TOWARDS A NEW ECONOMY: MINISTERIAL REPORT ON …
G:\TRC\AW\EAS\ICCP SOFTWARE.DOC
Better Understanding the Role of Software in the Information Economy:
A Preliminary Overview of Research, Policy and Measurement Challenges
Andrew W. Wyckoff and Nadim Ahmad
Organisation for Economic Co-operation and Development
Paris, France[1]
Abstract
As our power to measure and analyse the impact of information and communication technologies (ICT) on economic performance has increased, it is clear that our current tools are blunt and crude and do not allow us to differentiate between driving determinants, necessary compliments and extraneous factors. A necessary first step in this work is to move beyond broad aggregates such as IT or ICT and begin to focus on components such as computer equipment, communications equipment, computer services, communication services and software. Each of these elements has evolved from a distinctly different technological, economic and policy environment, necessitating a distinct treatment. This entails a need for significant improvements in their measurement and analysis. This paper examines this need using software as a case study both from the perspective of its development as well as its use.
Presented to “Transforming Enterprise and Beyond:
Connecting Research and Policy in the Digital Economy”
Post Conference Workshop at the National Science Foundation
29 January 2003
Washington, DC
Preliminary working draft. Please do not quote or cite without permission of the authors. Comments are welcomed and should be sent to Andrew.wyckoff@ or Nadim.Ahmad@.
The Changing Nature of Growth in the 1990s
1. 1. Now that the hype associated with the new economy has dissipated as the dotcom and telecom bubbles have burst and a series of accounting scandals have cast doubt on the true performance of some of the fast growing firms, a more critical assessment of the determinants of economic growth during the 1990s can occur with the aid of hindsight. While the fundamentals of a stable macro policy, competitive product markets, flexible labour markets and a well functioning financial system will be reasserted, it is clear that the role of science and technology policy, especially the technology of the 1990s -- information and communication technologies (ICT) -- was a key differentiating determinant of growth performance across countries. While the United States has been the laboratory for much of this analysis, in fact the importance of ICT to growth, both in terms of production and use, has been confirmed for several other countries as well (OECD, 2002).
2. Among the findings from this work are (see Table 1):
• Despite different timing of countries’ expansions and recessions in the 1990s, all the eight OECD countries analysed have witnessed a rapid increase, ranging from 10 to 25 percent, in constant price investment in information and communication equipment and software. In the US, Australia and Finland, ICT investment accounted for over 50 per cent of constant price growth of non-residential investment in the most recent years (1995-99);
• Over the past twenty years (1980-99) the contribution of ICT equipment and software to output growth of the business sector has been between 0.2 and 0.5 of a percentage point a year depending on the country. Over the past 4 years (95-99) the contribution of ICT and software has raised risen to annual values that range from 0.3 to 0.9 per cent. The United States is not alone in experiencing the “growth” effects of ICT, . ICT equipment and software have also played are playing a major role in driving output growth also in the case of Australia and Finland. In the case of Canada software investment series are not available but the contribution of ICT equipment is fairly strong and growing. However, Tthe impact of ICT in the other countries in the sample, instead, has shown little change in seems not to have increased in the latest years;
• Software has been the most dynamic component of ICT investment and, in the 1995-99 period, accounted for 25 to 40 per cent of the ICT contribution to overall investment growth across the eight countries; [Figure ?]
• Software capital accumulation accounted for a third of the overall contribution of ICT capital to output growth between 1995 and 1999. What is remarkable is that this result holds across all OECD countries in the sample, with the exception of Japan and the UK[2]. The United States is the most striking in that respect since the average percentage contribution of software to output growth in 1995-99 is four times up from its 1980-85 value.
3. The estimates obtained in this study are in the range of those obtained with similar methods and official statistics for the United States, France and the United Kingdom (see tTable 2). A point that fails to get much recognition is the important role played by software.
4. Although comparable international data is not yet available for updating this work and taking into account the downturn associated with the bursting of the “dotcom” and telecommunications investment bubbles, these results have been reaffirmed for the US. While the downturn in 2001 trimmed the increase in the average annual labour productivity growth achieved in the latter half of the 1990s as compared to 1973 to 1995 from about three-quarters of a percentage point to a little more than two-thirds, the contribution from IT to these productivity gains has not changed (Jorgenson, Ho and Stiroh, forthcoming).
5. Updating previous work, the Federal Reserve Board (Oliner and Sichel, 2002) shows a similar downward tick in the growth of labour productivity as new, more up-to-date data is included which contains the US downturn, but confirms the importance of ICT (Table 32). This study differentiates between computer hardware, software and communications equipment, revealing that while software made a contribution to labour productivity growth that was the same as that made by communications equipment between 1974 and 1990 (0.13 percentage points), and only about 40 percent of what computer equipment made during this period, its contribution from 1996 to 2001 had more than tripled, where it was accountable for more than four-tenths of a percentage point to labour productivity growth. This contribution was more than double twhat made by communications equipment made during this period and nearly two-thirds (63 percent) of that madeof by what computer hardware contributed. By 2001, US investment in software was more than double that or computers or communications equipment (US Department of Commerce, 2003).
6. It appears that many of the OECD economies have evolved from a “gold rush” period as regards ICT, fuelled by fears of a Y2K catastrophe and the excitement associated with the invention of the WWW, to a reflective period where organisations are learning how best to use this technology. In this sense, the economic impact of ICT may be just beginning and may be better analysed and understood in a more subdued economic period. It also gives the researchers and statistical agencies some time to consolidate the work done to date and consider thoughtfully the next steps that need to be taken to improve our understanding of this phenomenon.
Unanswered Questions
7. Inherent in any exercise that attempts to understand a new phenomenon, is that it generates more questions than it answers and is unable to analyse some factors simply because we lack the data and the proper conceptual framework. One of these unanswered questions is: “Why the sudden increase in productivity starting in the second-half of the 1990s?” The OECD recommendations suggest that “getting the economic fundamentals right,” “public investment in innovation”, “investing in up-skilling” or “injecting more competition into markets” are useful for creating economic frameworks conducive to innovation characterised by ICT in the 1990s, but these policies have a long gestation period. In this sense, they do not provide a completely satisfying explanation as to why a country like the US suddenly saw its productivity rate double in the late-1990s compared to a trend that had prevailed over the previous two-decades. Analytically, given the jump, one might expect to see would look for an “event” such as an oil shock, a dramatic change in exchange rates or a technological breakthrough that is quickly and widely diffused.
8. There is no shortage of plausible explanations. Some have suggested that it was the budget agreement that President Clinton made with Congress in 1993 that provided the foundation for an economic environment conducive to investment (Baily and Lawrence, 2001), others suggest that it was the cumulative effect of increased labour participation (Bosworth and Triplett, 2000). Many suspect that information and communication technologies played a significant role, but without much explanation as to why ICTs would have a more pronounced impact in the late-1990s than in the 1970s or the 1980s when their newness would have been expected to have a more visible impact. Some suggest that it was due to an acceleration of Moore’s Law at this time (Jorgenson, 2001). While others point to the analogy put forward by Paul David (David, 1990) that as in the case of electric dynamos it takes time for organisations to adapt their structure to fully exploit the potential of a new technology (Brynjolfsson and Hitt, 2002). Others suggest that the accumulation of falling costs led to capital deepening so that whereby 1995 the stock of ICT was of sufficient size to have an observable impact (capital deepening) (Oliner and Sichel, 2002). And some suggest that it was a combination of one-time events such as the Y2K bug, the invention of the Internet and a very favourable financial environment for investment as well as organisational innovations by “big box” US retailers (Gordon, 2003). Lastly, in a reversal from the circumstance of the early 1980s, the growth of the 1990s was due to a fortuitous confluence of good luck (favourable supply side shocks) (Bosworth and Triplett, 2000). In practiceactuality, it is likely that all of these explanations played a role. In this sense, it is insufficientwrong to look at the impact of ICT in isolation but rather how it interacts with a broader constellation of conditions and factors.
9. Upon closer examination of many of these studies, an important “red thread” running through many of the findings is the explicit and implicit role of software as an enabling technology. While the Internet is popularly thought of as servers, optic fibre cables and routers, it can and did work without any of this hardware. In fact, the Internet is probably more accurately thought of as software thanks to three separate software innovations: the transport control protocol / Internet protocoal (TCP / IP) that allows messages to be broken down into packets; hypertext mark up language (HTML) that allowed documents to be linked, creating the World-wide web (WWW) and the web browser that provided an easy-to-use form of access to the WWW. SMTP (simple mail transfer protocol), the standard for the exchange of e-mail can be credited with enabling the “killer application” that continues to fuel the demand for consumer Internet access[3]. This cluster of developments and their adherence to non-proprietary, open standards coalesced in 1994/95 and created a network that at a very low cost could tie together the existing computing stock through use of easy-to-use graphical software that was platform independent, non-proprietary and linked the existing communication systems (satellite, cable, telephone, etc.). This development vastly increased the functionality of the existing ICT capital, lowered the switching costs of moving from one IT technology to another and enabled new business practices that led to growth and increases in productivity[4].
10. All of these developments were undoubtedly contributing factors as to why ICT started to have an economic impact during the middle-of-the-1990s, although it is difficult to establish causality and it is clear that no one factor can be isolated as the key factor. This underscores the need to improve our analytical tools, especially the measurement of these different factors so that the role of ICT and the constellation of factors that surround it are better understood. A necessary first step in this work is to move beyond broad aggregates such as IT or ICT and begin to focus on components such as computer equipment, communications equipment, computer services, communication services and software. In an effort to sketch out the nature of this work, one element, software, is used as a case study. The policy issues posed by software are briefly explored from the perspective of its use as it becomes an integral part of the economic infrastructure as well as its development since this sector may be offer “strategic” advantages in maintaining technological competitiveness. The measurement issues are then outlined with a focus on the treatment of software in the national accounts given the importance of this source for economic analysis.
The Use of Software
11. As demonstrated by the Microsoft antitrust case, the Y2K scare, the widespread copying of music, the large-scale inconvenience associated with SPAM or the various viruses that take advantage of holes in popular applications, the use of software and its importance to the economy is already recognised. But it appears that these events are treated in an ad hoc, piecemeal basis and an overall appreciation for the role software plays as an enabling economic infrastructure, akin to ports and highways for transport, is lacking. This is probably because its role is largely invisible – the rocket blows up, the plane crashes or telephone service is disrupted (NAS, 2000; Huckle, 2002) -- but the role of software is only recognised as the critical factor later. Because of its increasing complexity both separately, and as applications begin to interact with each other, the role of software is difficult to comprehend, especially to policy makers. Sadly, heightened concerns about terrorism may be changing this view.
12. Policy issues associated with the use of software are as ubiquitous as is the technology, running the gamut from how it has changed communication patterns, and thus the sense of what constitutes a community, to how it may affect intellectual property rights. Even when the focus is more narrowly concentrated on the various policies designed to bolster the impact of ICT on economic growth and productivity, the scope can be broad. Two areas are briefly surveyed: 1) the importance of the use of software by service sectors and 2) the fact that being “ICT literate” in most cases means being able to use software.
Services and Software
13. Even though services represent the bulk of most OECD economies, their role as potential sources of innovation and productivity receives little attention both on research and policy agendas. But the inherent nature of services, where production and delivery happen at the same time and where products tend to be tailored to meet individual demand, necessitates making the collection and analysis of information a core element of their production process. ICT allows the automation of services, e.g. where an ATM can effectively replace a bank teller. No wonder therefore that about three-quarters of all ICT investments are made by sService industriess (Bonds and Aylor, 1998). Recent research finds that ssService industries s in general improved their productivity performance in the second half of the 1990s and some of the most important gains by some sectors (e.g. banks, wholesale trade, transportation services) were attributable to ICT (Bosworth and Triplett, 2002). But equally striking is the finding that some services still have poor productivity performance (e.g. hotels, health) even after significant investments in ICT.
14. Understanding why some service firms appear to achieve positive results from the adoption of ICT while others do not is an important research question. The role of software is seen as a key element since it is frequently cited as a barrier for fully exploiting theICT equipment (Bresnahan and Greenstein, 1997). At least three key characteristics of use are in need of further study: 1) the co-innovation process where by users work directly with producers to develop an application that works (Bresnahan, 1999; vonHipple, 2002) ; 2) the use of computer service providers such as system integrators, consultants, logistics or specialised software producers who act as intermediaries between producers and users and customize the system to the user, a process that usually entails software development and; 3) the in-house development of software by services, not only as a means to improve their production process but also as a tool for innovation. The use of “customized software” and “in-house development” are not trivial and currently constitute about one-third each of the total annual investment in software by US firms. In total, US investment in software is about the same as all investments made in trucks, buses, ships, boats and railroad equipment (US Department of Commerce, 2003).
15. The improved understanding of these vectors of technology adaptation and adoption are important for reorienting technology diffusion programs and policies designed to increase investment (e.g. investment tax credit) which have a tendency to focus on manufacturing even though the potential contribution to economy-wide productivity gains from services is much larger.
16. Software is both a traded product and is an enabler of trade, particularly in electronically delivered products, many of which are services like the distribution of music. In both cases, the electronic delivery raises issues as to whether the product should be treated as a good or a service under international rules governing trade and taxation. It also exposes sectors like retail to which have not yet been exposed to significant international trade but have been “traded” through foreign direct investment or have operated on a global level only for large corporate clients. This change may come as a shock to sectors that have been sheltered by logistical or regulatory barriers. In addition, it will generate pressures to reduce differences in regulatory standards such as accreditation, licensing, restrictions on activity for newly tradable products. This direct tradability of services could increase the frictions that exist between countries in areas that impinge on “culture” such as language, art and entertainment, sensitivities about pornography and gambling, and attitudes regarding health and education.
Human Capital needed to facilitate use
17. Software is at the root of the evolution of OECD economies towards “knowledge-based economies” that require the use of many abstract, analytical skills as opposed to manual dexterity. In this sense a large element of becoming “ICT literatecy” concerns is about the use of software as opposed to hardware. These skill requirements place new demands on schools and vocational training facilities and, by association, education policy. A system of education that familiarises young students with the use of software could reduce skills acquisition costs and decrease differences in participation rates across social stratathe various segments of a society’s population. Similarly, training in the use of software needs to be an integral part of any policy of life-long learning for citizens who have left the school system and face the continuing need to master new software or an upgrade of some existing application. Becoming computer-literate can be a significant additional cost, one that is likely to vary as a function of age and educational background.
The Development of Software
18. As the use of software becomes recognised as a key element in the economic infrastructure and a tool for improving the productivity of services, and thus the economy, the development of software and its integrity should become widely recognised as a “strategic” industry whose existence improves the competitiveness of other industries. Software development also holds the promise for addressing many of the policy issues such as security, privacy protection and protection for consumers engaging in e-commerce that plague the information society.
19. 19. The exact factors that lead to the successful development of software are not well understood and deserve additional research given the importance of this technology. Two factors that appear to be crucial are a large, competitive, market and the existence of sufficient human capital (software engineersprogrammers).
Competition Policy
20. 20. The importance of competition policy to the development of software has been recently highlighted by the Microsoft anti-trust case, but has been an important factor traceable to the birth of the industry and the pressure put on IBM to unbundled its hardware from its software. The experience of Japan and its numerous government led efforts to develop a software industry underscore the different nature of software development from other products (Anchordoguy, 2000).
21. [Breshanan /Tassey/ Shaperio.. to be added]
21. [needs more: Tassey, Shaperio & Breshnahan]
Human Capital needed to facilitate development
22. 22. As many have pointed out, a “knowledge paradox” exists where as more and more knowledge becomes codified, and thus accessible widely, the remaining non-codified, tacit knowledge, becomes a more crucial, differentiating factor. Developing, attracting and managing people who have this tacit knowledge is especially important in the area of software development; where one software engineer may be 10 times as productive as another. Software development policies oriented towards creating the next generation of software innovations must be oriented towards building or attracting this cadre of “star power” human capital either through the indigenous development of software engineers or their attraction from abroad.
23. 23. Policies to promote university quality in the area of computer science or software engineering are beyond the scope of this paper, but it is clear that federal funding (e.g. DARPA) has been an important factor in the creation of “centres of excellence” (e.g. MIT, UC Berkeley, Stanford and Carnegie Mellon) in this field and continue to play an important role: Federal support has constituted about 70% of total university research funding in computer science and engineering since 1976 (NAS, 1999). Research is needed to evaluate how this link between this funding and software development is evolving as the field matures.
24. 24. A wide number of policies ranging from preferential immigration polices to educational support exist to attract foreign talent in this area. In 1999, 46 percent of all computer science doctoral degrees were earned by foreign students (NSF, 2002). An important reason why the US attracts foreign students is the allure of the “centers of excellence” identified above and the chance to work at the scientific frontier with faculty that lead in field (OECD, 2001). An enabling factor is the receipt of financial support. More than 75% of the 1996 foreign doctoral recipients received financial support from the Universities, usually in the form of research assistantships which are largely Federally funded (NSF, 1998). Between 1985 and 1996, the number of foreign doctoral students who were awarded research assistant positions more than tripled (NSF, 1998).
25. 25. Education is an important channel for attracting the highly-skilled who then stay and find employment. The share of foreign students who stay in the US after receiving their degree differs by country and field, but of those foreign student who earned a computer science doctoral degree in 1994-95, 63 percent were still in the US in 1999 (NSF, 2002).
26. 26. It is clear that US software development has benefited from foreign immigration. A study using US Census and Dun and Bradstreet data show that nearly a third of Silicon Valley’s 1990 workforce was made up of immigrants, two-thirds of whom were from Asia, primarily either Chinesea or Indian descent (Saxenian, 1999). Between 1995 and 1998, Chinese and Indian engineers started 29 percent of Silicon Valley technology companies, up from 13 percent in 1980 to 1984 (IBID). Competition for this cadre of talent is likely to increase as other countries like the Japanese, the UK and the Germanys change their immigration laws to make it easier for the highly-skilled to immigrate, and as various countries and regions like the EU set goals for improving their innovation performance; which will necessitate a large influx of foreign scientists and engineers if these goals are to be met.
27. 27. Another channel for gaining access to this source of skilled labour has been to either establish development facilities abroad. Microsoft has software research facilities in Canada, outside the University of Waterloo;, the UK, in Cambridge; and in India, near the Indian Institute for Technology in Bangalore. Many other firms like Oracle, Nokia, HP, and IBM have also invested in facilities abroad. Linked to this has been the increasing tendency of firms to outsource to firms abroad, especially India (Hardy, 2002). While this initially began as outsourcing of routine, back office software, as these firms have matured, they have begun to take on more sophisticated work and the development of whole applications (OECD, 2000). This has met with concern from local developers that this practice of outsourcing undercuts their status and wages and fails to address the need for additional investment in local human capital. More recently, it has also raised security concerns about the possibility of thisof this software being susceptible to back doors, viruses and intentional errors.
Integrity
28. 28. As software becomes more integral to the economic infrastructure of advanced economies, its integrity and reliability becomes a policy issue. The US National Institute for Standards and Technology (testing) has estimated that the annual US costs of faulty software are in the range of 20 to nearly 60 billion dollars or 0.6% of GDP (NIST, 2002). This cost is a reflection of the growing complexity of various software systems (and their interaction) as well as a decreasing average market life, making it more difficult, and providing less of an incentive, for developers to adequately test software (NIST, 2002). Even though some estimate that typically half of the labour spent on developing a software program is devoted to testing (Beizer, 1990 in NIST, 2002), NIST has found that the existing methods and tools for testing are inadequate and let through too many bugs. The development of improved techniques that improve the efficacy, while reducing the cost of testing, may be a worthy policy objective and some firms are already launching initiatives in this area[5]. More critical observers believe that the culture of the industry needs to be changed through product liability lawsuits (Mann, 2002).
S&T Policy
29. 29. As is the case with many of the ICT innovations now having an effect on the economy, many of the fundamental break throughs in software -- windows, graphic user interfaces, relational databases, TCP/IP, HTML, and the browser -- were all developed in government laboratories with government money (NAS, 1999). This said, the notion that software development should be an explicit aim of S&T policy is not universally accepted and is frustrated by the notion that software is not a field in its own right but rather a tool for making advancements in traditional fields like physics or chemistry. This is aggravated by institutional rigidities in some universities where computer science is subservient to these traditional fiefdoms that fear competition from new fields that may eclipse their own. Continued government funding can help to change these institutional rigidities.
30. 30. Another issue is how to allocate funds for software R&D. While competitive, peer-reviewed funding programmes such as those conducted by NSF are models for academic research, the approach taken by the US Department of Defence’s Defence Advanced Research Projects Agency (DARPA), where funding was targeted towards certain topics, people and institutions and built select “centres of excellence”, must be credited with helping to create a strong software sector in the United States. The lesson is that while governments should not intervene excessively in this area, neither should they simply be spectators and should not fall prey to the mantra of “industry driven / market led” that many espouse in the ICT industry, ignoringant of the origin of many of the fundamental innovations they rely on.
31. 31. The growing importance of software as a tool in the innovation process needs to be better understood as well. As indicated, software development by services is a key source of innovation by itself, representing between 10 and 25 percent of R&D performed by services for countries that measure R&D for these sectors (Young, 1996; OECD, 2002c). To a large degree, service innovations such as “one-click” buying, instant messaging or financial derivatives are software. This form of innovation does not neatly conform to our innovation promotion policies (e.g. R&D tax incentives) or to our institutions that support the innovation process like patent offices.
32. 32. ICT in general and software in particular has been an important tool leading to innovations in a number of fields including biotechnology, nanotechnology, pharmacological research, aerospace, astronomy etc. Given that the essence of scientific research is to experiment, observe, analyse and report; information capture, manipulation and dissemination is essential and software systems that aid this endeavour are an important tool, leading some to suggest that “all science is becoming computer science.”, (Johnson, 2001). Beyond being a tool that improves the ability of researchers in a particular field, software has been an important factor in linking different fields. This has facilitated the exchange of ideas and borrowing of techniques from other fields, leading to advancements that would have been difficult to achieve if researchers had simply tried to forge ahead from their own knowledge base. These two characteristics – information processing and linking – have caused some to speculate that ICT may lead to a permanent improvement in the rate of innovation (Breshnahan, 2001; Jones, 19??). Further research is needed to explore this hypothesis given its significant implications for productivity and economic growth.
Statistical Challenges
33. 33. The important role software has played as a key part of investment’s contribution to recent growth as well as an element of change in many different policy issues illustrates the need for an explicit, comprehensive recognition ofas software as an integral part of the economic infrastructure of advanced economies. This requires a similar effort to make software a more common part of economic analysis; which in turn requires a comprehensive effort to identify software across a variety of economic and social statistics.
34. 34. There are not many statistical areas where accounting for software does not pose significant challenges. In the area of S&T indicators, new innovative software development is considered to be R&D, but is difficult to capture because frequently it is not considered to be R&D by the reporting firm. In the area of business statistics, the use of software, especially by service firms who use it to tailor a product to the consumer (e.g. travel preferencesnew service) or add convenience to a particular product (e.g. on-line banking) complicates the measurement of the output of that firm, especially if, over time, if the quality of the product being offered has been improved (Griliches, 1994). The ability of software to deliver intangible products like music across borders makes recording exports and imports difficult, complicating the calculation of balance of payments statistics.
35. 35. To gain a better appreciation of these challenges, one category – the treatment of software by the national accounts – will be explored because it is area where work is most advanced and is the set of data that is most traditionally used for analysis of growth and productivity, both by individual countries and for cross-country comparisons (Tables 1 and 2). This case study illustrates some of the measurement challenges that will have to be addressed as the measurement of the information economy moves out of the realm of case studies and ad hoc surveys and into mainstream economic statistical infrastructure that is used for day-to-day economic policy making.
Software in the System of National Accounts[6]
36. 36. With the latest revision of the system of national accounts (SNA93), it was recommended that software be capitalised and treated as an investment. This change reflected the ‘asset’ and ‘investment’ characteristics of software, its increasingly important contribution to economic growth and productivity, and brought the treatment of software purchased separately into line with software purchased as a bundle with hardware, which has always been capitalised.
37. 37. On average, the change increased GDP by over 1% in OECD economies but the variability is significant (see Figure 1) and difficult to explain. For example, according to official statistics, Denmark has three times the software investment levels of the UK (as a per cent of GDP) but has a software producing industry only ⅔ the size (as a per cent of GDP). The difficulty to explain these differences in an economically meaningful way has led many observers to question the methods used to make this measurement.
38. 38. In practice, two distinct methods exist for estimating software. The first is based on a conventional business survey of investment by asset type (known as a ‘Demand’ approach); . the traditional route for investment measurement. However the experience in most countries is that these surveys significantly underestimate investment, since businesses tend to adopt very prudent valuations. For example evidence suggests that many businesses do not capitalise own-account production of software at all. The widespread perception of inherent weakness of these surveys in measuring software investment has led many national statistical offices to estimate software investment using Supply based methods. These generally assume that investment levels can be determined by examining the total supply of software (broken down into sufficient product detail) entering an economy. So, for example, within the generic product description computer services, 100% of customised software services and 0% of hardware consultancy services might be capitalised. The second is to measure the total supply of computer services into an economy and estimate how much expenditure is on software that has asset characteristics (the ‘Supply’ approach). In practice Survey based methods are believed to deliver very low estimates of software investment, as businesses tend to adopt very prudent, (often zero), valuations. This widespread perception of the weakness of these surveys has led most countries to estimate software using a Supply approach, but underscores an important area in need of development.
39. [merge with para above ] Countries estimate investment in software using either a ‘Supply’ approach or a ‘Demand’ approach. The Demand approach is the traditional route for investment. Namely, that estimates are provided via businesses in investment surveys. However the experience in most countries is that these surveys significantly underestimate investment, mainly, but not exclusively, because businesses do not capitalise own-account production. As such most countries have resorted to using supply based methods. These generally assume that investment levels can be determined by examining the total supply of software (broken down into sufficient product detail) entering an economy. So, for example, within the generic product description computer services, 100% of customised software services and 0% of hardware consultancy services might be capitalised. Partly because of the differences described under ‘nomenclature’ above, these ratios differ across countries and contribute to the differences presented in Figure 2.
40.
39. HoweverNevertheless, tThhe sSupply approach has a number of shortcomings such as delineating between those expenditures on software that are capital expenditures, lasting more than one year, and those that are intermediate consumption (e.g. purchases of software that is bundled with, or embedded within, in anothersome final product, e.g. like a mobile phone). A priori, one would expect the ratio of final investments in software to intermediate expenditures on software to final investments in software towould be fairly similar across economies. But Figure 2 illustrates the difficulty countries have in making this distinction in a harmonised and consistent manner. It shows that where the ratio of capitalised software to total business and government expenditure on computer services varies by over a factor of over 10 across the 11 countries surveyed, median at 0.38; reflecting differences across statistical offices in the procedures, and conceptual framework, used to estimate software[?] .[?].
The difficulty of measuring intangibles
40. 41. Despite its intangible nature, the rationale for capitalising software is firmly based. Like tangible assets, software delivers a stream of capital services, and. ndeed the importance of these capital services on output and productivity has been well documented. However, transactions and production of tangible products can be easily measured, since they have clearly identifiable physical characteristics, conventional production processes, transparent service lives, and conventional markets but intangible products, in particular software, do not share these characteristics, making measurement problematic. In addition to the problems inherent in measuring intangibles, software has a number of other characteristics, which increases the difficulty of measurement:
• The markets for software are unconventional and complicated. For example, software is usually sold with a number of conditions attached, meaning that it is not immediately apparent where ownership resides, for example; the software producing company, the software purchaser, or the software publisher.
• Once a software ‘original’ has been produced, copies can be reproduced at minimal cost, and conditions are often included which restrict the subsequent use or resale of the copy.
• The service life of software is difficult to measure. Software does not age like tangible products, which tend to become less efficient as a result of ‘wear and tear’. In addition reproduced software is often purchased using license arrangements, begging the question of what happens to the software when the license is terminated or comes to an end. For example, a rented car continues to exist after the rental period ends and the car is returned to the car rental company but what happens to software?
• Finally it is not clear what the physical characteristics of reproduced software are? Do software copies actually exist or are they merely licenses/contracts that permit the use of, or access to, a software ‘original’.
41. 42. These characteristics complicate software-capitalisation but they do not invalidate it. The key problem for National Accountants however is how to interpret these characteristics, and for international comparability, how to interpret them in a consistent way.
The SNA definition
42. 43. Paragraphs 10.92 and 10.93 of the 1993 edition of SNA describe software as:
Computer software that an enterprise expects to use in production for more than one year is treated as an intangible fixed asset. Such software may be purchased on the market or produced for own use. Acquisitions of such software are therefore treated as gross fixed capital formation. Software purchased on the market is valued at purchasers’ prices, while software developed in-house is valued at its estimated basic price, or at its costs of production if it is not possible to estimate the basic price. Gross fixed capital formation in software also includes the purchase or development of large databases that the enterprise expects to use in production over a period of time of more than one year. These databases are valued in the same way as software.
43. A common interpretation of this is that software can be defined as follows:
• Pre-packaged or reproduced software.
• Own account software - Software produced in-house, not destined for final sale. This can include the production of any software ‘original’ intended for subsequent re-production.
• Customised software – This refers, in the main, to made-to-order software systems.
44. 45. While this is slightly more meaningful than the SNA description it is not particularly prescriptive, and this has led to problems with international comparability which highlight the measurement difficulties. Neither definition, fFor example, neither definition defines what is meant by ‘large’ when reference is made to large databases”. Equally it is not clear how certain, everyday, software transactions should be treated, e.g. software copies acquired using licensing arrangements. Is a license, a license to use a software copy, or is it a license to use a software ‘original’? With the first interpretation a separate product (a software copy exists) but with the second a device that allows access to the ‘original’ exists, not a copy. The lack of an exhaustive definition extends beyond specific transactions such as this however, indeed, it impacts on most software transactions.
Measuring Software Investment: Amount Sold (Supply) or Amount Used (Demand)
45. . 46. Only one country surveyed (Australia) uses a demand-based method. as routine. Other countries have experimented in the past with demand approaches using standard investment surveys but these attempts have generally been dropped in favour of supply approaches, since the demand methods were viewed as unreliable; producing estimates that were thought to be implausibly low.,
46. mainly, although not exclusively, because . business accounts, often, do not record the value of own-account software at all. Business accounting rules generally recommend that own-account software should be capitalised after ‘technical feasibility’ can be established. In practice however companies adopt a very prudent interpretation of when this occurs. Indeed in the case of many software-producing companies, e.g. Microsoft, own-account originals are not valued at all.
. 47. In fact only four countries were able to provide demand-based estimates at all – Australia, Canada, France and the US – giving, in itself, some indication on the dearth of demand information available. For Australia and Canada, respectively supply estimates were respectively 7 and 4 times as high as the demand based estimates.,, for Canada they were 4 times as high, and for France (despite excluding a large proportion of software supply from their calculations) supply was about ⅓ higher. For the US, the returned questionnaire referred to the US annual capital expenditure survey on businesses for 1998 which recorded an amount of 11.8 billion US$ as “capitalised software purchased separately”. This compares to the 123 billion US$ estimate made by BEA using the Supply approach.
47. 48. Every country measured own-account software using an input-method (except Japan, which was not able to provide estimates). In principle this is supposed to include all production costs, intermediate or primary income, (such as wages and salaries). However not all countries do.
49. Para A.1.2.3 pg. 30 = no record of own account
50. However business accounts rarely record the value of own-account software at all. Business accounting rules generally recommend that own-account software should be capitalised after ‘technical feasibility’ can be established. In practice companies adopt a very prudent interpretation of when this occurs. Indeed in the case of most software-producing companies, e.g. Microsoft, own-account originals are not valued at all.
51. Para A.3.2.1 p.47 = estimate of supply
52. The underlying principle for estimating investment in software using as Supply approach is simple. Investment in software is calculated as:
Total Domestic Supply + Imports
minus
Exports, Households’ expenditure, software purchased for bundling, subcontracted software, and software included within own-account production valuations.
(All at purchaser’s prices. Pre-packaged software (licenses-to-use) and customised software only, assuming an intention to use that software for more than one year and no ‘small tools’ purchases ).
48.
53. In practice, however, estimation is a little more complicated for a number of reasons. (1) It is not always possible to differentiate between pre-packaged and customised software and other computer services within total supply of computer services. (2) As demonstrated earlier iImport data is does not always separately identifiabley as software. And (3) Software purchased for bundling and subcontracted software expenditure are usually not separately identifiable.
49. 54. However tThese problems are not irresolvable. (1) Although information for pre-packaged software and customised software may not be directly observable, information is collected using standard product classifications (such as the European CPA classification) that broadly allow these categories of software to be identified[?]. - namely CPA 72.20.2, 72.20.32, 72.20.33. (Although this includes some software that does not satisfy asset requirements, for example payments for licenses to reproduce, which are included within CPA72.2, and so need to be excluded., see Annex A.) (2) In practice, many countries are able to estimate software contained within imports, for example Canada includes other ‘software’, reflecting software royalties, within its supply estimates. Finally, (3) Iit is possible to estimate bundled software, at least in a harmonised way across countries. The OECD final report recommends that 50% of all expenditure on pre-packaged software by the computer hardware industry should be assumed to be payments for bundles. In theory a similar adjustment could also be applied to the publishing and wholesale/retail industries; although anecdotal evidence suggests that the values involved are not significant.
55. Para A.3.2.2. through A.3.2.4 p.47 to 48
56. The following table provides a practical and simple guide to software estimation using the practical working assumptions set out above, providing links to US SIC and European CPA classification systems. (For further information see Chapter 5 of the OECD Final Report).
|Value of sales of capitalisable software services (SIC 73.71 + SIC 73.72; CPA 72.20.2 + 72.20.32 + 72.20.33 + |A |
|72.20.34), including royalties and license fees, including games software | |
| Inclusion of imports (including royalties and license fees and games) |B |
| Inclusion of trade margins and taxes on domestic supply and imports |C |
| Exclusion of software embedded by hardware industry (50% of purchases of pre-packaged software by hardware |D |
|industry), treated as intermediate consumption | |
| Exclusion of sub-contracting flows between “software companies” |E |
| Exclusion of household consumption in games and other pre-packaged software |F |
| Exclusion of exports (including royalties and license fees and games) |G |
| Exclusion of maintenance (CPA 72.20.34, 10-15% of SIC 73.71) |H |
|Total GFCF in purchased software |A+B+C-D-E-F-G-H |
|‘Capitalisable’ software – Supplied software that satisfies the asset requirements set out in Section 3 and in Annex A. |
Accounting for imports
50. 57. Para 2.4.1 through 2.4.7 p.11
58. Accounting for imported software is a not insignificant part of the Supply method but, Iin practice, there are significant difficulties in measuringement difficultiesproblems with the recording of imports of software, whether payments are for customised software and, royalty payments for the rights to reproduce a software ‘original’ (for example Microsoft Ireland paying Microsoft US for the rights to reproduce Windows 2000) or any other form of software. This mainly reflects the fact that ‘computer software’ is not well identified in current international trade codes or balance of payments (BOP) items. FClearly, for improved international comparability, software needs to be separately identifiable in international trade classification systems.
51. 59. For a Supply based method it is important to separately identify software royalties, since these may include payments for imported software that should, arguably, be recorded as investment. For example, companies may purchase software copies from abroad by purchasing licenses that allow them to use software delivered electronically, and under most conditions (see Annex A, A.2.5), the view of the Task Force is, that these payments should be recorded as investment, and, arguably as trade in goods[?][?]. On the other hand payments may be for the right to reproduce software, and in these cases, the Task Force took the view that these payments should be recorded as intermediate consumption of software, (see Annex A, A.2.3).
52. 60. Table 41 below shows data on software goods, computer services and software royalties for a number of OECD countries. Columns 2 to 4 show the values used in BoP statistics and Column 6 shows estimates of imported and exported software used in supply-use tables. Some countries, for example Greece, have consistent estimates suggesting that Supply based methods would not be affected by any misclassification of software. However others. E.g. Canada, Japan, and the UK do not.
61. Column 5 shows that most countries are not able to separately identify software royalties from total royalties in their BoP statistics, and for those that do, the contribution varies significantly, 19% in Canada to 59% in Denmark (for imports). How important this is depends on how much expenditure is on licenses-to-use, as opposed to licenses-to-reproduce. If predominantly the latter then there is less cause for concern. (as far as investment levels estimated using a Supply approach are concerned, see Annex A, A.2.3).
62. Clearly for improved international comparability software needs to be separately identifiable in international trade classification systems. Chapter 3 of the OECD Final Report on Software provides more detail and recommendations on the types of changes needed to achieve this.
Estimating Pre-packaged and Customised Software
53. 63. The SNA stipulates that (most) expenditure on products expected to be used in production for more than one year is an investment, where ownership (real or imputed) of the asset is transferred from the seller to the buyer. However, most purchases of software reproductions usually include a license that permits the purchaser (only) to use the software and so, arguably, ‘ownership’ is never transferred, stating that ownership resides with the owner of the software ‘original’. From a users’ perspective however an asset has been acquired. Statistical Offices in different countries overcome this dichotomy differently. For improved international comparability however a common interpretation is needed.
54. For customised software however the licensing arrangements are, in practice, less problematic since the production and subsequent delivery of software usually involves a transfer of full ownership. Where problems do arise, they are located in the as far as international comparability is concerned, is in the descriptions used for, and interpretation of, different types of software and computer services. The SNA does not describe what software is in a physical sense and, since the descriptions and interpretations used for software vary considerably across countries, this is particularly problematic for harmonisation. For example some countries capitalise ‘consultancy services’ but others do not.
55. Table 5 provides a simple guide to software estimation (for purchased software), building on the simple principles set out above, providing links to US SIC and European CPA classification systems. (For further information see Chapter 5 of the OECD Final Report).
Estimating Customised Software
64. The descriptions for different types of software and computer services play a big role in determining whether any expenditure on software is capitalised or not. The SNA does not describe what software is in a physical sense and, since the descriptions and interpretations used for software vary considerably across countries, this is particularly problematic. Consequently, some countries capitalise ‘consultancy services’ but others do not.
65. [ Needs more ]
Estimating Own-Account Software
56. . 66. Business accounting rules generally recommend that own-account (in-house production) software should be capitalised after ‘technical feasibility’ can be established. In practice however companies adopt a very prudent interpretation of when this occurs. Indeed in the case of many software-producing companies, e.g. Microsoft, own-account originals are not valued at all.
57. Because of this and the fact that there is no observable market price for own-account produced software, it is necessary to estimate the value of the software using an input-based methodOwn-account production is estimated in nearly all countries on the basis of the costs of some, or all, of the inputs used in the producion process:. In principle this is supposed to include all production costs, intermediate or primary income, (such as wages and salaries). However not all countries do include all costs, and certainly do not using the same underlying assumptions.. For example some countries include own-account production of ‘original’ software intended for reproduction (e.g. Microsoft Word) but others, until recently, have not. The BEA has recently begun to estimate the value of these originals within its estimates. e.g. remuneration of staff-involved in production. However the methods and assumptions used across countries vary significantly, (see paragraph 2.3.5).
67. insert para A.3.3.1 through 4.3.3.5 excluding para A.3.3.3
68. A.3.3 Own-account software
58. Because there is no observable market price for own-account produced software it is necessary to estimate the value of the software using an input-based method. Indeed SNA93 recommends that own-account production of software is valued at its estimated basic price or at its costs of production if it is not possible to estimate the basic price (10.92), .
TthereforeIn practice,, own-account software can be defined and estimated as the following:
Total number of employees working on own-account software production *
Average remuneration *
Proportion of time spent on own-account development +
Other intermediate costs used in own-account production +
Notional operating surplus related to own-account production.
59. 69. Average remuneration, or compensation of employees, should include wages and salaries, social contributions (including imputed social contributions) and any related compensations-in-kind.
60. 70. Not all of the time of each employee working on own-account production ithin ISCO 213 will be spent on own-account production. Some of their time will be spent working on software to be sold directly to an external customer (customised software that is, and not production of ‘originals’ for reproduction, which are own-account production), or, in conducting other in-house, but not own-account, activities such as maintenance work. Therefore any time not spent on own-account production has to be excluded. The US adopts a 50 percent deduction rule. The 50 percent share originates from a 20-year old study on the share of software development and maintenance costs in 487 business organizations reported in a study by Barry Boehm (Boehm, 1981)[?]. Other countries apply the same percentage e.g. Canada, France and Italy.
71. These two areas relate to the lack of a definitive conceptual framework for software in the SNA.
Building Capital Stocks
61. 72. For analysing the economic impact of software on growth the annual investments in assets like software need to be aggregated over times so that an estimate of the capital stock can be built up. This requires adjusting prices paid for the ofor the asset over time for any inflation (deflation) and for any change in the quality of the asset (e.g. so that orangesapples can be added to apples) – a process called deflation. Construction of the stock of a particular asset also requires deriving estimates of the depreciation that have may have occuroccurred over time. With these three pieces of information – the annual investment flow, the price deflator and its depreciation rate -, a stock of the asset can be calculated. Doing so for software is a non-trivial exercise as seen below in the example of the practices employed to derive price indices for software.
Deflators
62. 73. Measuring price changes in software is inevitably difficult, reflecting its intangible nature and its rapidly changing characteristics but doubly complicated by the fact that own-account software production has no observable market price.
63. 74. Figure 33 below compares price indices (fixed to 1995=100) across countries. The differences are significant. For example the index for Australia fell about 30 percent% by 2000 but rose by about the same amount for Sweden. This largely reflects the dearth of price index information available in this area, and the alternative proxy deflators used, for example: general inflation price indices; office machinery price indices; the US pre-recorded software price index; and input methods (see Table 62X below).
64. 75.
76.
Of the three software-types described earlier, observedactual price indices are generally only available for pre-packaged software, and even here currently only in the US (so far). The US experience is a good example of the difficulties involved: weighted hedonic and ‘matched-pair’ models[?] [?] were used for 1985-93 data, and matched-models with a (3.3%) bias adjustment thereafter; (based on the observed difference between hedonic and matched-pair indices in the years both were available; which differed significantly).
65. 77. For the period (1985-93) the hedonic index was not applied alone because of concern that it would overstate price falls. When the characteristics of high-priced packages with limited sales were incorporated with lower priced packages with much greater sales, values derived from the high-priced packages had too high a weight. On the other-hand because the ‘matched-pairs’ price index is based only on overlapping products in two periods (ignoring new products and so not sufficiently representing actual products sold) price movements tend to be underestimated, hence the ‘averaging’ of the two methods.
66. 78. Price indices for customised software and own-account software are generally not available in any country, instead countries largely use average earnings indices (sometimes weighted with pre-packaged indices) as proxies. Eurostat draft regulations recommend that a ‘model’ or ‘representative’ pricing approach should be used for both the customised and own-account cateagories where a “representative” software product is used as a proxy for price and quality changes for the broader category. But it is likely to be some time before countries are able to develop and implement price indices based on this approach. In the meantime therefore, at least, second-best methods are usedencouraged, such asfor example:
• The use of average earnings indices in the computer software sector for customised and own-account software for own-use; and
• The US index for pre-packaged software adjusted for exchange rates, for software reproductions and ‘originals’.
67. 79. Encouraging the use of second-best methods in the short to medium term may seem unsatisfactory, but they are preferable and to some extent it is, but when to the set of against the deflators currently in use across countries, even second best solutions represent a significant improvement, (see Table 6).2 below).
80. [to be added:
81. Needs a section on depreciation;Consumption of fixed capital
. Every country surveyed used a perpetual inventory method (PIM) to calculate capital stocks however some differences exist in other areas: asset lives; depreciation functions and mortality functions, (see Table 7). However as far as asset lives are concerned a central range of between 3-6 years is discernible, both for own-account and customised software, and packaged software. Differences in depreciation and mortality functions also vary but (given the relatively short asset-lives) differences here are not likely to make as much of an impact on capital consumption estimates. In any case most countries use similar depreciation patterns (straight line). Differences in mortality functions are related to the choice of depreciation function, and are in any case of secondary importance
▪ 82. Nneds para on US seen as best practice for the world, but nevertheless is far from perfect – see 2000 revisions – if known in advance may have provided an indication of impending recession March 2001]
83. [move?] To use as criticism of US, but RoW is worse.] Even in the US where arguably the measurement of software is most developed, the recent, relatively large downward of revision to output from 1998 to 2001: Q1, provoking some to call into question how strong the US economic performance was during the 1990s (FT), is attributed to the mismeasurement of software (BEA).
84. ===========
Conclusion
85. [To be completed]Even in the US where arguably the measurement of software is most developed, the recent, relatively large downward of revisions to annual output from 1998 to 2000, dropping the GDP growth rate by nearly a full-point in 2000. About one-quarter of the downward revision was due to lower than estimated software investment.[?]
. It is clear that at present estimates of software investment across countries are largely incomparable because of inconsistencies in the assumption used to estimate software, and this hampers analysis that focuses on estimating, and comparing, the actual contribution of ICT to growth across countries. National Statistical Offices are currently reviewing their procedures for estimating software. Indeed, even in the US, where arguably the measurement of software is most developed (and certainly this is true for price indices), this is welcome. The importance of software in this context can be seen in the recent, relatively large upward revisions to annual output from 1998 to 2000, dropping the GDP growth rate by nearly a full-point in 2000. About one-quarter of the downward revision was due to lower than estimated software investment.[?]
Figures
TABLES
Table 1. Percentage point contribution of ICT to output growth
Business sector, national price index1, 1980-2000 or latest available year
|National price index |
|[pic] |
|Table 2 (cont.) Percentage point contribution of ICT to output growth |
|Harmonised price index |
|[pic] |
Notes: Output is Gross Domestic Product, business sector, factor cost (OECD, ADB database); capital services refer to the accumulation of seven assets (software, equipment and non-residential structures) from National Accounts. Software is not included for Canada; only “order-made” software is included for Japan.
Source: Colecchia and Schreyer, 2002.
Table 2. Comparison with country-specific studies
| |Data source |Periods |ICT contribution to output growth (percentage |Main methodological differences |
| | | |points) | |
| | | |Software |IT Equipment |Communi-cations| |
| | | | | |equipment | |
|United States |
|This study |NIPA (BEA) |1990-95 |0.14 |0.20 |0.08 |Hyperbolic age-efficiency profile|
| |July 2001 revision |1995-00 |0.25 |0.47 |0.15 |Assets do not include land and |
| | | | | | |inventories |
| | | | | | |7 types of assets |
|Oliner and Sichel |NIPA (BEA) and BLS |1991-95 |0.25 |0.25 |0.07 |Hyperbolic age-efficiency profile|
|(2000) | |1996-99 |0.32 |0.63 |0.15 |58 types of assets |
|Jorgenson and Stiroh|NIPA (BEA) and BLS |1990-95 |0.15 |0.19 |0.06 |Output includes service flows |
|(2000) |with extensions |1995-98 |0.21 |0.49 |0.11 |from owner-occupied housing and |
| | | | | | |consumer durables |
| | | | | | |Geometric age-efficiency profile |
|France |
|This study |INSEE |1990-95 |0.02 |0.07 |0.04 | |
| |August 2001 |1995-00 |0.08 |0.11 |0.08 | |
|Cette et al. (2000) |INSEE |1989-95 |0.05 |0.09 |0.03 |Based on exchange rate-adjusted |
| | |1995-99 |0.09 |0.13 |0.05 |BEA deflator for IT equipment, |
| | | | | | |software |
|Heckel et al. (2000)|INSEE |1987-98 |0.32 |Not based on national accounts as|
| | | | |statistical source for capital |
| | | | |data |
|United Kingdom |
|This study |Oulton (2001) |1990-95 |0.31 (total ICT equipment) |Current price software |
| | |1995-99 |0.51 (total ICT equipment) |expenditure follows the ‘low |
| | |1999-00 | |software’ option presented by |
| | | | |Oulton. |
|Oulton (2001) |ONS and Bank of |1990-95 |0.34 (total ICT equipment) |Geometric age-efficiency profile |
| |England estimates |1995-98 |0.41(total ICT equipment) |Exchange rate adjusted US-BEA |
| | | | |deflators for ICT investment |
Source: Colecchia and Schreyer, 2002.
Table 3. Contributions to Growth in Labour Productivity: US
| |1974-1990 |1991-1995 |1996-2001 |Post-1995 change |
| |(1) |(2) |(3) |(3) minus (2) |
| | | | | |
|Growth of Labour Productivity |1.36 |1.54 |2.25 |0.71 |
| | | | | |
| Contributions from: | | | | |
|Capital Deepening |0.77 |0.52 |1.17 |0.65 |
| ICT |0.41 |0.46 |0.97 |0.51 |
| Computer hardware |0.23 |0.19 |0.50 |0.31 |
| Software |0.09 |0.21 |0.34 |0.13 |
| Communication equipment |0.09 |0.05 |0.13 |0.08 |
| Other Capital |0.37 |0.06 |0.20 |0.14 |
| | | | | |
| Labour Quality |0.22 |0.45 |0.25 |-0.20 |
| | | | | |
| Multifactor productivity |0.37 |0.58 |0.83 |0.25 |
| ICT |0.26 |0.41 |0.73 |0.32 |
| Semiconductors |0.08 |0.13 |0.42 |0.29 |
| Computer Hardware |0.11 |0.13 |0.18 |0.05 |
| Software |0.04 |0.09 |0.09 |0.00 |
| Communication equipment |0.04 |0.06 |0.04 |-0.02 |
| Other Sectors |0.11 |0.17 |0.10 |-0.07 |
| | | | | |
| Total ICT contribution |0.68 |0.87 |1.70 |0.83 |
Source: Oliner, Stephen and Daniel Sichel (2002), “Information Technology and Productivity in the United States: Where Are We Now and Where Are We Going?” Presented at the IAOS 2002 Conference, 28 August, London. Federal Reserve Working Paper #2002-29 at pubs/feds/2002/index.html
|Table 4: Trade in Software: Goods, Services and Royalties - US$ |
|Software Product |Software Goods|Computer |Software |Total |Royalties % of|Supply-Use |
| | |services |Royalties |software |Total | |
|Country | | | | | | | |
|Australia |
| |
|Table 5: Estimating Purchased software investment – The Supply approach |
|Value of sales of capitalisable software services (SIC 73.71 + SIC 73.72; CPA 72.20.2 + 72.20.32 + 72.20.33 + |A |
|72.20.34), including royalties and license fees, including games software | |
| Inclusion of imports (including royalties and license fees and games) |B |
| Inclusion of trade margins and taxes on domestic supply and imports |C |
| Exclusion of software embedded by hardware industry (50% of purchases of pre-packaged software by hardware |D |
|industry), treated as intermediate consumption | |
| Exclusion of sub-contracting flows between “software companies” |E |
| Exclusion of household consumption in games and other pre-packaged software |F |
| Exclusion of exports (including royalties and license fees and games) |G |
| Exclusion of maintenance (CPA 72.20.34, 10-15% of SIC 73.71) |H |
|Total GFCF in purchased software |A+B+C-D-E-F-G-H |
|‘Capitalisable’ software – Supplied software that satisfies asset requirement rules, see Section 3 of the OECD Task force final |
|report.. |
|Table 6: Comparison of deflators used for software |
|Country |Own-account |Customised |Pre-packaged |
|Australia |Prices are assumed to fall by 6% a year. |
|Canada |Weighted average (2:1) of |Weighted average of own-account and|Average of US index for |
| |programmer labour costs and |pre-packaged (1:3). |pre-packaged adjusted for exchange |
| |non-labour inputs to the computer | |rates. A new index is due for |
| |services industry. | |release next year. |
|Czech Republic |Price indices for the output of the computer services industry. |
|Denmark 1993-95 |Weighted average of labour costs and PC hardware (1:1). |
|1996-97 | |
|1998 + | |
| |Weighted average labour and PC hardware (3:1). |Weighted labour and PC hardware |
| | |(1:1). |
|Denmark 1993-95 |Weighted average of labour costs and PC hardware (1:1). |
|1996-97 | |
|1998 + | |
| |Weighted average labour and PC hardware (3:1). |Weighted labour and PC hardware |
| | |(1:1). |
| |Geometric average of labour and hardware (3:1). |
|Finland 1975-97 |Average earnings index for the computer services industry. |
|1998 + | |
| |Weighted average of labour costs of the computer services industry and US pre-packaged software index |
| |adjusted for exchange rates. |
|France 1995 (-) |US price index adjusted for exchange rates. |
|1995 + | |
| |Labour costs. |
|Greece |General (whole inflation) price index. |
|Japan |Corporate Service Price Index for “the development of computer software tailored for corporations”, based |
| |on the labour costs. |
|Netherlands |Labour costs of ICT personnel. |Producer price index. |Producer price index. |
|1. Spain |Based on producer price index for office machinery and the general consumer price index (excluding |
| |renting). |
|Sweden |Average earnings index for the computer services industry. |
|UK |Average earnings series adjusted for the computer services industry with 3% productivity adjustment since |
| |1996. |
|US |Weighted average (roughly 1:1) of |Weighted average of own-account and|Directly collected price index (see|
| |programmer labour costs and |pre-packaged (1:3). |above). |
| |non-labour inputs to the computer | | |
| |services industry. | | |
|Table 7: Capital consumption and Asset lives |
|Country |Asset lives |Depreciation function |Mortality function |
| |Own-acc’t & |Pre-recorded/ | | |
| |Customised |packaged | | |
|Australia |Pre 89/90 - 8 |6 |Hyperbolic for age efficiency |Skewed retirement for packaged & other |
| |Post 89/90 -6 |4 |function | |
|Canada |5 |3 |Straight line |Truncated normal |
|Czech Republic |5 |Business accounts |Any |
|Denmark |6a |4b |Straight line |Winfrey S3 |
|Finland |5 |Straight line |Skewed Weibull |
|France |5 |Straight line |Lognormal. |
|Italy |5 |Straight line |Truncated normal |
|Japan |5 |Straight line |None |
|Netherlands |3 |Straight Line |Weibull |
|Spain |4 |Straight Line |Delayed linear |
|Sweden |10a |5b |Geometric |None |
|United Kingdom |5 |Straight Line |Normal |
|United States |5 |3 |Geometric |None |
1. Figure 1. Growth of NASDAQ indices and worldwide Internet hosts, February 1999 - May 2001
2. [pic]
Source: OECD based on Netsizer ( ) and NASDAQ.
3. Figure 6. Growth in Internet host computers and major software developments
[pic]
* New methodology used in January 1998.
Source: Network Wizards.
Notes
xxx
Anchordoguy, Marie (2000) “Japan’s Software Industry: A Failure of Institutions?,” Research Policy, No.29.
Ahmad, Nadim (forthcoming) “Measuring Investment in Software,” STI Working Papers, .
Baily, Martin and Robert Z. Lawrence (2001) “Do We Have a New E-conomy?” NBER Working paper 8243, .
Boehm, Barry M. (1981) Software Engineering Economics (Englewood, NJ: Prentice-Hall): 533-35, 548-50.
Bonds, Belinda and Tim Aylor (1998), “Investment in New Structures and Equipment in 1992 by Using Industries,” Survey of Current Business, December 1998.
Bosak, Jon and Tim Bray (1999), “XLM and the Second-Generation Web,” Scientific America, May.
Bosworth, Barry and Jack Triplett (2000), “What’s New About the New Economy? IT, Economic Growth and Productivity,” Brookings Institution Paper, .
Bosworth, Barry and Jack Triplett (2002), “ ‘Baumol’s Disease’ Has Been Cured: IT and Multifactor Productivity in US Services Industries”, Prepared for the Texas A&M Conference, April, brook.edu/dybdocroot/es/research/projects/productivity/workshops/20020517_triplett.pdf
Breshnahan, Timothy and Shane Greenstein, 1997, “Technical Progress and Co-invention in Computing and the Use of Computers, Brookings Papers on Economic Activity: Microeconomics.
Breshnahan, Timothy F. (1999), “Computing” in US Industry in 2000, National Academy Press, Washington, DC.
Breshnahan, Timothy F. (2001), “Prospects for an Information Technology-led Productivity Surge,” May;
Brynjolffson, Erik and Shinkyu Yang (1996), “Information technology and Productivity: A review of the Literature”, Advances in Computers, 43 (February).
Brynjolffson, Erik and Lorin M. Hitt (2002), “Computing Productivity: Firm-level Evidence,” MIT Working paper, No. 4210-01,
Information technology and Productivity: A review of the Literature”, Advances in Computers, 43 (February).
Colecchia, Alessandra and Paul Schreyer (2002) “ICT Investment and Economic Growth in the 1990s: Is the United States a Unique Case?” Review of Economic Dynamics, No. 5.
David, Paul A. (1990), “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox”, The American Economic Review, May.
Eichengreen, Barry, James Tobin and Charles Wyplosz (1995), “Two Cases for Sand in the Wheels of International Finance”, The Economic Journal, 105, January, p.162-172.
Greenspan, Alan (1998), “The Implications of technological changes”, speech delivered at the Charlotte Chamber of Commerce, Charlotte, North Carolina, 10 July,
Griliches, Zvi (1994) AEA Address.
Gordon, Robert J. (2003), “Hi-tech Innovation and Productivity Growth: Does Supply Create Its Own Demand?” NBER Working Paper 9437,
Greenspan, Alan (1999), Testimony before the Joint Economic Committee, US Congress June 14, 1999,
Hardy, Quentin (2002) “The New HP Way: World’s Cheapest Consultants,” Forbes, 12 May, .
Huckle, Thomas (2002), “Collection of Software Bugs,” .
Jorgenson, Dale (2001) “Information Technology and the US Economy,” American Economic Review, Vol.91, No.1.
Jorgenson, Dale, Mun S. Ho and Kevin J. Stiroh (forthcoming), “Lessons from the US Growth Resurgence,” presented at the First International Conference on the Economic and Social Implications of Information Technology, US Departmnet of Commerce, Washington, DC 27-28 January 2003.
Johnson, George (2001) “All Science Is Computer Science,” The New York Times, March 25.
Mann, Charles (2002) “Why Software Is So Bad,” Technology Review, July / August.
Mowery, David C. (1999), “America’s Industrial resurgence: How Strong, How Durable?”, Issues in Science and Technology, Spring.
OECD (1999), The Economic and Social Impact of Electronic Commerce, Paris.
OECD (2000), The OECD Information Technology Outlook, Paris,
OECD (2001), The International Mobility of the Highly Skilled, Paris.
OECD (2002), “The Contribution of Information and Communication technologies to Economic Growth in Nine OECD Countries”, OECD Economic Studies, No.34, 2002/1.
OECD (2002b) Measuring the Information Economy, Paris. sti/measuring-infoeconomy
OECD (2002c) Research and Development Expenditure by Industry, 1987-2000, Paris
Oliner, Stephen and Daniel Sichel (2002), “Information Technology and Productivity in the United States: Where Are We Now and Where Are We Going?” Presented at the IAOS 2002 Conference, 28 August, London. Federal Reserve Working Paper #2002-29 at pubs/feds/2002/index.html
National Academy of Sciences, 2002, “Making IT Better,” National Academy Press, Washington, DC.
National Academy of Sciences (1995) Evolving the High Performance Computing and Communications Initiative to Support the Nation’s Information Infrastructure, Washington, DC.
National Academy of Sciences (1999) Funding a Revolution: Government Support of Computer Research, Washington, DC.
National Science Foundation (1998) “International Mobility of Scientists and Engineers to the United States -- Brain Drain or Brain Circulation?,” Issue Brief, NSF 98-316, June
Saxenian, AnnaLee (1999), “Silicon Valley’s New Immigrant Entrepreneurs”, Public Policy Institute of California,
Sichel, Daniel E. (1999), “Computers and Aggregate Economic Growth: An Update”, Business Economics, April, p.19.
Sichel, Daniel E. (1997), “The Computer Revolution: An Economic Perspective”, The Brookings Institution, Washington, DC.
Stiroh, Kevin (1998), “Computers, Productivity, and Input Substitution,” Economic Inquiry 36, 175-191.
Triplett, Jack E. (1998), “The Solow Productivity Paradox: What Do Computers Do to Productivity?”, presented at the American Economic Association Meetings, Chicago, January.
US Department of Commerce, NIST (National Institute of Standards & technology) (2002) “The Economic Impacts of Inadequate Infrastructure for Software Testing” (Planning Report 02-3), May.
US Department of Commerce (2003), Bureau of Economic Analysis, National Income and Product Accounts, , NIPA table 5.8.
Wessel, David (1996), “Greenspan’s Optimism Finds Underpinning in Fed Study”, Wall Street Journal of Europe, 18 November.
Young, Alison (1996), “Measuring R&D in Services”, STI Working Papers 1996/7, Paris.
Table 10 Comparison with country-specific studies
| |Data source |Periods |ICT contribution to output growth (percentage |Main methodological differences |
| | | |points) | |
| | | |Software |IT Equipment |Communi-cations| |
| | | | | |equipment | |
|United States |
|This study |NIPA (BEA) |1990-95 |0.14 |0.20 |0.08 |Hyperbolic age-efficiency profile|
| |July 2001 revision |1995-00 |0.25 |0.47 |0.15 |Assets do not include land and |
| | | | | | |inventories |
| | | | | | |7 types of assets |
|Oliner and Sichel |NIPA (BEA) and BLS |1991-95 |0.25 |0.25 |0.07 |Hyperbolic age-efficiency profile|
|(2000) | |1996-99 |0.32 |0.63 |0.15 |58 types of assets |
|Jorgenson and Stiroh|NIPA (BEA) and BLS |1990-95 |0.15 |0.19 |0.06 |Output includes service flows |
|(2000) |with extensions |1995-98 |0.21 |0.49 |0.11 |from owner-occupied housing and |
| | | | | | |consumer durables |
| | | | | | |Geometric age-efficiency profile |
| | | | | | | |
|France |
|This study |INSEE |1990-95 |0.02 |0.07 |0.04 | |
| |August 2001 |1995-00 |0.08 |0.11 |0.08 | |
|Cette et al. (2000) |INSEE |1989-95 |0.05 |0.09 |0.03 |Based on exchange rate-adjusted |
| | |1995-99 |0.09 |0.13 |0.05 |BEA deflator for IT equipment, |
| | | | | | |software |
|Heckel et al. (2000)|INSEE |1987-98 |0.32 |Not based on national accounts as|
| | | | |statistical source for capital |
| | | | |data |
|United Kingdom |
|This study |Oulton (2001) |1990-95 |0.31 (total ICT equipment) |Current price software |
| | |1995-99 |0.51 (total ICT equipment) |expenditure follows the ‘low |
| | |1999-00 | |software’ option presented by |
| | | | |Oulton. |
|Oulton (2001) |ONS and Bank of |1990-95 |0.34 (total ICT equipment) |Geometric age-efficiency profile |
| |England estimates |1995-98 |0.41(total ICT equipment) |Exchange rate adjusted US-BEA |
| | | | |deflators for ICT investment |
| | | | | |
-----------------------
[1] The views expressed here are those of the author and do not necessarily reflect the opinions of the OECD or its Member countries.
[2] . This is partly due to the fact that software investment in Japan in underestimated. See footnote 4.
[3] See p.47 “Use of the Internet by Individuals” in OECD, 2002b.
[4] Extensible Markup Language (XLM), created in 1998, is the next step forward that provides a standard method for describing (tagging) information so it can be exchanged across the Internet more efficiently (Bosak and Bray, 1999).
[5] For example, the Sustainable Computing Consortium. See
[6] . See also This section is directly drawn from AAhmad 2003, (forthcoming).
[7] Part of the reason is the difficulty in achieving exactly comparable definitions of computer services, although this cannot explain the differences for EU countries that use exactly the same definition. At the more detailed level differences are starker. For example for a given expense of 100 on similar (detailed) types of software services, the US will capitalised 100, while France will capitalise only 50.
[8] Namely CPA 72.20.2, 72.20.32, 72.20.33. Although this list includes some software that does not satisfy asset requirements, for example payments for licenses to reproduce, which are included within CPA72.2, and so need to be excluded.
[9] See "International Merchandise Trade Statistics - Concepts and Definitions" 27. Goods used as carriers of information and software. (HS heading 85.24) This category includes, for example, (a) packaged sets containing diskettes or CD-ROMs with stored computer software and/or data developed for general or commercial use (not to order), with or without a users' manual, and (b) audio- and videotapes recorded for general or commercial purposes….. However, (i) diskettes or CD-ROMs with stored computer software and/or data, developed to order, (ii) audio- and videotapes containing original recordings, and (iii) customised blueprints etc. are to be excluded from international merchandise trade statistics.
[10] Matched-pair models measure the relative price change by selecting products whose qualities and characteristics are constant over two periods of time. Where these qualities/characteristics change rapidly however this may not be possible. Hedonic price indices attempt to overcome this by employing regressions between any product and its qualities/characteristics such that relative price changes with constant quality can be measured. However hedonic price indices are extremely data-intensive and the regressions require regular re-evaluation.
[11]. US Department of Commerce, Bureau of Economic Analysis, 2001, BEA News Release, “National Income and Product Accounts Second Quarter 2001 GDP (Advance) Revised Estimates: 1998 through First Quarter 2001” .
[12]. US Department of Commerce, Bureau of Economic Analysis, 2001, BEA News Release, “National Income and Product Accounts Second Quarter 2001 GDP (Advance) Revised Estimates: 1998 through First Quarter 2001” .
[i]. Part of the reason is the difficulty in achieving exactly comparable definitions of computer services, although this cannot explain the differences for EU countries that use exactly the same definition. At the more detailed level differences are starker. For example for a given expense of 100 on similar (detailed) types of software services, the US will capitalised 100, while France will capitalise only 50.
[ii] See "International Merchandise Trade Statistics - Concepts and Definitions" 27. Goods used as carriers of information and software. (HS heading 85.24) This category includes, for example, (a) packaged sets containing diskettes or CD-ROMs with stored computer software and/or data developed for general or commercial use (not to order), with or without a users' manual, and (b) audio- and videotapes recorded for general or commercial purposes (see paragraph. 123 below for recommendation on valuation).….. However, (i) diskettes or CD-ROMs with stored computer software and/or data, developed to order, (ii) audio- and videotapes containing original recordings, and (iii) customised blueprints etc. are to be excluded from international merchandise trade statistics (see paragraph. 48 below)..
123. There are international transactions which present special difficulties or questions regarding valuation of the goods involved. Some of the difficulties are due to the complexity of the transaction or the peculiarity of the goods. (...) The valuation of all goods should be made in accordance with the WTO Agreement on Valuation and the recommendations contained in the present publication (see paragraphs. 116 and 121 above). In addition, it is recommended that: ….. (b) Goods used as carriers of information and software, such as packaged sets containing diskettes or CD-ROMs with stored computer software and/or data developed for general or commercial use (not to order), be valued at the their full transaction value (not at the value of the empty diskettes or CD-ROMs, paper or other materials (see paragraph. 27 above))
[iii] Barry W. Boehm, Software Engineering Economics (Englewood, NJ: Prentice-Hall, 1981): 533-35, 548-50
[iv]. Matched-pair models measure the relative price change by selecting products whose qualities and characteristics are constant over two periods of time. Where these qualities/characteristics change rapidly however this may not be possible. Hedonic price indices attempt to overcome this by employing regressions between any product and its qualities/characteristics such that relative price changes with constant quality can be measured. However hedonic price indices are extremely data-intensive and the regressions require regular re-evaluation.
-----------------------
[pic]
[pic]
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related download
- chinese attitudes towards work and vocation a north
- history enhanced scope and sequence whi
- densho civil liberties curriculum constitutional issues
- towards a new economy ministerial report on
- info sheet 3 asian american demographics in queens
- homepage u s national park service
- the shaping of north america
Related searches
- new york school report cards
- new kentucky school report card
- new york school report card
- new york state report card
- new jersey school report card
- new york state report cards
- make a new email account on yahoo
- why a traditional economy is best
- how does a command economy work
- in a market economy quizlet
- an example of a market economy is
- why is a good economy important