shares of reported exports of other business services and computer and information services, 1995 and 2004 (2004 data not yet available for all countries). For Hong Kong (China), India, and the Slovak Republic, data are for 2003. with type replaced Republished with permission from OECD Economic Outlook (2006). Based on IMF Bal- ance of Payments Database, March 2006.

60 INNOVATION IN GLOBAL INDUSTRIES puter and information services are shown in Table 1. Computer and information services (NAICS 518) include the categories computer and data processing ser- vices (NAICS 5181) and database and other information services (NAICS 5182). This table was reorganized based on the tables of Business, Professional, and Technical Services with Unaffiliated Foreigners from BEA. Ireland is included in all other EU and is not identified in BEA’s tables. These export and import transactions with unaffiliated foreigners are interfirm transfers, which are tradi- tional trades. Note that “affiliated foreigners” are locally established affiliates of multinational firms. The Asian Tigers consist of Korea, Singapore, Taiwan, and Hong Kong. There are three things to notice about this table. First, at present the numbers are small relative to total U.S. trade in services: exports and imports of software services represent 3.3 and 1.0 percent of total exports and imports of services, respectively. Second, the United States maintains a positive overall balance in trade and services; moreover, over the period 1998-2004 exports of computer services grew at a faster rate than imports (7.86 vs. 4.48 percent aver- age annual growth rate [AAGR]). Third, although imports of computing services from India grew rapidly from 1994 to 2004, overall U.S. imports from India and the other software underdogs are small relative to other estimates. Data from other sources suggest that the U.S. data may underestimate imports of software services. An OECD estimate indicates that over 90 percent of Indian service exports to OECD countries are not accounted for in the data on service imports published by these countries (OECD, 2004). Other analyses report similar difficulties in tracking Indian software services exports to the United States. A recent General Accounting Office (GAO) report notes that, for 2002, the United States reported $240 million in unaffiliated imports of business, professional, and technical (BPT) services from India, whereas India reported about $6.5 billion in affiliated and unaffiliated exports in similar services categories (GAO, 2005).  For 2003, the United States reported $420 million in unaffiliated imports of BPT services from India, whereas India reported approximately $8.7 billion in affili- ated and unaffiliated exports of similar services to the United States. The bulk (40-50 percent) of the difference, according to the GAO, is because the United States does not count the earnings of temporary workers resident in the United States in services imports. Other sources include differences in coverage (e.g., embedded software is counted as exports of goods by the United States, or IT- enabled financial services are not classified as IT services by the United States), and because U.S. data do not indicate affiliated imports by country of origin. As noted earlier, services trade data do not capture intrafirm migration of software activity abroad. The BEA data on U.S. MNCs provide detailed informa- tion on the investment and production activities of U.S. companies abroad.   Affiliated trade occurs between U.S. parent firms and their foreign affiliates and between foreign- owned firms in the United States and their foreign parent. Unaffiliated trade occurs between U.S. entities and foreign entities that neither own nor are owned by the U.S. entity.

TABLE 1  Computer and Information Services with Unaffiliated Foreigners (million dollars) Years 1994 1998 2004 AAGR, 1998-2004 Computer and Computer and Royalties Computer and Royalties Computer and Royalties Information Information and License Information and License Information and License Services Services Fees Total Services Fees Total Services Fees Total Exports All countries 2,332 3,705 3,195 6,900 6,601 4,261 10,862 10.10 4.92 7.86 Canada 333 430 125 555 1,144 279 1,423 17.71 14.32 16.99 Europe 899 1,767 1,508 3,275 3,281 1,328 4,609 10.87 –2.10 5.86 Japan 177 306 724 1,030 327 1,568 1,895 1.11 13.75 10.70 Asian Tigers 117 200 … … 163 … … –16.34 … … Underdogs Brazil 48 136 ... ... 149 81 230 1.53 .... ... Israel 51 24 32 56 38 13 51 7.96 –13.94 –1.55 China 17 29 46 75 48 51 99 8.76 1.73 4.74 India 9 38 17 55 227 29 256 34.70 9.31 29.21 Imports All countries 286 1,494 498 1,992 2,002 589 2,591 5.00 2.84 4.48 Canada 34 589 9 598 1,189 12 1,201 12.42 4.91 12.32 Europe 122 259 449 708 400 562 962 7.51 3.81 5.24 Japan 20 41 26 67 15 1 16 –15.43 –41.90 –21.23 Asian Tigers 6 18 … … 31 … … 55.98 … … Underdogs Brazil 1 1 1 2 1 ... ... 0.00 ... ... Israel 0 9 2 11 7 3 10 –4.10 6.99 –1.58 China 2 6 ... ... 7 ... ... 2.60 ... ... India 7 100 ... ... 315 6 321 21.07 ... ... NOTE: Omitted cells include either transactions below $500,000 or data that were omitted to maintain confidentiality. AAGR, average annual growth rate. 61 SOURCE: BEA Data on U.S. International Trade in Services.

62 INNOVATION IN GLOBAL INDUSTRIES Table 2 shows that growth in employment in IT services and computer de- sign industries has been faster for foreign affiliates of U.S. firms than for their domestic operations (AAGR 5.1 vs. 3.9 percent) due to faster growth among foreign affiliates in computer design and related services. Financing of Software Products and Services Table 3 includes data on one of the inputs to software product and service firms: financial capital. It includes data on disclosed rounds of venture capital fi- nancing by year and by destination country as reported in the Venture Economics VentureXpert database. As is well known, venture financing exhibits significant yearly variation (e.g., Gompers and Lerner, 2006) and our data may not capture all venture financing rounds. However, some broad trends are suggested. First, similar to our data on inventive outputs (described in further detail later), the United States clearly dominates in inputs of financial capital to emerging soft- ware firms. However, based on data from 2002-2005, there is some evidence that rounds of venture financing to the software underdogs declined less from their 2000 peak than did financing to U.S. firms.10 However, there was an apparent decline in venture financing to these countries in 2005. In short, more years of data are needed to discern whether there is a trend of increasing venture capital financing to the software underdogs. Regional Trends in Packaged Software and Software Services In the previous section we showed that the United States represents the ma- jority of world sales in packaged software. However, other regions of the world have a large and increasing percentage of software services. In this section we discuss some regional trends that are partially responsible for the geographic variance in economic activity in packaged software and services. Software Producers in Europe and Japan In Western Europe, the software industry has long been dominated by cus- tom software development and software services (Malerba and Torrisi, 1996; Steinmueller, 2004). Table 4 shows sales of software products and IT services in the EU15 dur- ing 2003-2005.11 IT professional services such as consulting, implementation, 10  Thesoftware underdogs consist of India, Ireland, Israel, Brazil, and China. 11  TheEU15 comprised the following 15 countries: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, and the United Kingdom.

SOFTWARE 63 TABLE 2  Growth in Employment for Foreign Affiliates of U.S. Firms vs. Growth for All U.S. Establishments, Selected Industries, 1999-2002 1999 2002 AAGR Information services and data processing services Foreign affiliates of U.S. firms 104.5 132.0 8.1 All U.S. establishments 371.9 473.8 8.4 Computer system design and related services Foreign affiliates of U.S. firms 157.9 172.9 3.1 All U.S. establishments 997.0 1,061.3 2.1 Total Foreign affiliates of U.S. firms 262.4 304.9 5.1 All U.S. establishments 1,368.9 1,535.1 3.9 NOTE: AAGR, average annual growth rate. SOURCE: Data on foreign affiliates of U.S. firms from table on selected data for majority-owned nonbank foreign affiliates and nonbank U.S. parents in all industries, 2003. From BEA International Economic Accounts, U.S. Direct Investment Abroad: Financial and Operating Data for U.S. Multina- tional Companies. Data on all U.S. establishments from U.S. County Business Patterns data. TABLE 3  Disclosed Rounds of Venture Financing by Country, 1988-2005 (thousands of dollars) United States Other G-7 Underdogs All Other Total 1988 2,565 660 0 0 3,225 1989 15,000 2,465 0 0 17,465 1990 6,350 464 248 0 7,062 1991 1,100 0 0 0 1,100 1992 1,607 1,418 0 0 3,025 1993 15,247 582 0 0 15,829 1994 7,403 138 0 0 7,541 1995 14,340 0 0 0 14,340 1996 92,784 1,466 0 2,766 97,016 1997 242,873 0 0 7,049 249,922 1998 300,355 9,359 0 6,039 315,753 1999 1,068,310 68,011 28,666 21,102 1,186,089 2000 2,036,591 221,297 73,307 169,636 2,500,830 2001 460,911 83,944 32,256 16,629 593,740 2002 99,836 23,295 6,831 3,815 133,777 2003 173,205 14,607 15,251 167 203,230 2004 151,025 9,492 10,600 1,848 172,965 2005 138,428 2,000 2,000 59 142,487 SOURCE: Venture Economics VentureXpert database, and author’s calculations. Software in- cludes rounds of financing from software and e-commerce software firms. Dates are round date of financing.

64 INNOVATION IN GLOBAL INDUSTRIES TABLE 4  Sales of Software Products and IT Services in the EU15 2003 2004 2005 Average Growth (%) Software products 59,235 61,707 64,979 4.74 System software 30,944 32,537 34,536 5.64 Application software 28,291 29,169 30,443 3.73 IT services 112,472 116,149 120,913 3.68 Professional services 81,376 84,380 88,147 4.08 Support services 31,096 31,769 32,766 2.65 Total software 171,707 177,856 185,892 Percent services 52.67% 53.13% 53.74% SOURCE: European Information Technology Observatory (2006). and operations management are larger than the entire software products market. Malerba and Torrisi (1996) identify several reasons for this focus on software services, including a weak local IT hardware industry, first-mover advantages by U.S. software product firms, fragmentation of local demand, and relatively little interaction between European universities and industry. The largest European producer of packaged software is SAP, the producer of enterprise software. SAP is currently the third largest software product company by sales, behind Microsoft and Oracle. One surprising result in Figures 1 and 2 is that, in contrast to many other technology industries, Japanese firms account for a very small share of the total export market for packaged software. This is not a recent result; Japanese firms have not ever been major players in the world market for packaged software, despite their success in video games and in other IT markets. Japan runs a sig- nificant negative trade imbalance in software: In 1997, Japan imported US$3.93 billion of software but exported only US$23.33 million (Asahi Shimbun, reported in Anchordoguy, 2000). A number of reasons have been provided for the relative weakness of Japa- nese software producers, including challenges created by the Japanese language, weak venture capital markets, weakness in intellectual property protection, and weak university computer science education (Anchordoguy, 2000; Baba et al., 1996; Cottrell, 1996; Fransman, 1995). Cottrell (1996) argues that weakness in Japanese PC software production was due historically to a fragmented standards environment, while Anchordoguy (2000) argues that the aforementioned proxi- mate reasons were ultimately caused by Japan’s economic system of “catch-up capitalism.”12 12  particular, she argues that some of the key elements of the Japanese economic system—includ- In ing state targeting policies, its keiretsu industrial groups, bank-centered financial system, and weak intellectual property system—have been benefited by its development of successful industries in steel, semiconductors, and IT hardware but have hindered the development of its IT software industry.

SOFTWARE 65 Other Countries That Are Large Software Producers Rapid growth in the size of the Indian software industry has recently at- tracted much attention in the academic and popular press (e.g., Athreye, 2005a; Arora et al., 2001). Data from NASSCOM show that Indian IT services exports grew from $22 million in 1984 to $10 billion in 2005, with an additional $3 bil- lion due to R&D services, engineering services, and software products. As this makes clear, the Indian software industry has largely been built around software services rather than products. Athreye (2005a) estimates that in 2000, revenue per employee among Indian software firms was approximately $35,100, up from only $6,200 in 1993. Some anecdotal evidence suggests that Indian firms are increasingly per- forming more R&D-intensive activities. Athreye (2005a) notes the growth of a new innovative sector of small niche companies. Moreover, there is evidence of a deepening of R&D skills and the emergence of informal networks among local firms in India. This is also some evidence of success in certain niche technolo- gies such as wireless and embedded systems (Parthasarathy and Aoyama, 2006; Ilavarasan, 2006); software for mobile phones represents a substantial category. Some Indian firms have also had success in developing software products for the developing countries market: one example is CITIL (now i-flex), a Citibank sub- sidiary that initially produced software products for developing country markets before eventually moving on to head-to-head competition with the established incumbent producers in developed countries (Arora, 2006; Athreye, 2005b). There are also some data on substantial and growing R&D activities in countries such as India; Arora (2006) reports that total revenues for engineering services and R&D by Indian producers in 2006 were estimated to be US$4.8 billion, a 23.1 percent increase over the prior year. In the next section we attempt to shed some additional light on this issue by examining U.S. patent data. The Irish software industry consists of two very separate subindustries, each with very different characteristics. First, there is an overseas sector that is domi- nated by MNCs. These firms primarily are engaged in software logistics (such as media replication and printing and packaging production and distribution), localization (such as translating and adapting software to suit European markets), and development (O’Riain, 1999). Second, there is an indigenous sector that is populated by smaller firms that is engaged in software development and product development activities. The number of MNCs in Ireland grew rapidly throughout the 1990s, from 74 foreign firms in 1991 to 140 foreign firms in 2000. As Arora, Gambardella, and Torrisi (2004) note, this rapid growth was due to a number of factors, in- cluding the liberalization of economic policies that began in 1991, a large and well-educated English-speaking workforce, an advantageous site for localization activities, as well as potential agglomeration economies that were ignited after the Irish software-producing industry reached sufficient scale. MNC subsidiaries

66 INNOVATION IN GLOBAL INDUSTRIES are engaged primarily in “low-value-added, low-skill activities such as porting of legacy products on new platforms, disc duplication, assembling/packaging, and localization” (Arora et al., 2004). Revenues and exports in the Irish software industry have always been dominated by these MNCs. Sands (2005) notes that total industry revenues grew from $2.66 billion in 1991 to over $18 billion in 2002, with MNCs continuously accounting for over 90 percent of the total. In contrast, the indigenous sector is more product-based: it accounts for just under half of employment; however, it accounts for only 9 percent of revenues. Indig- enous companies are usually young and small, and often produce primarily for niche or vertical (i.e., industry-specific) markets (Sands, 2005). The software industry in Israel looks considerably different from that in either Ireland or India. Compared to locally owned Indian or Irish firms, Israeli firms are more product-based and are more R&D intensive. Breznitz (2005) notes that revenue per employee for Israeli software firms was US$255,172 in 2000. By his calculations, the similar statistic in 2000 for U.S. software publishers was US$231,621 and for locally owned Irish software producers was US$90,000. Breznitz (2005) examines the reasons for Israel’s product-based industry. He provides several reasons: tight links between the R&D activities of Israeli univer- sities and high-tech industries in the country; the presence of a highly successful indigenous hardware industry; the presence of local market demand for new products; the presence of American MNCs locating R&D facilities in Israel; and the ability of the Israeli IT industry to raise capital in U.S. financial markets. EMPIRICAL EVIDENCE ON THE LOCATION OF INVENTIVE ACTIVITY In this section we examine the global geographic distribution of inventive ac- tivity in software. The data presented in the preceding section pointed to expand- ing markets for software services abroad. Those data also show that the market for packaged software continues to be highly concentrated in the United States, and little evidence indicates that this trend is reversing. However, authors such as Athreye (2005a) report increasing inventive activity in Indian firms, and other authors have reported similar trends in Ireland (Sands, 2005) and China (Tschang and Xue, 2005), as well as well-established software product industries in Israel (Breznitz, 2005) and Brazil (Botelho et al., 2005). Software product sales are a lagging indicator of inventive activity in software: Could inventive activity in software be picking up in other areas of the world but not yet reflected in product sales? If so, how significant are these developments in terms of number of inven- tions and their importance? To answer these questions, one needs a measure of R&D and inventive activity that is comparable across countries. Patent data have long been used as one measure of inventive activity. Patents have also been found to be correlated, although weakly, with R&D spending, so they provide a weak measure of raw inputs into innovation (Griliches, 1990).

SOFTWARE 67 There are, of course, significant limitations to the use of software patents as a measure of inventive activity. As Jaffe and Trajtenberg (2002) note, not all inventions meet the U.S. Patent and Trademark Office (USPTO) criteria for pat- entability,13 and inventors must make an explicit decision to patent an invention, as opposed to relying on some other method of intellectual property protection. In particular, there may be incremental inventive activity that is not patented and therefore is not reflected in patent statistics. Moreover, firms may sometimes choose to use trade secrecy rather than patenting to protect groundbreaking inven- tions because of incomplete enforcement of property rights. To the extent that intellectual property regimes differ across countries, this may make comparison of levels of patents across countries more difficult. Conversely the high growth rate of patenting that we observe in our sample may be influenced by strategic patenting behavior. As Hall and Ziedonis (2001) document in the semiconductor industry, firms may patent not to protect stand- alone technological inventions but rather to protect against holdup by external patent holders or to negotiate access to external technologies. Thus, when inter- preting our results, readers should be aware of how patent statistics may deviate from the level of inventive activity across countries. However, so long as the propensity to patent does not change significantly over time, these biases should not appreciably affect our interpretation of the time trends of patenting behavior across countries (and their interpretation as a metric of inventive activity). Historically, inventions in software were not patentable14 and for a time copyright was the predominant form of formal intellectual property protection in software. However, a series of court decisions widened the scope of software patents. Eventually, this culminated in the Commissioner of Patents issuing guidelines for the patenting of software that allowed inventors to patent any software embodied in physical media (Hall and MacGarvie, 2006). In contrast, over the same period, a series of cases, including several copyright infringement cases brought by Lotus Development, weakened the intellectual property pro- tection offered by copyrights. Graham and Mowery (2003) show that over this period the number of granted software patents has increased dramatically while the propensity of firms to copyright has declined.15 Recent research has shown that the stock of patents is correlated with firm success in the software industry 13  Note that not all inventions also meet the criteria for patentability for the European Patent Office (EPO) and Japanese Patent Office (JPO). 14 The following provides a necessarily brief overview of the history of intellectual property pro- tection in software. For a more detailed overview, see Graham and Mowery (2003) and Hall and MacGarvie (2006). 15 The set of patentable inventions is narrower in Europe than in the United States. To be patent- able, the European Patent Convention requires that inventions address a particular technical problem and suggest a technical means to solve this problem (Thoma and Torrisi, 2006). The implication of this requirement is that “inventions having a technical character that are or may be implemented by computer programs may well be patentable” (EPO, 2005).

68 INNOVATION IN GLOBAL INDUSTRIES (Merges, 2006), suggesting that patents may be a potentially useful metric of the inventive output of firms. A second issue in using software patents to measure inventive activity in software is identifying exactly which patents are software patents.16 Software patents are not assigned to a particular class or subclass in either the USPTO or International Patent Classification (IPC) schemes. Moreover, there is no unique software classification field for patents. Graham and Mowery (2003) were the first to systematically identify software patents for research purposes. They identified the IPC classes used by the six largest producers of PC software over the period 1984 to 1995. This search resulted in a list of 11 IPC classes, which account for over half (57 percent) of the more than 600 patents assigned to the 100 largest packaged software firms in 1995 (as identified in the trade news publication Softletter). The Graham-Mowery approach of using the patent classification system to identify software patents has been used and revised by others. Graham and Mowery (2005) identify software patents using USPTO classifications. Hall and MacGarvie (2006) identify software patents by finding the USPTO class-subclass combinations in which 15 large software firms patent. To identify their final sample, they intersect the resulting set of patents with another keyword definition used by Bessen and Hunt (2004). Bessen and Hunt (2004) identify software patents through the use of a Bool- ean query that searches for keywords in the text of patents. They arrive at a patent sample that is broader than that used by other researchers (Layne-Farrar, 2005). Other researchers have identified a smaller sample of patents by reading them manually. Allison and Tiller (2003) identify Internet business method patents and Allison et al. (2005) identify university software patents.17 For this chapter, we use a version of the Graham-Mowery approach based on the IPC system. We began by identifying the top 10 firms by revenue volume in 1995 according to the Corptech Directory of Technology Companies.18 We then examined the IPC classes in which they patented. Because we found that the Gra- ham-Mowery set of IPC classes covered only 46 percent of the patents of these top 10 firms, we added two additional IPC categories. Our complete list of patent classes covered more than 80 percent of the patents of these top 10 firms. Table 5 provides a list of the included IPC classes and subclasses and their descriptions. 16  This section provides an overview of the issues in identifying software patents. For a more com- plete discussion, see Layne-Farrar (2005) and Hall and MacGarvie (2006). 17  Thoma and Torrisi (2005) compare several of these methods in a study of European software patents. 18  These are Adobe, Autodesk, Cadence, Macromedia Inc., Microsoft, Novell, Oracle, SAP, Sybase, and Symantec Corp. Note that Corptech’s coverage of foreign firms is more limited than its coverage of U.S. firms; however, so long as the distribution of patent classes used by software patenting firms does not vary substantially for U.S. and non-U.S. firms, this issue should not appreciably influence our results.

SOFTWARE 69 TABLE 5  List of IPC Patent Classes Used in Analyses Class/Subclass Description G06F 3/00 Input arrangements for transferring data to be processed into a form capable of being handled by the computer; output arrangements for transferring data from processing unit to output unit (e.g., interface arrangements) G06F 5/00 Methods or arrangements for data conversion without changing the order or content of the data handled G06F 7/00 Methods or arrangements for processing data by operating upon the order or content of the data handled G06F 9/00 Arrangements for program control (e.g., control unit) G06F 11/00 Error detection; error correction; monitoring G06F 12/00 Accessing, addressing, or allocating within memory systems or architectures G06F 13/00 Interconnection of, or transfer of, information or other signals between memories, input/output devices, or central processing units G06F 15/00 Digital computers in general G06F 17/00 Digital computing or data processing equipment or methods, specially adapted for specific functions G06K 9/00 Methods or arrangements for reading or recognizing printed or written characters or for recognizing patterns (e.g., fingerprints) G06K 15/00 Arrangements for producing a permanent visual presentation of the output data G06T 11/00 Two-dimensional image generation (e.g., from a description to a bit-mapped image) G06T 15/00 Three-dimensional image rendering (e.g., from a model to a bit-mapped image) G09G 5/00 Control arrangements or circuits for visual indicators common to cathode-ray tube indicators and other visual indicators H04L 9/00 Arrangements for secret or secure communication NOTE: Class names are as follows: G06F, Electric Digital Data Processing; G06K, Recognition of Data, Presentation of Data, Record Carriers, Handling Record Carriers; G06T, Image Data Processing or Generation, in General; G09G, Arrangements or Circuits for Control of Indicating Devices Using Static Means to Present Variable Information; H04L, Electric Communication Technique. SOURCE: International Patent Classification System, World Intellectual Property Organization, http://www. wipo.int/classifications/ipc/ipc8/?lang=en. By using a broader set of IPC classes than Graham and Mowery, we are more likely to include patents that may be assigned to the aforementioned classes but are not software patents. As we will see, software patenting outside of the United States is relatively rare, so we utilize a conservative definition that includes as many such patents as possible in hopes of achieving an “upper bound” on the stock of software patents invented outside of the United States. However, we recognize that, if the rate of patenting in related technologies outside of software is higher than that inside and if the share of inventive activity in these other tech- nologies is higher in the United States than abroad, then our measure may artifi- cially inflate the gap in software patenting between the United States and other nations. To address this possibility, we compare our results using several software patenting definitions, including those of Graham and Mowery (2003, 2005).

70 INNOVATION IN GLOBAL INDUSTRIES 70 65 Percent U.S. Patents 60 55 50 45 1988 1992 1996 2000 2004 Year GM 03 GM 05 AFY FIGURE 4  Percent of U.S. patents invented in United States under different software software-4.eps definitions. SOURCE: USPTO data and authors’ calculations. As an illustration, we computed the percentage of patents produced by in- ventors who reside in the United States (regardless of assignee location) under different definitions and then compared them. Figure 4 presents these results. All three definitions show similar percentages for U.S. patents. Moreover, the three definitions have similar trends: increas- ing throughout the 1990s before reaching a peak around 2000 before declining slightly. Given the similarity in results across these different definitions, we will continue to focus on our original definition described earlier. Results of Patent Data Analysis Figure 5 shows the number of U.S. patents invented in the United States, Ja- pan, other G-7, and all other nations (based on inventor address) by year of patent grant. The steep increase in the number of patents granted after 1995 is consistent with prior work that has shown an increase in the propensity to patent software after increases in the scope of intellectual property rights afforded by software patents (Graham and Mowery, 2003; Hall and MacGarvie, 2006). In 2004, 4,695 software patents were issued to inventors in the United States—a larger number of patents than inventors from all other areas of the world combined (2,811). The average annual growth in software patenting between 1988 and 2004 was also greater in the United States than in all other G-7 nations: patenting by U.S. inven-

SOFTWARE 71 5000 4000 Number of Patents 3000 2000 1000 0 1988 1992 1996 2000 2004 Year Japan Other G-7 All Others US FIGURE 5  U.S. software patents invented in United States and other countries. SOURCE: software-5.eps USPTO data and authors’ calculations. tors grew at an average annual rate of 19.5 percent, compared to 16.1 percent for inventors in Japan and 18.0 percent in other G-7 nations. These figures may reflect a “home country bias”: U.S. firms may be more likely to patent in the U.S. market than foreign firms. Thus, in our data on patent- ing by location of inventor, the high percentage of U.S. patents may reflect (1) higher rates of U.S. patenting by U.S. firms (compared to firms in other countries) and (2) a higher propensity for U.S. firms to invent in the United States. More broadly, there may be some concern that there are potential differences between the site of inventive activity in U.S.-assigned U.S. patents that have EPO or JPO equivalents and the site of inventive activity in U.S.-assigned U.S. patents that do not have such equivalents. We address this potential concern in two ways. First, we look at the location of inventive activity for patents assigned to firms from outside of the United States. Second, we compare our results to recent work that has examined software patenting behavior in European patents. We examined the percentage of patents assigned to the home country by country of assignee firm, based on year in which the patent was granted. Figure 6 shows that Japan-assigned U.S. software patents are predominantly invented in Japan, although this share appeared to decline during 2000-2004. Similarly, the location of invention in Israeli- and G-7-assigned patents (excluding the United

72 INNOVATION IN GLOBAL INDUSTRIES 100 Percent invented in home country 95 90 85 80 1988 1992 1996 2000 2004 Year US Japan Israel G-7 Excluding US and Japan software-6.eps FIGURE 6  Percentage of U.S. software patents invented in home country by country of assignee. SOURCE: USPTO data and authors’ calculations. States and Japan) is predominately sited in those countries and regions. To be clear, comparing the propensity of U.S. software patents assigned to U.S. firms to be invented in the United States with the propensity of U.S. software patents assigned to firms from other countries to be invented in that (home) country is not an “apples to apples” comparison. However, given this important caveat, this figure does not suggest that patents assigned to U.S. firms are significantly more likely to be invented in the home country (United States) than are the patents from other countries. In fact, for several years, the U.S. patents assigned to Japanese and Israeli firms were more likely to be invented in the home country than U.S. patents assigned to U.S. firms. In recent years, however, this “home” percentage has been higher for patents assigned to U.S. firms than for others, though this is largely attributable to a decline in the home invented share for patents assigned to firms from other countries. Thoma and Torrisi (2006) examine the rate of software patenting in Euro- pean patents. Figure 7 shows the number of patents granted by country of patent assignee and year of patent application. There are some differences in the way Thoma and Torrisi define software patents and other differences in their sample construction: in particular, Thoma and Torrisi examine the distribution of patent-

SOFTWARE 73 12000 10000 8000 6000 4000 2000 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 DE FR GB IT US JP Others FIGURE 7  European Patent Office software patent grants by country of the assignee and year of application. SOURCE: Thoma and Torrisi (2006). software-7.eps bitmap image with type replaced ing activity by site of assignee rather than inventor.19 However, the broad trends are very similar to those in Figure 5: U.S. firms are responsible for the majority of software patenting activity, followed by Japanese firms, and then all others. Moreover, Thoma and Torrisi (2006) note that, of the European software patents in their database, 80.3 percent have also been granted by the USPTO and 73.8 percent have also been granted by JPO. If the majority of European software patents assigned to U.S. firms are also invented in the United States (not an un- reasonable assumption given evidence presented in the earlier paragraph that the majority of U.S. software patents assigned to U.S. firms are invented in the United States), then the graph suggests that, even using European software patent data, a large share of the inventive activity in software takes place in the United States. Further, we note that while the levels of software patenting expressed in Figures 4 and 5 may be influenced by home country bias, so long as this bias does not change systematically over time, the time trends shown in these figures will not be as influenced by such bias. 19  In particular, Thoma and Torrisi use a variant of Hall and MacGarvie’s (2006) method of con- structing a software patent sample based on patent classes and Bessen and Hunt’s (2004) keyword method. Moreover, this graph shows patenting by assignee country rather than inventing country; however, according to our data 93.4 percent of patents assigned to U.S. firms were also invented in the United States. Last, this figure shows patenting by year of application rather than year of granting; however, the broad trend of greater patenting among U.S. assignees is robust to this difference.

74 INNOVATION IN GLOBAL INDUSTRIES 100 80 Number of Patents 60 40 20 0 1988 1992 1996 2000 2004 Year Brazil India China Ireland Israel FIGURE 8  Number of U.S. software patents invented in underdog countries. SOURCE: USPTO data and authors’ calculations. software-8.eps Patenting Activity by Region Figure 8 shows the number of U.S. patents invented in the underdog coun- tries based on inventor location. Israel is the only one among them to have a significant number of U.S. patents. Israeli patenting activity increased from 3 in 1998 to a high of 90 in 2003. No other country has had more than 20 patents in any one year, though the number of patents invented in India has risen slightly in recent years, from an average of 0.5 throughout the 1990s to 16 in 2004. Figure 9 shows the number of patents invented in the East Asian Tigers based on inventor location.20 The number of patents invented in these countries is significantly higher than that of the underdogs. However, evidence suggests that many of these patents may be related to electronics.21 Patenting among these countries is dominated by inventors from indigenous electronics companies in Korea and Taiwan: in 2004, 264 of the 280 patents granted were from this set of assignees. 20  For the purposes of this paper, the Asian Tigers consist of Korea, Taiwan, Singapore, and Hong Kong. This is a separate and distinct set from the software underdogs. 21  The top patenting firms in these countries include Daewoo Electronics Co. Ltd. (33), Electron- ics and Telecommunications Research (60), Hyundai Electronics Industries Co. Ltd. (57), Industrial Technology Research Institute (55), Inventec Corporation (25), LG Electronics (102), and Samsung Electronics (463). All of these companies are heavily involved in electronics research.

SOFTWARE 75 150 100 Number of Patents 50 0 1988 1992 1996 2000 2004 Year Korea Taiwan Singapore Hong Kong FIGURE 9  Number of U.S. software patents invented in East Asian Tigers. SOURCE: software-9.eps USPTO data and authors’ calculations. Assignee Location for Patents Invented Abroad As noted earlier, multinational firms have played a major role in the devel- opment of software industries in other countries such as India and Ireland and may be driving the patenting activity by overseas inventors. To investigate this question further, we examined the location of U.S. software patent assignees for U.S. software patents invented in different countries. The overwhelming majority of patents invented in the United States were also assigned to U.S. firms. This fraction ranges from 93 to 97 percent over the period 1988-2004. No other region ever exceeded 6 percent in these data. Figure 10 shows the distribution of assignee country for patents invented in the underdog countries. Here, the fraction of patents assigned to U.S. firms has generally been increasing over time, ranging from 20 percent in 1990 to a high of 65.7 percent in 2002. Excluding Israel (which has a robust software product industry) from the software underdogs, the top assignees in the software under- dogs are 3Com (12), IBM (25), and Texas Instruments (12); no other company has more than five patents. The percentage of patents invented in underdog na- tions that are assigned to underdog firms has similarly been declining over time, from 80 percent in 1990 to 32.7 percent in 2004. The increasing share of patents invented abroad in one of the software un-

76 INNOVATION IN GLOBAL INDUSTRIES 100 Percentage of 80 Total Patents 60 40 20 0 1988 1992 1996 2000 2004 Year Percentage of Total Patents 100 80 60 40 20 0 1988 1992 1996 2000 2004 Year Asian Tigers Other G-7 Underdogs All Others FIGURE 10  Distribution of assignee country for U.S. software patents invented in soft- ware underdogs. Top panel, United States; bottom panel, all other countries. software-10.eps derdogs but assigned to U.S. firms suggests that there may be some shift in the location of inventive activity for U.S. firms to offshore locations. There is some evidence of a shift to more offshore invention for patents assigned to U.S. firms. However, the shift is small and offshore software invention in underdog countries by U.S. firms accounts for a very small share of the total patents assigned to U.S. firms. We also examine the trends in the site of inventive activity for U.S. software patents assigned to U.S. firms; note that these trends, because they only examine the site of inventive activity for patents assigned to U.S. firms, are not subject to concerns of home country bias. The percentage of U.S.-assigned patents invented in the United States fell from 93.5 percent in 1996 to 92.1 percent in 2005.22 This decrease in the share of U.S. patents is due in large part to the increase in offshore activity in the underdogs: the percent of U.S.-assigned patents invented in the underdogs rose from 1.1 percent in 1996 to 1.8 percent in 2005. We next examined whether there were any systematic differences in the industrial classification of the patent assignees by region where the patent was in- 22  The share of U.S.-assigned patents invented in the United States was 96.2 percent in 1988 and 93.0 percent in 1989, though the number of software patents in these years was much lower than in 1996 (260 in 1988 and 387 in 1989 compared to 1,519 in 1996) which, as described earlier, was one of the first years in which software patenting began to grow rapidly.

SOFTWARE 77 TABLE 6  Assignee Industry for U.S. Patents by Region, 1988-2005 Electrical & Industrial Electronic Holding Software Machinery Equip. Companies Publishers (SIC 35) (SIC 36) (SIC 67) (SIC 7372) All Other United States 9,741 5,291 4,770 2,217 13,434 (27.48) (14.92) (13.45) (6.25) (37.89) Other G-7 1,023 153 12,561 98 3,585 (5.87) (0.88) (72.11) (0.56) (20.58) Asian Tigers 47 51 781 1 818 (2.77) (3.00) (46.00) (0.06) (48.17) Software underdogs 133 136 56 59 354 (18.02) (18.43) (7.59) (7.99) (47.97) All other 85 44 609 15 570 (6.42) (3.33) (46.03) (1.13) (43.08) SOURCE: Author’s manipulation of data from USPTO and Corptech Database of Technology Com- panies. Numbers represent frequencies of row and column combinations. Numbers in parentheses represent the percentage of assignees in an industry conditional on invention in the country in the row. The unit of observation in the table is a patent. vented. To do this, using assignee (company) name, we merged our U.S. software patent data with the Corptech Database of High Technology Companies. Table 6 shows the distribution of assignee industry for patents by inventor region. Due to the way the Corptech data are collected, the industries of many Asian and G-7 countries in the Corptech database are classified as holding com- panies, so we focus our analyses on patents invented in the United States and in the underdog countries. One fact that is immediately apparent across all rows of the table is that patents are assigned to companies that belong to a variety of industries. Outside of the “Holding Company” category, most patents are from the “Other” category. Moreover, most patents are not assigned to firms in the “Software Publishers” industry (SIC 7372)—the SIC industry for packaged software producers. Second, the distributions of industries in the United States and underdog countries are broadly similar, with U.S. firms slightly more likely to be in industrial machinery and equipment and the underdogs more likely to be in electronics. In Table 7 we provide some descriptive statistics on top patenting firms in major software-producing countries. To construct this table, we identified the five firms with the largest number of U.S. patents in each of nine countries: China, Germany, the United Kingdom, Ireland, Israel, India, Japan, South Korea, and the United States. Two major facts emerge. First, as noted earlier, the top patent- ing firms in software are usually not packaged software producers. Second, the top patenting firms in the underdog countries are usually large U.S. producers of

TABLE 7  Leading Recipients of U.S. Software Patents, by Country of Inventor, 1988-2005 78 Year Company was Number of Number Name Founded Industry Employees Revenue Home Country of Patents CHINA Microsoft 1975 Software 71,553 $44 billion United States 6 IBM 1888 IT hardware, software, services 330,000 $91 billion United States 5 United Microelectronics Corp. 1980 Electronics 12,000 Taiwan 4 Intel 1968 Electronics 99,900 $39 billion United States 2 Huawei Technologies 1988 Telecommunication 44,000 $8.2 billion China 1 GERMANY Siemens 1847 Conglomerate 472,000 $75 billion Germany 252 Robert Bosch GmbH 1886 Automotive 251,000 $55 billion Germany 178 IBM 1888 IT hardware, software, services 330,000 $91 billion United States 98 Infineon Technologies 1999 Electronics 36,000 $7 billion Germany 41 Daimler Chrysler AG 1998 Automotive 383,000 $150 billion Germany 38 UNITED KINGDOM IBM 1888 IT hardware, software, services 330,000 $91 billion United States 140 International Computers Limited 1968 Computers United Kingdom 40 British Telecommunnications PLC 1846 Telecommunications 104,400 $37 billion United Kingdom 38 Sun Microsystems Inc. 1982 IT hardware 31,000 $11 billion United States 35 Philips Corporation 1891 Electronics 159,226 $36 billion The Netherlands 32 IRELAND 3Com Corporation 1979 Networks 1,925 $800 million United States 11 Analog Devices Inc. 1965 IT Hardware 8,800 $2.4 billion United States 3 Richmount Computers Limited 3 Hitachi Ltd. 1920 IT hardware, electronics 323,072 $80.5 billion Japan 3 IBM 1888 IT hardware, software, services 330,000 $91 billion United States 3

ISRAEL IBM 1888 IT hardware, software, services 330,000 $91 billion United States 69 Intel 1968 Electronics 99,900 $39 billion United States 58 Motorola Inc. 1928 Electronics 88,000 $37 billion United States 32 Scitex Corporation (now Scailex IT hardware, now venture capital $128.2 million Israel 12 Corporation)a Applied Materials Inc. 1967 Semiconductor 12,576 $7 billion United States 11 INDIA IBM 1888 IT hardware, software, services 330,000 $91 billion United States 17 Texas Instruments 1930 Hardware 30,300 $13 billion United States 12 Honeywell International Inc. 1886 Aerospace 116,000 $26 billion United States 3 Veritas Operating Corporation (acquired by 1989 Software 16,000 $4.1 billion United States 3 Symantec)b Sun Microsystems Inc. 1982 Hardware 31,000 $11 billion United States 2 JAPAN Hitachi Ltd. 1920 IT hardware, electronics 323,072 $80.5 billion Japan 1,403 Canon 1937 Imaging 100,000 $35 billion Japan 1,286 Fujitsu 1935 Hardware 158,000 $40 billion Japan 1,127 NEC Corporation 1899 Electronics 148,540 $41 billion Japan 976 Toshiba 1904 Electronics 165,000 $60 billion Japan 820 SOUTH KOREA Samsung 1938 Electronics $80 billion South Korea 460 LG Electronicsc 1958 Electronics 66,614 $23.5 billion South Korea 100 Electronics and Telecommunications 1976 South Korea 55 Research Institute Hyundai Electronics (now Hynix Semiconductors 13,000 $5.6 billion South Korea 49 Semiconductor) Hyundai Motor Company 1967 Automotive 51,000 $57 billion South Korea 30 79 continued

TABLE 7  Continued 80 Year Company was Number of Number Name Founded Industry Employees Revenue Home Country of Patents UNITED STATES IBM 1888 IT hardware, software, services 330,000 $91 billion United States 4,981 Intel 1968 Electronics 99,900 $39 billion United States 1,648 Microsoft 1975 Software 71,553 $44 billion United States 1,136 Sun Microsystems Inc. 1982 Hardware 31,000 $11 billion United States 1,088 Hewlett-Packard Inc. 1939 Hardware 150,000 $89 billion United States 682 aDataare for Scailex. bDataare for Symantec. cLG Electronics and LG Semicon Co. Ltd. are each part of the LG Group. LG Phillips is a joint venture with the LG Group and Philips. Data missing are because data on some subsidiaries of the LG Group are not separately available. SOURCE: The top five firms with the largest number of U.S. patents, identified from our calculations of USPTO data. Company data are from Hoover’s Online, company annual reports, company web pages, and Wikipedia. Revenues are in U.S. dollars and for the most current year available. Missing cells represent firms for which we were unable to recover data.

SOFTWARE 81 electronics—and to a much lesser extent European and Japanese producers—such as IBM, Intel, Texas Instruments, and Sun Microsystems. One exception is China, where one Taiwanese and one Chinese firm are included among the leading pro- ducers. However, as noted earlier, the number of U.S. software patents produced in China is very small. U.S. MARKET ADVANTAGES FOR INNOVATIVE ACTIVITY The data in the prior two sections show two very different stories occurring in the globalization of software activity. On one hand, as has been well docu- mented, there has been increasing growth in the production of IT service activity outside the United States. This trend has been going on for some time now and shows no signs of abating. Second, there is evidence that inventive activity in software development (at least as measured by patents) is highly concentrated in the United States and heavily controlled by U.S. firms. Though there is some evidence that inventive activity is picking up outside the United States, at current rates of growth this activity will not catch up with the U.S. software industry any time soon. Moreover, though there is some evidence that some inventive activity by U.S. firms has shifted abroad, at present the shift is small and this remains a small share of U.S. firms’ overall inventive activity. However, these trend rates of growth can change, so it is useful to examine the conditions that are widely thought to be conducive to innovation and inven- tive activity in new technologies. The literature has long examined some of the factors influencing the variance in innovative activities across countries.23 These include R&D investments and human capital (e.g., Romer, 1990), supportive pub- lic policies (e.g., Nelson and Rosenberg, 1994; Mowery and Rosenberg, 1998), and more localized factors supporting the growth of clusters, including spillovers and user-producer interactions (Porter, 1990). In general, the United States has advantages over other rich and poor countries in all of these dimensions. We focus our attention on one area that we believe has received insufficient attention: the importance of geographic proximity to lead user innovation. A key factor in the development and growth of a local software industry is the relationship with users. The transition of new inventions to usable economic products is a difficult process. Solving the problems that remain after initial con- ceptualization requires sustained innovative activity. User innovation and input are often an important part of this process (Rosenberg, 1963), and the willingness and ability of individuals to acquire and use new products and technologies is often as important as the developments of such products and technology them- selves (Rosenberg, 1983). Such user activity is particularly important in software. Business software in particular is often bundled with a set of business rules and assumptions about 23  For a recent overview and review of this literature, see Furman et al. (2002).

82 INNOVATION IN GLOBAL INDUSTRIES business processes that must be integrated with the existing business organiza- tion, its activities, and its processes. Recent research indicates that proximity between software developers and users is particularly important for this activity to occur. The software industry has a long history of user innovation and interac- tions with users leading to path-breaking new products. For example, IBM’s collaboration with American Airlines on the SABRE airline reservation system in the 1950s and 1960s was an important early use of information technology in “real-time” applications that would later be used in airline reservations, bank automation, and retail systems (Campbell-Kelly, 2003; Copeland and McKenney, 1988). The genesis of this project was a serendipitous event: the chance meet- ing on a flight of R. Blair Smith of IBM’s Santa Monica sales office with C. R. Smith, the president of American Airlines. The eventual outcome of this project was the SABRE system. Both IBM and American Airlines made extensive invest- ments and contributions to the project: “We tapped almost all types of sources of programming manpower. The control (executive) program was written by IBM in accordance with our contract with them. We used some contract programmers from service organizations; we used our own experienced data processing people; we tested, trained, and developed programmers from within American Airlines, and hired experienced programmers on the open market.”24 Similarly, the early development of ERP software by SAP occurred through a series of incremental improvements during development of real-time software for clients (Campbell- Kelly, 2003). One major challenge to offshoring software product development work will result from the difficulty of coordinating software development activity across a globally distributed team. As is well known, partitioning complicated software development projects across multiple team members is difficult and often sub- stantially increases the costs of software development (Brooks, 1995). These problems may become still greater when management of such projects is attempted at a distance. Globally distributed team members do not have access to the rich communication channels that co-located developers have. Moreover, differences in language and culture may make it much more difficult to establish common ground among team members and ensure that miscommunications do not occur (Armstrong and Cole, 2002; Olson and Olson, 2000). These projects face other challenges, including an inability to engage in informal communication as well as the difficulty of managing team members who may believe that such projects are a prelude to job cuts. A number of techniques have been proposed for lowering the costs of distrib- uted software development. Going back as far as March and Simon (1958), one common technique in distributed development is to reduce the interdependencies among software components. The increasing modularization of software code and 24  Parker (1965), as quoted in Campbell-Kelly (2003).

SOFTWARE 83 the use of object-oriented software development techniques has likely reduced some of the costs of distributed development over time. However, schedules and feedback mechanisms are necessary when interdependencies are unavoidable (March and Simon, 1958). The recent successes of large-scale open-source proj- ects such as Linux and Apache have led some to consider whether open-source project management methodologies could be utilized in traditional corporate software development. Globally distributed teams rely heavily on coordination tools such as e-mail, phone, and more recently instant messaging as well as con- figuration management tools. However, several authors have shown that initial meetings are often necessary both to detail project requirements and for project members to become familiar with one another (e.g., Herbsleb et al., 2005). In general, the literature has demonstrated that, despite the continued development of tools and techniques to manage distributed projects, globally distributed work is difficult and can involve significant coordination costs. Despite the considerable work that has been done in examining the chal- lenges of software project management in a distributed environment, there has been heretofore relatively little systematic widespread empirical evidence on how distance from software suppliers impacts firm decisions to offshore software development. Arora and Forman (2007) attempt to gather such systematic evidence by examining which IT services can be effectively performed from a distance or, to put it another way, which IT services are tradable. One way of examining the tradability of IT services is to examine the extent to which they are clustered near local demand. If markets for IT services are local, then we should expect the entry decisions of IT services firms to depend in part on the size of the local market. If markets are not local, then the composition of local demand should matter little; rather, suppliers should locate in low-cost regions. By providing evidence of the geographic reach of markets, this analysis also provides evidence on the tradability of services: Markets for services that are not tradable will be local, whereas those for services that are tradable need not be local. Arora and Forman examine the clustering of local market supply for two types of IT services: programming and design and hosting. “Programming and design” refers to programming tasks or planning and designing information systems that involve the integration of computer hardware, software, and com- munication technologies. These projects require communication of detailed user requirements to the outsourcing firm in order to succeed. Hosting involves man- agement and operation of computer and data-processing services for the client. 25 After an initial setup period, the requirements of such hosting services will be relatively static and will require relatively little coordination between client and 25  While hosting activities do not fit most definitions of “innovation” or “invention” in software per se, they do provide a useful benchmark to compare tradability of services that require complex communication and coordination between supplier and customer and those that do not.

84 INNOVATION IN GLOBAL INDUSTRIES TABLE 8  Average Outsourcing by Size of Metropolitan Statistical Area Programming Hosting Ex Programming (%) and Design (%) Internet (%) Rural Area 17.81 24.30 15.91 (0.38) (0.43) (0.37) Small MSA (< 250,000) 17.87 23.85 15.04 (0.54) (0.60) (0.50) Medium MSA (250,000 to 1 million) 18.48 26.30 16.41 (0.35) (0.40) (0.34) Large MSA (> 1 million) 18.54 26.08 15.31 (0.21) (0.24) (0.20) NOTE: Calculations are for 2002; standard errors in parentheses. The difference between rural/small and medium/large is significant at the 5 percent level for all three types. SOURCE: Arora and Forman (2007). service provider. Thus, ex ante we would expect that hosting activities may more easily be conducted at a distance than other activities. Using data from U.S. Cen- sus County Business Patterns, Arora and Forman (2007) find that the elasticity of local supply to local demand characteristics is higher for programming and design (0.806) than for hosting (0.1899). That is, a 10 percent increase in local market demand will translate into an 8.1 percent increase in the supply of programming and design firms but only a 1.9 percent increase in the supply of hosting firms. Arora and Forman also examine whether firm decisions to outsource pro- gramming, design, and hosting services depend on local market supply. Table 8 shows how 2002 outsourcing varied by the size of geographic area in the United States. Average outsourcing of programming and design is clearly increasing in the size of a location, though the pattern for hosting is less clear. Outsourcing of programming and design increases from an average level of 24.2 percent in small metropolitan statistical areas (MSAs) and rural areas to 26.1 percent in medium and large MSAs, and these levels are significantly different from one another at the 1 percent level. In contrast, outsourcing of hosting declines slightly from an average level of 15.61 percent in rural areas and small MSAs to 15.60 percent in medium and large MSAs; these levels are not statistically different from one other. Since the supply of outsourcing establishments is increasing in location size, these results suggest that the decision to outsource programming and design is increasing in the local supply of outsourcing firms. Controlling for industry differences, establishment size, and other factors yields the same conclusion. This evidence, combined with that on the costs of distributed software development described earlier, suggests that proximity to users is an important determinant of inventive activity in software. The contrast with other products and industries in this volume is informative. For other products, such as wireless devices or PDAs, lead users have significant concentration in locations outside

SOFTWARE 85 of the United States such as East Asia. However, the lead users of software are predominantly large organizations, and the leading large organizations in use of software and IT remain in the United States. This is especially true for the large market segment of business applications software, for which software products and services are frequently embedded in business process. User requirements in this setting often involve the transfer of tacit knowledge, and so proximity to lead users is particularly salient. Thus, as long as the United States remains the major market for software products, and the locus of the vast majority of lead users, it is unlikely to lose its technical leadership. SOME RECENT TRENDS AND PROJECTIONS FOR THE FUTURE Trends in Computer Science Education Continued success in any innovative industry like software requires a tal- ented and highly educated workforce. There is widely reported concern about a perceived shortage of domestic-born scientists and engineers in the United States (e.g., Ricadela, 2005). Figure 11 shows data from the National Center for Education Statistics (NCES) on the number of undergraduate and master’s degrees in computer sci- ence earned in the United States over the period 1983-2002.26 The numbers of both undergraduate and master’s degrees rose sharply from a combined figure of 35,200 in 1996 to 65,700 in 2002. This increase was influenced by the boom in the IT sector in the late 1990s. More recent indicators of undergraduate- and master’s-level enrollments in computer science are currently unavailable using official U.S. statistics. Figure 12 presents data from an annual survey of incoming freshmen. Mirroring the NCES statistics, these data show intention to major in computer science rising through- out the late 1990s and remaining high until 2001. However, intentions to major in computer science drop sharply thereafter. The Computing Research Association’s Taulbee Survey shows similar findings. These data survey Ph.D.-granting institu- tions in the United States. Aspray et al. (2006) argue that data from the Taulbee Survey closely match trends in the NCES data, and so these data are a good leading indicator of the national educational statistics. Figure 13 shows a sharp decline in newly declared computer science majors after 2000. Somewhat more recent official data are available for doctoral degrees con- ferred by U.S. universities. Figure 14 shows the number of doctoral degrees earned in computer science and mathematics during the period 1983-2003. In contrast to bachelor’s or master’s degrees, the number of doctoral degrees granted has generally been on the decline in the United States over the past decade. The 26  Data from the NCES and other official government statistics in this subsection are from the National Science Foundation publication Science and Engineering Indicators.

86 INNOVATION IN GLOBAL INDUSTRIES 70.00 60.00 50.00 Thousands 40.00 30.00 20.00 10.00 0.00 2000 2002 2001 1983 1984 1985 1986 1987 1988 1989 1990 1992 1991 1993 1994 1995 1996 1997 1998 Bachelor’s Master’s FIGURE 11  Undergraduate and master’s degrees earned in computer science. SOURCES: U.S. Department of Education, National Center for Education Statistics, Integrated Post- software-11.eps secondary Education Data System, Completions Survey; and National Science Foundation, Division of Science Resources Statistics, WebCASPAR database, http://webcaspar.nsf. gov. See appendix Table 2-26 from Science and Engineering Indicators 2006 for further details; 1999 data are not available. figure shows that the number of computer science Ph.D.s peaked in 1995 at about 1,000 and then has fallen over time. In 2003 the number of such degrees advanced slightly, from 810 to 870. However, due to the very long lag between entry and graduation in doctoral programs, this increase likely reflects enrollment decisions in the middle to late 1990s, when demand for computer scientists was particularly strong. While the number of students entering computer science programs appears to have fallen recently, there is evidence that such enrollments have been picking up in other countries. Figure 14 also shows the number of doctoral degrees granted in mathematics and computer science in selected countries other than the United States. The number of doctoral degrees in computer science and mathematics has recently been increasing in Asian countries such as China, Korea, and Taiwan. 27 Unfortunately, similar statistics are not easily available for the production of 27  These statistics, presented in Science and Engineering Indicators and collected from a variety of places, are unfortunately available only with some lag, and may not be strictly comparable. Moreover, they do not provide educational statistics on computer science graduates for India.

SOFTWARE 87 FIGURE 12  Freshman intentions to major in computer science. SOURCE: Globalization and Offshoring of Software: A Report of the ACM Job Migration Task Force (2006), eds. Aspray, Mayadas, and Vardi. FIGURE 13  Newly declared computer science majors. SOURCE: Computing Research Association and Globalization and Offshoring of Software: A Report of the ACM Job Migration Task Force (2006), eds. Aspray, Mayadas, and Vardi.

88 INNOVATION IN GLOBAL INDUSTRIES 2500 2000 1500 1000 500 0 1983 1985 1987 1989 1991 1993 1995 1997 1999 2000 2001 2002 2003 Thousands US Germany & UK China Korea & Taiwan FIGURE 14  Doctoral degrees in mathematics and computer science by region. SOURCES: China—National Research Center for Science and Technology for Development and Edu- software-14.eps cational Yearbook, 2002; Division of Higher Education, special tabulations (2005); South Korea—Organisation for Economic Co-operation and Development, Center for Education Research and Innovation, Education database, http://www1.oecd.org/scripts/cde/members/ EDU_UOEAuthenticate.asp; and Taiwan—Ministry of Education, Educational Statistics of the Republic of China (annual series). bachelor’s and master’s degrees. Gereffi and Wadhwa (2005) provide evidence on the number of bachelor’s and subbaccalaureate engineering, computer science, and IT degrees for the United States, India, and China in 2004. Figure 15 shows that the number of degrees awarded in engineering by India and the United States are roughly similar. Although the numbers of engineering graduates in China are much larger than that of either the United States or India, Gereffi and Wadhwa (2005) note that educational statistics on engineers from China include degrees from 2- or 3-year programs that include students graduat- ing from technical training programs that may be qualitatively different from baccalaureate programs in the United States. When normalized by population, the United States continues to lead in the production of bachelor’s degrees in engineering, producing 468.3 bachelor’s degrees per million compared to 103.7 in India and 271.1 in China. However, recent work by Arora and Bagde (2006) shows that the number of engineering baccalaureate degrees awarded in India is growing much faster than in the United States. Table 9 shows that, although the number of engineer-

SOFTWARE 89 700,000 600,000 500,000 292,569 400,000 300,000 200,000 84,898 103,000 351,537 100,000 137,437 112,000 0 U.S. India China Bachelors Degrees Subbaccalaureate Degrees FIGURE 15  Bachelor’s and subbaccalaureate degrees in engineering, 2004. SOURCE: Gereffi and Wadhwa (2005). software-15.eps ing baccalaureate degrees awarded in 2003 is roughly the same as that reported by Gerrifi and Wadhwa, this number has grown steeply over time. From about 42,000 in 1992, the total more than tripled to greater than 128,000 in 2003. 28 Moreover, since the number of baccalaureates produced reflects the capacity added with a 4-year lag, it is important to note that sanctioned engineering bac- calaureate capacity in India now exceeds 440,000, although a substantial portion is of dubious quality. Figure 16 shows that the number of foreign students enrolled in graduate computer science programs in the United States declined in 2003 for the first time since 1995, reflecting visa restrictions imposed after September 11, 2001, the growth in degree-granting programs in other countries, as well as declines in the demand for engineers and computer scientists that took place in the early years of the most recent decade (NSF, 2006). Overall, the data show that the United States continues to maintain a lead in the production of computer science graduates at all levels. However, recent data suggest that enrollments in computer science may be declining in the United States and picking up in other nations. As we will show in the next section, how- ever, these changes in domestic supply are likely not due to long-term declines in the demand for computer science graduates within the United States. 28  These numbers are based on data reported by 14 states, which include all the major states except Bihar, and probably represent 80-90 percent of the engineering baccalaureates produced in India.

90 INNOVATION IN GLOBAL INDUSTRIES TABLE 9  Output of Engineering Graduates (B.S. and B.E.) in India, Various Years Year Total Number of Engineering Graduates Produced 1990 42,022 1991 44,281 1992 46,762 1993 48,281 1994 52,905 1995 56,181 1996 57,193 1997 61,353 1998 67,548 1999 75,030 2000 79,343 2001 97,942 2002 107,720 2003 128,432 NOTES: These data are based on the figures for the 14 major states (except the State of Bihar) in India, which account for 80% of the gross domestic product and likely more than that number of the total production of engineering gradu- ates. These data are based on “Annual Technical Manpower Review” (ATMR) re- ports published by National Technical Manpower Information System (NTMIS), India. These reports are prepared by a state-level nodal center of NTMIS and give details of sanctioned engineering college capacity and outturn for all undergradu- ate technical institutions in the state. See cited source for more details. SOURCE: Arora and Bagde (2006). Labor Market Trends There is some evidence that growth in the number of computer science de- grees awarded over the past 25 years has not been fast enough to keep pace with demand for workers with computer science training. Figure 17 shows that the annual growth rate in the production of all mathematics and computer science degrees averaged 4.2 percent during the period 1980-2000, significantly less than the average annual growth of 9.3 percent in occupations directly associated with these fields.29 In comparison, over the same period, growth of all science and en- gineering graduates (including math and computer science) averaged 1.5 percent while growth in all science and engineering occupations averaged 4.2 percent (Table 10). Thus, the difference between degree growth and employment growth is larger in mathematics and computer science than it is for science and engineer- 29  Occupationaldata from these figures were compiled by the National Science Foundation, Division of Science Resources Statistics, from U.S. Census data.

SOFTWARE 91 30.00 25.00 20.00 Thousands 15.00 10.00 5.00 0.00 2000 2003 2002 2001 1990 1996 1984 1986 1988 1998 1999 1983 1985 1989 1994 1995 1992 1993 1987 1997 1991 Foreign Citizens FIGURE 16  U.S. graduate enrollment in computer science by citizenship. SOURCE: Science and Engineering Indicators 2006. software-16.eps Employment Doctoral Master’s Bachelor’s All 0 1 2 3 4 5 6 7 8 9 10 FIGURE 17  Average annual growth of degree production and occupational employment in mathematics and computer science, 1980-2000. SOURCE: Science and Engineering software-17.eps Indicators 2006.

92 INNOVATION IN GLOBAL INDUSTRIES TABLE 10  Output of Engineering Graduates (B.S. and B.E.) in India, Various Years Degree Growth Employment in Occupation All science and engineering 1.5 4.2 Mathematics/computer science 4.2 9.3 NOTES: Degree growth includes undergraduate, master’s, and doctoral degrees. SOURCE: Science and Engineering Indicators 2006. ing overall. These data are now several years old and do not account for students receiving degrees from outside of computer science but moving into computer science professions. However, despite these qualifications, they do suggest that the United States may have relied in part on workers from abroad to make up for the shortfall of native workers with computer and math skills. Recent data suggest that the inflation-adjusted median salaries for master’s graduates in mathematics and computer science rose 54.8 percent between 1993 and 2003, higher than any other broad class of science and engineering graduates and higher than the average across all non-science and engineering graduates. 30 Growth in salaries was similarly competitive for graduates with bachelor’s de- grees (28.0 percent AAGR, second only to engineering graduates among science and engineering graduates) and those with doctoral degrees (18.6 percent AAGR, second only to graduates in engineering and physical sciences among science and engineering graduates). Furthermore, 2003 median salaries for computer science master’s graduates are higher than any other broad category of science and engineering graduates ($80,000), whereas levels for bachelor’s ($50,000) and doctoral ($67,000) degree graduates remain similarly competitive. Thus, even when one uses data that include the recent technology downturn, salaries of occupations requiring skills in mathematics and computer science have remained quite competitive when compared to other occupations in science and engineering and compared to the national average. As noted earlier, there has been a significant shortfall in the rate of com- puter science degrees conferred relative to the rate of employment growth, and this excess demand for workers with computer science and engineering skills has been partially offset by the immigration of skilled workers from abroad. In fiscal year 2001 there were 191,397 H-1B visa admissions to the United States from computer-related occupations, 57.8 percent of total such admissions and 30  The source for these data is the National Science Foundation, Division of Science Resource Statistics, National Survey of College Graduates.

SOFTWARE 93 the largest of any such category.31 Kapur and McHale (2005a) list the top com- panies that petitioned for H-1B visas in October 1999 through February 2000, a list that includes some of the leading IT hardware and software firms: Motorola (618 petitions), Oracle (455 petitions), Cisco (398 petitions), Mastech (398), Intel (367), Microsoft (362), Rapidigm (357), Syntel (337), Wipro (327), and Tata Consulting (320). Changes in immigration represent one mechanism that has the potential to affect the U.S. software industry in the relatively short term, and recent changes in the environment outside the United States can potentially affect immigra- tion flows. The rapid growth in the software industries of countries like India and Ireland has increased the attractiveness of those countries to highly skilled indigenous workers. This has been particularly evident in Ireland, where rapid growth has encouraged an increasing number of highly skilled workers to remain in Ireland or return to Ireland from the United States. Kapur and McHale (2005a) report that emigration of male Irish graduates fell from about 25 percent in 1987 to under 15 percent in 1997, with similar trends for female graduates. Of the 644,444 Irish who had spent one year outside of Ireland in a 2002 census, 42 per- cent reported taking up residence in Ireland between 1996 and 2002, suggesting that a large fraction are recently returning Irish (Kapur and McHale, 2005a). 32 With the continuing growth of the software industries in India and Ireland, it is likely that these historically important sources of highly skilled software professionals will retain a growing fraction of their indigenous software workers. Moreover, as noted by Kapur and McHale (2005b), the international market for software professionals is increasingly competitive. Richer countries such as the United States, Canada, Australia, Germany, and the United Kingdom increasingly compete for talent from other countries. In many cases, this competition has manifested itself as a decline in the traditional barriers to short- and long-term migration (Kapur and McHale, 2005b). This competition is likely only to increase with the aging demographics of these countries as well as the increasing require- ments for a skilled workforce in software and in other industries. Federal Government Spending on Software R&D U.S. federal government investment in computer hardware and software R&D is thought to be one of the contributing success factors to both industries (Flamm, 1988; Langlois and Mowery, 1996). Early government R&D investment in software provided the computer facilities for universities to conduct early software research (Langlois and Mowery, 1996) and federal agencies such as 31  Administrative data from the U.S. Department of Homeland Security, Bureau of Citizenship and Immigration Services. 32  These data include migration of Irish citizens that have returned after studying in U.S. universi- ties, including those studying for computer science degrees.

94 INNOVATION IN GLOBAL INDUSTRIES National Aeronautics and Space Administration (NASA) and Defense Advanced Research Projects Agency (DARPA) have been long-standing supporters of com- puter-related research. Federal grants remain a major source of funding for doc- toral students in computer science: in 2003, 17.4 percent of full-time computer science graduate students reported that their primary source of funding was from the federal government.33 In the 1990s, though funding from the Department of Defense had largely flattened out, R&D spending grew rapidly throughout the decade through ex- panded funding from agencies such as the Department of Energy and NSF. However, over the period 2001-2003 (the most recent data available), government R&D spending in computer science remained largely flat. Moreover, the percent of total R&D spending on computer science (relative to other fields) declined over the period 2001-2003, from 4.5 to 4.0 percent.34 We discuss the implications of these spending patterns in the next section. CONCLUSIONS AND IMPLICATIONS Public Policy Implications The trends that we have described in this paper have several public policy implications. First, our results have provided evidence of a sizable export-driven software services sector in countries like India and Ireland, though there is less evidence of substantial inventive activity in software going on outside of the United States. These results suggest that entry- and mid-level programming jobs can be performed away from the point of final demand, though inventive activ- ity that requires proximity with lead users is most effectively done in the United States. However, these entry- and mid-level programming jobs have traditionally provided U.S. IT workers with the skills needed to perform more complicated development activities such as creation of new software programs (Levy and Murname, 2004). In other words, training by U.S. firms has traditionally be- stowed a beneficial externality upon entry-level workers by providing them with general human capital that workers appropriate later in their careers. This human capital is not easily provided by traditional publicly funded primary or secondary school education programs (Levy and Murname, 2004). As a result, a declining demand for entry-level programming jobs could hurt U.S. workers’ future abil- ity to perform more complex software development activity (e.g., new packaged software development). If this is true, then there are two ways that U.S. workers could obtain the general human capital needed. One would be for U.S. workers 33  National Science Foundation, Division of Science Resource Statistics, Survey of Graduate Stu- dents and Postdoctorates in Science and Engineering, WebCASPAR database (Science and Engineer- ing Indicators, 2006). 34  Science and Engineering Indicators 2006.

SOFTWARE 95 to internalize the externality by accepting jobs for lower salaries. Of course, in the short run, workers may prefer instead to accept jobs in other (relatively higher-paying) fields. Alternatively, the government could attempt to subsidize entry-level employment, for example, by raising the costs of H-1B visas or by direct labor market subsidies. However, if the cost of remote software develop- ment remains lower than that in the United States, then clearly implementation of this policy may be problematic. We have provided evidence of recent declines in computer science enroll- ments at the graduate and undergraduate levels. In our view, it is too soon to speculate whether these changes are evidence of a new trend or instead reflect temporary student reactions to business cycle fluctuations, particularly the IT downturn that began in the early part of this decade. Still, there is evidence that for some time, U.S. software developers have been using skilled labor from abroad as inputs into their innovation production function, presumably in part to supplement the pool of skilled labor available locally. As noted earlier, there is increasing competition from other industrialized countries for these skilled workers, and there is no sign that this competition will abate in the near future. Decreasing the costs of H-1B visas or lowering the costs of permanent migration is unlikely to be feasible in the short run because of the aforementioned concerns of labor substitution between foreign and indigenous workers. As a result, ensur- ing an adequate supply of local workers with sufficient basic or enabling skills (Levy and Murname, 2004) in mathematics, computer science, and related fields taught in the nation’s school and university system will be important to the long- term success of software producers in the United States. Another area of public policy concern is in government funding of computer science research. As noted earlier, federal funding of computer science has flat- tened out in recent years. A continuation of this trend could negatively impact innovative activity in software in the United States in two ways: by decreasing an important source of financial capital for basic research as well as potentially accentuating the negative downturn in enrollments in computer science graduate programs in the United States through a decline in graduate student funding. Summary and Conclusions There are currently two very different stories in the globalization of soft- ware development. On the one hand, the IT services industries in countries such as India, Ireland, and other countries continue to grow rapidly. The production of IT services is quite dispersed globally, and this dispersion will only increase over time. In contrast, both sales and inventive activity in packaged software are localized in the United States and undertaken primarily by U.S. firms. While differences in the levels of patent activity across countries should be interpreted with some caution because of divergence between the rate of patenting and inven- tive activity, examination of patent growth rates is less subject to these concerns

96 INNOVATION IN GLOBAL INDUSTRIES and they show there is no sign of these trends reversing in the short to medium term. Recent trends in computer science enrollments have attracted considerable attention in the popular press. We do find evidence of some declines in enroll- ments in U.S. computer science in recent years.35 However, of likely equal or greater importance in the short run may be the increasing incentives for skilled foreign workers to remain in their home countries or to depart from the United States immediately or some years after degree conferral. There is already some evidence that improving educational systems and employment opportunities in the underdog countries is causing some skilled software professionals to remain at home or to return. Nonetheless, there are powerful forces at work that are likely to keep the development of new software products and software innovation concentrated in the United States for some time to come. Despite recent trends, the United States continues to have the best postsecondary educational systems in the world for training computer scientists, and it continues to enjoy substantial albeit declin- ing inward migration that benefits the software (and other) industries. Beyond the education and human capital issues, U.S. software innovators continue to enjoy substantial advantages due to agglomeration economies arising from the preexisting concentration of the industry, as well as a generally favorable business environment. Perhaps the most significant advantage that U.S. software product innovators enjoy is proximity to lead users. U.S. firms have been among the most innovative users of IT in the world, and these users have benefited U.S. software producers in the past and will continue to do so for some time to come. ACKNOWLEDGMENTS We thank Jeffrey Macher and David Mowery for their comments and sug- gestions, and Nicholas Yoder and Kristina Steffenson McElheran for outstanding research assistance. REFERENCES Allison, J., and E. Tiller. (2003). Internet business method patents. Pp. 259-257 in Patents in the Knowledge-Based Economy, W. M. Cohen and S. A. Merrill, eds. Washington, D.C.: The National Academies Press. Allison, J., A. Rai, and B. N. Sampat. (2005). University Software Ownership: Trends, Determinants, Issues. Working Paper, Columbia University. Anchordoguy, M. (2000). Japan’s software industry: A failure of institutions? Research Policy 29:391-408. 35  In graduate programs these declines appear to be concentrated primarily among immigrants. Among undergraduate degree programs current data are not available to indicate whether these de- clines are from U.S. nationals or immigrants.

SOFTWARE 97 Armstrong, D.J., and P. Cole. (2002). Managing distances and differences in geographically distrib- uted work groups. Pp. 167-186 in Distributed Work, P. J. Hinds and S. Kiesler, eds. Cambridge: MIT Press. Arora, A. (2006). The Indian Software Industry and its Prospects. Working Paper, Heinz School of Public Policy & Management, Carnegie Mellon University. Arora, A., and S. Bagde. (2006). The Indian Software Industry: The Human Capital Story. Working Paper, Heinz School of Public Policy & Management, Carnegie Mellon University. Arora, A., and C. Forman. (2007). Proximity and Software Programming: IT Outsourcing and the Local Market. Proceedings of the 40th Annual Hawaii International Conference in System Sci- ences, Waikoloa, Hawaii. Available at http://www.hicss.hawaii.edu/diglib.htm. Arora, A., V.S. Arunachalam, V.S. Asundi, and R. Fernandes. (2001). The Indian software services industry. Research Policy 30(8):1267-1287. Arora, A., A. Gambardella, and S. Torrisi. (2004). In the footsteps of Silicon Valley? Indian and Irish software in the international division of labor. Pp. 78-120 in Building High-Tech Clusters: Silicon Valley and Beyond, T. Bresnahan and A. Gambardella, eds. Cambridge, UK: Cambridge University Press. Aspray, W., F. Mayadas, and M. Y. Vardi. (2006). Globalization and Offshoring of Software: A Report of the ACM Job Migration Task Force. Available at http://www.acm.org/globalizationreport/ pdf/fullfinal.pdf. Athreye, S. (2005a). The Indian software industry. Pp. 7-40 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press. Athreye, S. (2005b). The Indian Software industry and its evolving service capability. Industrial and Corporate Change 14(3):393-418. Baba, Y., S. Takai, and Y. Mizuta. (1996). The user-driven evolution of the Japanese software industry: The case of customized software for mainframes. Pp. 104-130 in The International Computer Software Industry, D. C. Mowery, ed. Oxford, UK: Oxford University Press. Bessen, J., and R. M. Hunt. (2004). An Empirical Look at Software Patents. Working Paper 03-17/R, Research on Innovation. Botelho, A. J. J., G. Stefanuto, and F. Veloso. (2005). The Brazilian software industry. Pp. 99-130 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press. Bresnahan, T., and S. Greenstein. (1996). Technical progress in computing and in the uses of comput- ers. Brookings Papers on Economic Activity, Microeconomics 1-78. Breznitz, D. (2005). The Israeli software industry. Pp. 72-99 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press. Brooks, F. (1995). The Mythical Man-Month: Essays on Software Engineering, 20th Anniversary Edition. Reading, MA: Addison-Wesley. Campbell-Kelly, M. (2003). From Airline Reservations to Sonic the Hedgehog: A History of the Software Industry. Cambridge: MIT Press. Copeland, D. G., and J. L. McKenney. (1988). Airline reservations systems: Lessons from history. MIS Quarterly 12(3):353-370. Cottrell, T. (1996). Standards and the arrested development of Japan’s microcomputer software indus- try. Pp. 131-164 in The International Computer Software Industry, D. C. Mowery, ed. Oxford, UK: Oxford University Press. European Information Technology Observatory. (2006). European Information Technology Observa- tory 2006. Berlin, Germany: European Information Technology Observatory (EITO)—European Economic Interest Grouping (EEIG).

98 INNOVATION IN GLOBAL INDUSTRIES European Patent Office (EPO). (2005). Computer-Implemented Inventions and Patents. Law and Practice at the European Patent Office. Flamm, K. (1988). Creating the Computer: Government, Industry, and High Technology. Washington, D.C.: Brookings Institution Press. Fransman, M. (1995). Japan’s Computer and Communications Industry: The Evolution of Industrial Giants and Global Competitiveness. New York: Oxford University Press. Furman, J. L., M. E. Porter, and S. Stern. (2002). The determinants of national innovative capacity. Research Policy 31(6):899-933. GAO (General Accounting Office). (2005). U.S. and India Data on Offshoring Show Significant Dif- ferences. GAO Report GAO-06-116. Gereffi, G., and V. Wadhwa. (2005). Framing the Engineering Outsourcing Debate: Placing the United States on a Level Playing Field with China and India. Report, Master of Engineering Program, Duke University. Gompers, P., and J. Lerner. (2006). The Venture Capital Cycle: Second Edition. Cambridge: MIT Press. Gormely, J., W. Blustein, J. Gatloff, and H. Chun. (1998). The runaway costs of packaged applica- tions. The Forrester Report 3(5). Graham, S. J. H., and D. C. Mowery. (2003). Intellectual property protection in the U.S. software industry. Pp. 219-258 in Patents in the Knowledge-Based Economy, W. M. Cohen and S. A. Merrill, eds. Washington, D.C.: The National Academies Press. Graham, S. J. H., and D. C. Mowery. (2005). Software patents: Good news or bad news. Pp. 45-80 in Intellectual Property Rights in Frontier Industries: Software and Biotechnology, R. W. Hahn, ed. Washington, D.C.: AEI-Brookings Joint Center for Regulatory Studies. Griliches, Z. (1990). Patent statistics as economic indicators. Journal of Economic Literature 28(4): 1661-1707. Hall, B. H., and M. MacGarvie. (2006). The Private Value of Software Patents. NBER Working Paper 12195. Hall, B. H. and R. H. Ziedonis. (2001). The patent paradox revisited: an empirical study of patenting in the U.S. semiconductor industry, 1979-1995. RAND Journal of Economics 32(1): 101-128. Herbsleb, J. D., D. J. Paulish, and M. Bass. (2005). Global software development at Siemens: Expe- rience from nine projects. Pp. 524-533 in International Conference on Software Engineering. St. Louis, Mo. Ilavarasan, P. V. (2006). R&D in Indian software industry. Pp. 134-143 in Managing Industrial Research Effectively, R. Varma, ed. ICFAI University Press. Jaffe, A., and M. Trajtenberg. (2002). Patents, Citations, and Innovations: A Window on the Knowl- edge Economy. Cambridge: MIT Press. Kapur, D., and J. McHale. (2005a). Sojourns and software: Internationally mobile human capital and high-tech industry development in India, Ireland, and Israel. Pp. 236-274 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press. Kapur, D., and J. McHale. (2005b). Give Us Your Best and Brightest: The Global Hunt for Talent and Its Impact on the Developing World. Washington, D.C.: Center for Global Development. Langlois, R. N., and D. C. Mowery. (1996). The federal government role in the development of the U.S. software industry. Pp. 53-85 in The International Computer Software Industry, D. C. Mowery, ed. Oxford, UK: Oxford University Press. Layne-Farrar, A. (2005). Defining Software Patents: A Research Field Guide. Working Paper 05-14, AEI-Brookings Joint Center for Regulatory Studies. Levy, F., and R. J. Murname. 2004. The New Division of Labor: How Computers Are Creating the Next Job Market. New York: Russell Sage Foundation.

SOFTWARE 99 Malerba, F., and S. Torrisi. (1996). The dynamics of market structure and innovation in the Western European software industry. Pp. 165-196 in The International Computer Software Industry, D. C. Mowery, ed. Oxford, UK: Oxford University Press. March, J. G., and H. A. Simon. (1958). Organizations. New York: Wiley. Merges, R. P. (2006). Patents, Entry, and Growth in the Software Industry. Working Paper, University of California, Berkeley. Messerschmitt, D. G., and C. Szyperski. (2002). Software Ecosystem: Understanding an Indispensible Technology and Industry. Cambridge: MIT Press. Mokyr, J. (2002). The Lever of Riches. Oxford, UK: Oxford University Press. Mowery, D. C., and N. Rosenberg. (1998). Paths of Innovation. Cambridge, UK: Cambridge Uni- versity Press. Nelson, R., and R. Rosenberg. (1994). American universities and technical advance in industry. Re- search Policy 23(3):323-348. NSF (National Science Foundation). (2006). Science and Engineering Indicators 2006. Washington, D.C. OECD (Organisation for Economic Co-operation and Development). (2002). OECD Information Technology Outlook. Paris: OECD. OECD. (2004). OECD Information Technology Outlook. Paris: OECD. OECD. (2006). OECD Information Technology Outlook. Paris: OECD. Olson, G. M., and J. S. Olson. (2000). Distance matters. Human-Computer Interaction 15(2-3): 139-178. O’Riain, S. (1999). Remaking the Developmental State: The Irish Software Industry in the Global Economy. Doctoral Dissertation, Department of Sociology, University of California, Berkeley. Parker, R. W. (1965). The SABRE System. Datamation (September):49-52. Parthasarathy, B., and Y. Aoyama. (2006). From software services to R&D services: Local en- trepreneurship in the software industry in Bangalore, India. Environment and Planning A 38(7):1269-1285. Porter, M. E. (1990). The Competitive Advantage of Nations. New York: Free Press. Ricadela, A. (2005). Q&A: Bill Gates on supercomputing, software in science, and more. Information Week. November 18. Romer, P. (1990). Endogenous technological change. Journal of Political Economy 98:S71-S102. Rosenberg, N. (1963). Technological change in the machine tool industry, 1840-1910. Journal of Economic History 23:414-443. Rosenberg, N. (1983). Inside the Black Box: Technology and Economics. Cambridge, UK: Cambridge University Press. Sands, A. (2005). The Irish software industry. Pp. 41-71 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press. Scaffidi, C., M. Shaw, and B. Myers. (2005). Estimating the numbers of end users and end user pro- grammers. Pp. 207-214 in Proceedings of the 2005 IEEE Symposium on Visual Languages and Human-Centric Computing. Los Alamitos, Calif.: IEEE Computer Society. Schumpeter, J. A. (1934). The Theory of Capitalist Development. Cambridge, MA: Harvard Univer- sity Press. Steinmueller, E. (2004). The European software sectoral system of innovation. Pp. 193-242 in Sec- toral Systems of Innovation: Concepts, Issues, and Analyses of Six Major Sectors in Europe, F. Malerba, ed. Cambridge, UK: Cambridge University Press. Thoma, G., and S. Torrisi. (2006). The Evolution of the Software Industry in Europe. Working Paper, CESPRI, Bocconi University. Tschang, T., and L. Xue. (2005). The Chinese software industry. Pp. 171-206 in From Underdogs to Tigers: The Rise and Growth of the Software Industry in Brazil, China, India, Ireland, and Israel, A. Arora and A. Gambardella, eds. Oxford, UK: Oxford University Press.

Next: 3 Semiconductors »
Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies) Get This Book
×
 Innovation in Global Industries: U.S. Firms Competing in a New World (Collected Studies)
Buy Paperback | $73.00 Buy Ebook | $59.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The debate over offshoring of production, transfer of technological capabilities, and potential loss of U.S. competitiveness is a long-running one. Prevailing thinking is that "the world is flat"—that is, innovative capacity is spreading uniformly; as new centers of manufacturing emerge, research and development and new product development follow.

Innovation in Global Industries challenges this thinking. The book, a collection of individually authored studies, examines in detail structural changes in the innovation process in 10 service as well as manufacturing industries: personal computers; semiconductors; flat-panel displays; software; lighting; biotechnology; pharmaceuticals; financial services; logistics; and venture capital. There is no doubt that overall there has been an acceleration in global sourcing of innovation and an emergence of new locations of research capacity and advanced technical skills, but the patterns are highly variable. Many industries and some firms in nearly all industries retain leading-edge capacity in the United States. However, the book concludes that is no reason for complacency about the future outlook. Innovation deserves more emphasis in firm performance measures and more sustained support in public policy.

Innovation in Global Industries will be of special interest to business people and government policy makers as well as professors, students, and other researchers of economics, management, international affairs, and political science.

READ FREE ONLINE

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!