[Job Market Paper]
ICT and Global Economic Growth By Khuong Vu November 2004
Abstract How and to what extent has the Information and Communication Technology (ICT) revolution impacted economic growth? This question has been well addressed for the US and several developed countries but remains challenging for the global economy. This paper examines the impact of ICT on the global economy using a large sample of economies. The study reveals that ICT contribution to economic growth is a global phenomenon, which is evident not only in developed economies but also in developing ones. The ICT contribution to growth in most economies drastically increased from the period 1990-1995 to the period 1995-2000, while its variance was also strikingly widened. The key determinants of the variance in ICT contribution to growth across economies are education, institutional quality, openness, and English fluency. ICT has impact on economic growth not only in quantity but also in quality. ICT capital is superior to Non-ICT capital in enhancing economic growth: a higher level of ICT capital stock per capita allows a typical economy to achieve a higher growth rate for given levels of growth in labor and capital inputs.
JEL Classification: O3, O4 Key words: ICT, Economic Growth, Capital Stock
Program on Technology and Economic Policy, Harvard Kennedy School of Government; email:
[email protected]. I am grateful to Dale Jorgenson, Dwight Perkins, Robert Jensen, and Robert Barro for advising me on this research. I would also like to thank seminar participants at the Conference Board and the World Bank E-Development Group for their helpful comments.
Table of Contents I. INTRODUCTION..................................................................................................................................... 3 I.1.THE ICT REVOLUTION AND THE GLOBAL ECONOMY ............................................................................ 3 I.2.DATA SOURCES AND EMPIRICAL STRATEGY ....................................................................................... 12 II. DECOMPOSING THE SOURCES OF GROWTH ........................................................................... 14 II.1. METHODOLOGY ................................................................................................................................ 14 II.2. MEASURING ICT INVESTMENT FLOWS.............................................................................................. 16 II.3. CAPITAL STOCKS .............................................................................................................................. 18 II.4. CAPITAL SERVICES ........................................................................................................................... 21 II.5. HOURS WORKED AND LABOR QUALITY ............................................................................................ 25 II.6.THE SOURCES OF OUTPUT GROWTH .................................................................................................. 29 III. ICT AS A SOURCE OF ECONOMIC GROWTH........................................................................... 34 III.1. GLOBAL ECONOMIC GROWTH ......................................................................................................... 34 III.2.MAGNITUDE OF ICT CONTRIBUTION TO OUTPUT GROWTH .............................................................. 35 III.2.THE SHARE OF ICT CONTRIBUTION IN THE OVERALL RATE OF OUTPUT GROWTH ........................... 42 IV. DETERMINANTS OF ICT CONTRIBUTION TO OUTPUT GROWTH .................................... 44 IV.1.FACTORS INFLUENCING THE MAGNITUDE OF ICT CONTRIBUTION TO GROWTH ............................... 44 IV.2. RESULTS .......................................................................................................................................... 51 V. IMPACTS OF ICT ON ECONOMIC GROWTH.............................................................................. 55 V.1. ICT AS A DETERMINANT OF OUTPUT GROWTH ................................................................................. 55 V.2. IMPACT OF ICT ON THE QUALITY OF ECONOMIC GROWTH ............................................................... 59 VI. CONCLUSION..................................................................................................................................... 64 BIBLIOGRAPHY ...................................................................................................................................... 65 APPENDICES ............................................................................................................................................ 75
2
I. Introduction
I.1.The ICT revolution and the Global Economy The remarkable progress in Information and Communication Technology1 (ICT) witnessed in the past decade has had an increasingly profound impact on economic activity and the way people work and communicate across countries around the world.
While there are still debates on whether the impact of ICT matches that of the Great Inventions, such as the steam engine, electricity, and the railroad, in terms of transforming social and economic structures, factual evidence of ICT impact on the economy is compelling in all three dimensions of a typical analysis: Input, Process, and Output.
On the input side, technological progress in the ICT sector in the last few decades was remarkable. Gordon Moore in 19652 projected the rapid pace of technology innovation, making a then-unbelievable prediction that computer processing power, or the number of transistors on an integrated chip, would double nearly every two years. This statement, known as Moore's Law, has held firm in the last four decades. The number of transistors on integrated circuits increased from 50 in 1965 to 29,000 in 1981 (Intel i8088), to 125 million in 2004 (Intel Pentium 4). Computer chips have made their way into modern 1
ICT is defined as Information Technology (IT) plus Telecommunication Equipment and Services. The IT, in turn, refers to a combined industry, which includes IT hardware (office machines, data processing equipment, data communication equipment), IT software, and IT services (WITSA, 2000). 2
“ Cr a mmi ngmor eCompon e nt son t oI n t e g r a t e dCi r c u i t s , ”Electronics, Volume 38, Number 8, April 19, 1965
3
products ranging from computers to greeting cards, from automobiles to microwaves, and played a major role in driving down the costs and improving the performance of ICT products; the price of IT hardware declined by over 2000 times over the 40-year period from 1960 to 2000; the semiconductor industry has grown from infancy in the late 1950s to more than $200 billion in annual revenue today and it has been at the center of the explosive growth of the ICT sector.
On the process side, the last decade witnessed a rapid increase in ICT spending and investment. The world ICT expenditure nearly doubled, from $1,300 billion in 1993 to $2,400 in 2002 (WITSA, 2002) at a compound annual growth rate of 8 percent, which far exceeded the growth rates in the same period of global gross product and international trade, which are about 3% and 5%, respectively. The ICT intensity, measured as a percentage of ICT spending in GDP, increased (on average for 50 countries with available data) from 4.4 percent in 1992 to 7.0 percent in 2000.3 The computer penetration rate, aggregated for all 207 countries and territories in the world, increased 7.6 times, from 1 per 100 inhabitants in 1990 to 7.6 per 100 inhabitants in 2000; the pace was even faster for mobile phones (57 times, from 0.26 to 14.9) and the Internet (1600 times, from 0.004 to 6.4).4 In addition, the number of Internet hosts rose 300 times, from about 350,000 in 1990 to nearly 110 million in 2000. Numerous individuals, firms, industries, and nations have aggressively embraced the opportunities brought about by
3
Computed from WDI (2002).
4
ibid.
4
the ICT revolution, especially in communication, marketing, and learning, and achieved phenomenal success.5
On the output dimension, the impact of ICT on economic structure and economic growth performance has been palpable in a number of countries. Particularly, in the U.S., labor productivity revived in the 1990s with a considerable acceleration during the period 1995-2000, and ICT capital input accounted for more than one fifth of GDP growth during the periods 1990-95 and 1995-2000 (Council of Economic Advisors, 2001; Jorgenson and Stiroh (2000); Oliner and Sichel, 2001). The contribution of ICT to growth was also significant in Australia (Parham et al, 2001), Canada (Armstrong et al, 2002; Khan and Satos, 2002), Korea (Kim, 2002), United Kingdom (Oulton, 2001), Finland (Jalava and Pohjola, 2002), and the Netherlands (Van der Wiel, 2002). In particular, Ireland’ se xt r e mes uc c e s si nattracting an enormous amount of FDI to its ICT sector in the 1990s6 and that small country has quickly become the eighth largest exporter of computer equipment and the fifth largest producer of software in the world (Tallon and Kraemer, 2003). India enjoyed dramatic growth in the software industry with its software exports increasing from $105 million in 1990 to $6.2 billion in 2000 and $9.2 billion in 5
Here are a few prominent examples: Dell, a U.S. computer vendor founded in 1984, has become a $35 billion company and market leader in the U.S. computer market through its unique web-based marketing model (Dell, 2003); Legend, a Chinese computer maker, founded in 1984 with less than $100,000, produced 1 million PCs in 1998 and had a market capitalization value of $3 billion in 2003 (Legend, 2003); Nokia, a Finnish telecom producer, by capturing the dramatic technological progress and market growth in the mobile phone industry, increased its sales from $4 billion in 1994 to $20 billion in 2000 and has become a major contributor to the Finish economy, composing 4 percent of GDP, 25 percent of export, and 5 percent of manufacturing employment (Finland, 2003; Nokia, 2003). 6
For example, Intel invested $2.5 billion in Ireland during 1990-1997 (Intel, company website, www.intel.com, October 10, 2003).
5
2002. I ndi a ’ ss of tware industry has surpassed major traditional industries, such as steel a nda ut omot i ve ,t obe c omet h ec ount r y ’ sl a r g e s tvalue-added industry (Nasscom, 2003; Dataquest).
All the above evidence and studies suggest that ICT might have had a significant role in fostering economic growth in both developed and developing countries. However, measuring and assessing the impact of ICT on economic growth remains a challenging task, even for the US economy. Robert M. Solow, Economics Nobel Laureate, in 1987 r a i s e dac onc e r na boutt he“ pr oduc t i vi t ypa r a dox” :“ y ou c a ns e et hec omput e ra g e everywhere but in the productivity statistics.”7
Lately, Gordon (2000) has argued that ICT does not measure up to the great inventions of the past in transforming society and lifting long-term productivity growth and that the TFP acceleration in the US in the late 1990s was entirely concentrated in the ICTproducing sector. Furthermore, Bosworth and Triplett (2000) contend that the impact of I CTi snots omuc h“ ne w”butj us tl a r ge rt ha nbe f or e ” .Gor don( 2002) ,a l t houg hmor e positive in assessing the impact of ICT on the productivity revival in the US, still ma i nt a i nst ha t“ c omput e rc a pi t a ldi dnotha vea nyki ndofma g i c a lore xt r a or di na r y effect—i te a r ne dt hes a mer a t eofr e t ur na sa nyot he rt y peofc a pi t a l ”( p.25) .Ot he rI CT skeptics, who make similar points, include Kiley (1999) and Roach (1998).
7
Robe r tM Sol ow,“ We ’ dBe t t e rWa t c hOu t ” ,New York Times, July 12, 1987, Book Review, No.36.
6
The potential impacts of ICT on productivity and economic growth have spawned a wealth of studies. These studies can be divided into four principal groups based on their methodological approach.
The first and major group includes studies employing the growth accounting approach to capture the contribution of ICT to output growth and labor productivity at the national level. This set of studies was initiated by the work of Oliner and Sichel (1994) and Jorgenson and Stiroh (1995, 1999) for the U.S economy. Other notable studies in this group focus on either a single nation or a group of countries. Among single country studies, Jorgenson and Stiroh (2000), Oliner and Sichel (2001, 2002), Whelan (2000), Jorgenson (2001), and Jorgenson, Ho, and Stiroh (2002) focus on the US; Oulton (2001) on the UK; Miyagawa et al (2002) and Jorgenson and Motohashi (2003) on Japan; RWI and Gordon (2002) on Germany; Cette et al (2002) on France; Armstrong et al (2002) and Khan and Satos (2002) on Canada; Parham et al (2002), Simon and Wardrop (2001), and Gretton et al (2002) on Australia; Jalava and Pohjola (2002) on Finland; and Van der Weil (2002) on the Netherlands. The studies examining a group of economies include Schreyer (2000), Colecchia and Schreyer (2001), Ark et al (2002), and Daveri (2002) for selected EU economies; and Jorgenson (2003)G7 economies.
The second group includes studies at the industry and firm levels. These studies use a combination of growth accounting methods and econometric models to examine a sample of firms or industries and single out the impact of ICT on their growth and productivity.
7
Brynjolfsson and Hitt (1995, 2000, 2003), examining US firm-level data, prove that ICT has a solid impact on productivity; furthermore, the impact of ICT is enhanced by additional investment in organizational capital and returns on ICT investment are larger over long periods. Lichtenberg (1995), inspecting US firm-level data, finds excess returns on capital and labor deployed in information systems. McGuckin and Stiroh (2000), addressing t he“ a ggr e ga t i on”bias problem, produce the results confirming that computers have a large impact on output. Nordhaus (2001) and Stiroh (2002), inspecting US industry-level data, reveal the link between IT and the US productivity revival in the late 1990s. Motahashi (2001), examining firm-level data of Japanese manufacturers and distributors, finds that the productivity impact of information network use varies according to application and timing of its introduction. O’ Ma honya ndVe c c hi (2002), investigating panel data of 31 sectors in the US and 24 sectors in the UK, provide strong evidence of the impact of ICT on TFP. Hubbard (2003) shows that that installing advanced on-board computers in trucks have increased capacity utilization by 13 percent. Pilat and Wolfl (2004) examine the roles of the ICT-producing industry and of key ICTusing industries in overall productivity growth in OECD countries; the authors find that the contribution of the ICT-producing industry is most impressive in Finland, Ireland, and Korea while certain ICT-using services in a some countries, notably the United States and Australia, have experienced a significant pick-up in productivity growth in the second half of the 1990s. Gretton et al (2004), investigating firm-level data from the Australian Business Longitudinal Survey, find positive and significant links between ICT use and productivity growth in manufacturing and a range of service industry sectors. Hempell
8
and Wiel (2004), analyzing comparable panel data for German and Dutch firms in the service sector, find that ICT use and innovation have a complementary impact on productivity growth.
Baldwin et al, studying firm-level data in the Canadian food
processing sector, reveal that process control and network communications technologies are particularly important to productivity in this sector. Arvanitis (2004), analyzing data of 1382 Swiss firms in 2000, shows that labor productivity is positively correlated with the intensity of Internet and Intranet penetration. Maliranta and Rouvinen (2004), examining firm-level data of Finish enterprises, find that ICT induces excess productivity and the excess rate is higher in services than in manufacturing. Milana and Zeli (2004), inspecting firm-level data collected by the Italian National Statistical Institute (ISTAT), reveals that ICT has a positive and significant impact on TFP growth in all industries. Finally, Atrostic and Nguyen (2002), using firm-level data from the US manufacturing sector, find that computer networks have a significant impact on productivity.
The third group of studies uses non-growth accounting approaches to investigate the impact of ICT on growth and productivity. Most studies in this group employ econometric models with national panel data for their analysis: Roller and Waverman (2001) examine the data on telephone penetration for 36 countries over 20 years (19701990) and find that telephone penetration has a significant effect on growth and that the effect is higher when the penetration passes a critical mass; Caselli and Coleman (2001) inspect the data on computer penetration for 43 countries in the period 1970-1990 and ascertain that computer penetration significantly enhances growth; Pohjola (2000) uses IT
9
spending as a proxy for IT capital and finds that net return on IT investment was much larger than net return on non-IT investment in 39 countries during the period 1980-1995. Fare et al (1994) explore a different approach, using a nonparametric programming method to analyze productivity growth in 17 OECD countries over the period 1979-1988; the authors find that US productivity growth is more likely to be driven by technological c ha ng ewhi l eJ a pa n’ sproductivity growth largely relies upon efficiency improvement.
The fourth group contains case studies proving that ICT has an impact on productivity, income, and growth in a country, a sector, an industry, or a locality. Hanna et al (1996) examine the process through which the ICT sector developed and came to play a major role in the miraculous growth and development taking place in East Asian countries, including Japan, Korea, Taiwan, Hong Kong, and Singapore. Kapur (2002) explains the causes and consequences of the boom in the Indian software sector. Tallon and Kraemer (2003) depict the case of Ireland, which is enjoying a fascinating growth spurt driven by a strategic focus on the ICT-producing sector. Danzon and Furukawa (2001) find that the Internet fosters competition and productivity in the health care industry; Eggleston, Jensen, and Zeckhauser (2001) analyze the impact of ICT on the income of rural villages in China; Clemons and Hitt (2001) show that the Internet enhances transparency, differential pricing, and disintermediation in the financial service sector in the US, McAfee(2001) examines the concrete ways that the Internet has affected the US manufacturing sector and the mechanism by which the Internet is achieving this impact; Fine and Raff (2001) find a major impact of the networked communications enabled by
10
the Internet on supply chain management in the automotive industry, which potentially boosts the efficiency of that industry, and hence the entire economy; Bailey (2001) indicates that the Internet fosters competition and efficiency in the US retail sector.
Besides the four main groups of empirical studies, there are also a number of papers providing a general view of the role of ICT in promoting economic growth and productivity. DeLong and Summers (2001) assert that the impact of ICT is important and that it will have a lasting effect on economic growth, and they project that the ICT sector will continue to grow faster than the rest of the U.S. economy. Baily and Lawrence show that the change in TFP growth in the U.S. between 1995 and 2000 was primarily structural, not cyclical, and the change was significantly related to IT-enabled services. Roeger (2001) compares the technological and economic development related to ICT in the 1990s in the US and Europe to explain why the US was better than Europe in reaping the benefits of ICT to promote economic growth.
While the previous studies have confirmed the important role of ICT in promoting economic growth, several challenging questions remain: (1) What are the magnitude and dynamics of ICT contribution to global economic growth? (2) Is ICT an important determinant of the variation in output growth across economies? (3) Is ICT superior to Non-ICT in enhancing the quality of output growth? This paper aims to answer these questions.
11
I.2.Data Sources and Empirical Strategy The main sources of data used for this study include the following: (i) The World Bank Development Indicators (2002 and 2003) for economic, social, and ICT indicators; supplementary sources for these indicators include Penn World Tables version 5.6, UN Human Development Reports (various issues), and the ITU8 Statistics. (ii) The WITSA9 Digital Planet Reports (1998, 2000, and 2002) for ICT spending data; (iii)The World Bank Governance Indicators for governance indices; (iv)The Barro-Lee Education dataset for educational attainment level (measured as average years of schooling); a supplementary source is the Cohen-Soto educational attainment dataset.
Because the WITSA reports provide information for estimating investment flows in ICT, which is essential for this paper, the 50 economies10 covered by the WITSA reports constitute the core sample of this study (Appendix A provides the list of these 50 economies with their selected indicators). The sample consists of 22 industrialized and 28 developing economies based on a categorization defined by the United Nations (UN, 2002). My paper further splits the sample into six subgroups: (i) G7 (7 economies): Canada, France, Germany, Italy, Japan, the UK, and the US;
8
International Telecommunication Union is an organization within the United Nations specializing in services related to ICT policy development and regulatory guidance. 9
WITSA stands for World Information Technology and Services Alliance –a private consortium of 48 global ICT industries. The WITSA data on ICT are based on the work of International Data Corporation (IDC), which is a global market intelligence company specializing in information and telecommunication technology with offices in 50 major ICT-spending countries. 10
According to WITSA, these 50 economies accounted for over 98% of the global ICT market in 2000.
12
(ii) Non-G7 (15 non-G7 industrialized economies): Australia, Austria, Belgium, Denmark, Finland, Greece, Ireland, Israel, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, and Switzerland; (iii) Developing Asia or Asia-11 (11 economies): China, Hong Kong, India, Indonesia, Korea, Malaysia, the Philippines, Singapore, Taiwan, Thailand, and Vietnam; (iv)Latin America or LA-6 (6 economies): Argentina, Brazil, Chile, Colombia, Mexico, and Venezuela; (v) Eastern Europe or EE-8 (8 economies): Bulgaria, Czech Republic, Hungary, Poland, Romania, Russia, Slovakia, and Slovenia; and (vi)Other Economies or Other-3 (3 economies): Egypt, South Africa, and Turkey.
My paper uses the growth accounting framework to decompose the sources of economic growth and to single out the ICT contribution. Various econometric models are then used for analyzing the determinant of variance in ICT contribution across economies and the impacts of ICT on economic growth. The remainder of this paper is organized as follows. Section II presents the methodology and steps to decompose the sources of output growth for the 50 individual economies over the two periods 1990-1995 and 1995-2000. Section III depicts a global picture of the ICT contribution during the two periods; Section IV examines the determinants of the variations of ICT contribution across economies. Section V explores the impacts of ICT on the quantity and quality of output growth. Section V concludes.
13
II. Decomposing the Sources of Growth II.1. Methodology This study decomposes the sources of economic growth using the Production Possibility Frontier (PPF) model, which was introduced by Jorgenson (1966) and first employed by Jorgenson and Griliches (1967). The model was introduced into productivity measurement by Jorgenson (1996) and has been used in numerous studies in the field. The alternative to the PPF model is the Solow Aggregate Production (SAP) model, which was first employed by Solow (1957, 1960). As an improvement to the SAP model, the PPF model allows changes in capital and labor quality to absorb a significant portion of the residual, which used to be attributed to technological progress and unexplained elements. Furthermore, the PPF model captures the substitutions among outputs of investment and consumption goods and among inputs of capital and labor. This advantage of the PPF model is more pronounced in the face of a drastic substitution of ICT for non-ICT capital in the past 10-15 years in many economies. It is also important to note that this substitution has been driven not only by the sharp decline in the price of ICT assets but also the remarkably increased benefits from investing in ICT assets.
Like the SAP model, the PPF model requires three key assumptions: (i) Constant returns on scale in production; (ii) The existence of a competitive equilibrium, in which the private marginal product of each input factor is equal to its real rental price; and (iii) the absence of externalities to assure that the social marginal product is equal to the private marginal product. Combining assumptions (i) and (ii) implies that the social marginal product of each input equals its real rental price.
14
ThePPFmode la s s ume sa“ Hi c ks -ne ut r a l ”t e c hnol og i c a la ug me nt a t i ona ndfor this paper, it takes the following form: [II-1]
Y= A.X(Kn, Kc, Ks, Kt, L),
where Y is aggregate output, X is a function of capital (K) and labor services (L), and A is the “ Hicks-ne ut r a l ”t e c hnol og i c a la ug me nt a t i onofa ggr e g a t ei nputorTot a lFa c t or Productivity (TFP). The capital services are divided by the asset types of interest: nonICT capital (Kn), hardware (Kc), software (Ks), and telecommunication (Kt).
Suppose that capital Ki has rental price ca ( t hes ubs c r i pt“ a”i ndi c a t e st het y peofc a pi t a l , whi c hi s“ n” f or non-I CT, “ c ”f or ha r dwa r e ,“ s ”f or s of t wa r e ,a nd “ t ”f or telecommunication equipment), and labor L (measured as hours worked) earns the average wage of w. In the nominal output, the share of capital Ka is computed as sKa = ca Ka/YP and the share of labor is sL = wL/YP, where P is the price of output Y. Differentiating equation II-1 with respect to time yields [II-2]
Y= sKn K n + sKc K c + sKs K s + sKt K t + sL L + TFP
where a dot above each variable indicates its growth rate. In particular, TF P (or A) is the growth rate of technology level or Total Factor Productivity (TFP).
Equation II-2] thus indicates that the growth rate of output can be decomposed into the contributions of major types of capital assets, labor, and productivity growth: hardware, sKs K- sKn K n , the contribution of the non-ICT asset, sKc K c – s
software, and sKt K –telecommunication equipment; t sL L, the contribution of the labor input; and
15
TF P , the contribution of TFP growth.
Labor input, in turn, can be decomposed into labor stock (measured in hours worked) and labor quality. Jorgenson, Gollop, and Fraumi (1987) introduce the concept of labor quality index, which is measured as the ratio between labor input and labor stock; that is, [II-3]
Q = L/H
where L, H, and Q denote, respectively, labor input, labor stock (hours worked), and labor quality index.
Equation [II-3] can be rewritten as [II-4]
L=H*Q
That is, labor is the product of labor stock and labor quality index. Transforming equation [II-4] into the log form, then taking the derivative with regard to time for both its sides leads to the equation for computing the growth rate of labor input: [II-5]
L= H+ Q
That is, the growth rate of labor input can be split into the growth rate of hours worked and the growth rate of labor quality. Plugging L in equation [II-5] into equation [II-2] further elaborates the sources of output growth: [II-6]
Y= sKn K n + sKc K c + sKs K s + sKt K t + sL H + sL Q + TFP
II.2. Measuring ICT Investment Flows Efforts to analyze the impacts of ICT on economic growth with a cross-country approach face two major statistical challenges: the availability and comparability of data on 16
investment flows into ICT assets. For developing countries, these data are virtually unavailable, especially from official sources. For some developed countries, the data have been collected for a while but the statistical methods employed for assembling the data differ considerably across countries11. Therefore, it is impossible to expect to have some official comprehensive dataset on ICT investment series for a broad set of countries, including developing economies.
A cross-country analysis of the impacts of ICT on economic growth, however, requires an ICT dataset that meets the following two principal criteria: (i) the dataset covers a large number of countries that well represent the global market; (ii) the dataset is assembled based on a consistent methodology, which allows meaningful cross-country comparisons.
The ICT data provided by the Digital Planet reports published by WITSA (World Information Technology and Services Alliance, a private consortium of 48 ICT industries) appear to best satisfy the above two requirements. The ICT data from WI TSA’ sDi g i t a lPl a ne tr e por t sa r ec ompi l e dba s e dont hema r ke ts t udi e sc onduc t e dby International Data Corporation (IDC), a global market intelligence company specializing in information and telecommunication technology with offices in 50 major ICT-spending economies. The 50 economies together account for about 98 percent of the global ICT market12 and hence well represent the ICT-i nf l ue nc e dwor l d.WI TSA’ sDi g i t a lPl a ne t
11
12
See Colecchia and Schreyer (2001) and Ahmad, Schreyer, and Wolfl (2004) for helpful discussions (WITSA, 2002)
17
reports13 cover ICT data for these 50 economies since 1992. WITSA/IDC data are collected and compiled based on a consistent methodology adopted by IDC for all the economies. WITSA/IDC ICT data have been widely used by researchers for analyzing the contribution of ICT to growth in a group of economies14. Furthermore, the World Bank, IMF, and the UN, in their recent major reports, used WITSA/IDC data to monitor the global ICT development.15
The WITSA data source, however, provides only ICT data related to total spending without a breakdown into investment and consumption. Therefore my paper has to estimate ICT investment flows from the WITSA ICT spending series. The relationship between the ICT investment flows16 and the WITSA ICT spending data for the U.S. is used as the main benchmark for estimating ICT investment flows for other economies. Details of the estimation are presented in Appendix B.
II.3. Capital Stocks
The Perpetual Inventory Method (PIM) The Perpetual Inventory Method (PIM) allows one to calculate the capital stock of an asset type as the accumulated sum of its past real investment flows, weighted to reflect
13
WI TSA’ sDi g i t a lPl a n e tr e por t swe r epu bl i s h e di n1998,2 00 0,a n d200 2.
14
Schreyer (2000), Daveri (2002), and Lee and Khatri (2003) are among the most notable examples.
15
For examples, OECD (2000, 2001, 2002), World Bank (2000, 2001), IMF(2001), UN(2001).
16
This data is compiled by BEA (Bureau of Economic Analysis, website www.BEA.gov)
18
the loss of productive efficiency of the installed asset over time. The stock of capital a at period t can be estimated as Sa,t = =0, (1- a) Ia, t-
[II-7]
where Ia,t-is the flow of investment in capital a at time t- . Supposing that the average service life of the asset is m years, we can assume that this asset will be definitely discarded after m years in service. That is, investments in asset a in year t-m or earlier are eradicated from the capital stock of asset a in year t. Therefore, Equation [II-7] can be simplified into [II-8]
Sa,t = =0, m-1 a ,Ia, t-
where a =(1- a) is the rate of efficiency remaining at year t of asset a invested in year
t-for =1, 2,.., m. The value of a 0 needs to be adjusted for the earlier years to take into account the assumption that the asset is definitely discarded when its age reaches m years. For computational convenience, the remaining efficiency of the asset when it is m discarded after its mth year in service, (1- should be added to the efficiency of the a) m asset in its first year in use, that is, a, 0 = 1 + (1- a) .
Now we start to compute the capital stock of the four types of asset, including three types of ICT asset–hardware, software, and telecommunication; and the aggregate non-ICT asset. Following widely-used practice, originating from the work of Fraumeni (1997), we assume the service lives and depreciation rates of the above assets17 as follows: The service life is 7 years for hardware, 5 years for software, 11 years for telecommunication equipment, and 30 years for aggregate asset.
17
These assumptions are widely used in studies on the contribution of ICT to economic growth.
19
The geometric depreciation rate is 31.5 percent for hardware and software, 11 percent for telecommunication equipment, and 7.5 percent for the aggregate asset.
Consider asset i with an average service life of m years. Measuring the capital stocks for this asset at year 1990 requires the investment flow into this asset for m years, from 1990-m-1 to 1990. Specifically, one needs to estimate investment flow starting at 1984 for hardware (m=7), 1986 (m= 5) for software, 1980 for telecommunication equipment (m=11), and 1961 for the aggregate asset (m=30).
Investment Deflators The investment flows {Ia,t-} in equation II-8 are measured in their real values. For the aggregate asset, this investment flow is the GFCF flow in its constant value. For an ICT asset type (hardware, software, and telecommunication equipment), its nominal investment flows estimated from the previous section need to be deflated to their real values. The rapid progress of ICT technology has caused the real prices of ICT goods, especially computer hardware, to decline remarkably in the past few decades. For example, the real price of computer hardware in 2000 was nearly 100 times lower than in 1975 (BEA, 2004). However, constructing price indices for ICT assets remains a formidable challenge for most countries. The US was the pioneer in this effort. It has used the hedonic approach since 1985 and this method has gained wide acceptance. The hedonic approach estimates price changes for an ICT asset such as computer equipment, for which direct comparison is impossible, by estimating an implicit price on the basis of the coefficients of particular features of a product (such as speed, memory size, and other technical features). The hedonic price deflators, therefore, substantially improve the
20
accounting for rapid quality changes in ICT assets. Because the features and prices of ICT products do not vastly differ from country to country18, it is reasonable to assume that the prices of ICT products behave in a consistent way across economies. Therefore, I assume that the U.S. ICT hedonic price indices are applicable for other economies. Using the hedonic deflator for each of the three ICT assets, I can deflate its nominal investment flows to real values. Formula [II-8], therefore, estimates the real values of the capital stocks of four asset types: the gross asset (GA), computer hardware (HW), computer software (SW), and telecommunication equipment (TEL).
From the real value of the capital stock for an asset type (GA, HW, SW, or TEL), one can compute its current value in a given year by multiplying its real value by the current price index of that asset type. The current value of the ICT asset is the sum of the current values of the capital stocks of its three main components, HW, SW, and TEL. The current value of the Non-ICT asset’ sc a p i t a ls t oc kis computed as the current value of the gross asset’ sc a pi t a ls t oc kmi nusthat of the ICT asset.
II.4. Capital Services Capital Services Methodology When a firm hires a worker, the wage paid to this worker is a measure of his/her labor service and is added to GDP. In a similar way, when a firm purchases or rents a piece of equipment, the capital services rendered by this equipment, then, should be added to
18
ICT products are supplied by a small number of global vendors; for example, the five largest PC vendors (Dell, Hewlett-Packard, IBM, Fujitsu/Siemens, and Toshiba) together account for nearly 41 percent of the worldwide PC market, which was about 169 million in 2003 (Source: Gartner Dataquest, January 2004).
21
GDP. The capital services methodology is based on the economic theory of production and has been comprehensively described in Jorgenson and Stiroh (2000) and OECD (2001a, 2001b).
Capital services are defined as the productive inputs, per period, that flow to production from a capital asset. The value of capital services rendered by an asset is the quantity of services provided by the asset multiplied by the price of those services. The framework for computing capital services is as follows19: The quantity of capital services of asset a in period t is proportional to the average of the stocks of this asset available at the end of periods t and t-1; that is [II-9]
Ka,t=qa*( Sat + Sa,t-1)/2
where Ka,t and Sa,t are, respectively, quantity of capital services and capital stock measured for asset a at time t, and qa denotes the constant of proportionality, which can be set equal to unity without losing generality.
The capital service price is the unit cost for the use of a capital asset for one period; that is, the price for employing or obtaining one unit of capital services. Thes e r vi c epr i c ei sa l s or e f e r r e dt oa st he“ r e nt a lpr i c e ”ofac a pi t a lg oodort he “ us e rc os tofc a pi t a l ” .Toi nf e rt hec a pi t a ls e r vi c epr i c eofac a pi t a lg ood,o ne assumes that a typical investor in period t-1 makes her decision by considering only two alternatives available to her: earning a nominal interest rate of return, rt, on her money, Pa, t-1, in the money market to get (1+rt)Pa, t-1, or buying a unit of capital, collecting a rental fee, ca,t, and then selling the depreciated asset in period 19
See Jorgenson and Stiroh (2000) for more details.
22
t for (1- a)Pa, t. Under the equilibrium condition, the investor is indifferent between the two alternatives; that is [II-10]
(1+rt)Pa, t-1=ca,t + (1- a)Pa, t
where rt is the nominal interest rate at time t, Pa, and ca, are the prices of capital stock and capital services, respectively, of asset a at time , and a is the geometric depreciation rate for asset a.
Rearranging equation [II-10] provides a formula for computing rental price ca,t [II-11]
ca,t = rtPa, t-1- a, t Pa, t-1+ aPa,t
where a, t = (Pa, t - Pa,t-1)/Pa,t-1 is the investment deflator or capital gain/loss for owning asset a over period t. Normalizing the price of output to 1, the income share sKa of capital a in output at time t is computed as
[II-12] sKa, t = Ka, t*ca, t /Yt = Ka, t* (rtPa, t-1- a, t Pa, t-1+ aPa,t) /Yt
The interest rate rt can be computed from the following equation: [II-13]
sK =shw+ssw+stel+snict
which shows that the income share of aggregate capital (sK) equals the sum of the income shares of the four types of capital under discussion: hardware (shw), software (ssw), telecom equipment (stel), and non-ICT (snict).
23
Plugging sK a, t from equation [II-12] into equation [II-13] leads to the equation for computing the interest rate rt as follows: [II-14] rt={sK Yt+
[Ka, t a, t Pa, t-1] - [Ka, t aPa,t]} / [Ka, t* Pa, t-1]
where the summation symbol in front of each component in [ ] indicates the sum of this component across the four types of capital, a{n, c, s, t}, and the income share of capital is assumed to be constant over the period under consideration.
User Cost of Capital Computing capital services requires not only capital service quantity, which is deduced from capital stock, but also capital service price or the user cost of capital. There are three important elements in equation [II-11], which is used for computing the user cost of capital asset a at time t: interest rate rt, the asset’ sde pr e c i a t i onr a t e , a, t, and its capital gain, a, t. The depreciation rate, a, t , for capital goods a at time t, as discussed in section II.3, is assumed to be constant at 31.5 percent for computer hardware and computer software, and 11 percent for business telecommunication equipment. The capital gain a, t can be computed straightforwardly from its definition, a, t = (Pa, t - Pa, t-1)/Pa, t-1, where Pa, t and Pa, t-1 are the price indices of asset a at times t and t-1, respectively. The interest rate, rt, which is applied to all types of capital goods, is computed from Equation [II-14]. In this equation, only the income share of aggregate capital, sK, needs to be estimated.
24
For estimating the income share of capital, there are three typical approaches. The first approach is to compute the income share of capital for an economy from its national account. This approach, however, appears suitable only for the OECD countries, for which national account statistics are standardized and well recorded. The second approach relies on econometric regressions. In this approach, researchers regress output Y on capital stock K, employment L, and other possible factors such as land, human capital index, and time, using the time series data for each country under study (for example, Senhadji, 2000; Chow, 1997, 2000). This approach can be used for a broad set of countries but its accuracy remains questionable. The third approach assumes that capital shares for all economies take the typical value observed in most OECD countries, which is around 0.30 to 0.40; the income share of labor, therefore, is around 0.60 or 0.70. This approach has been widely used in studies on economic growth, which include Woo(1997), Perkins(2000), Schreyer(2000), and Bosworth and Collins(2003). In this study, I assume that the capital share, sK, equals 0.35 across economies (and therefore, their labor share, sL, is 0.65). The interest rate, therefore, can be computed from Equation [II-14]. In addition, I also assume that the expected interest rate (in the US$ terms) earned from investing in each individual economy is stable for the entire period 1990-2000. II.5. Hours Worked and Labor Quality Measuring Hours Worked Data on employment and workforce can be found in the database of the International Labor Organization (ILO)20 and its yearly-publ i s he d“ Ye a r bookofLa borSt a t i s t i c s ” .The data on hours worked are available for most economies. However, the usability of this 20
Se et h eI LO’ swe bs i t e( www.ILO.org)
25
data in terms of their time series frequency, breakdown, and representation of overall economy varies largely among the economies. For example, yearly data on hours worked per week for each category of economic activity are available for most OECD countries; but for China, the data are intermittent and collected from surveys confined to a few cities or state-owned sector. Therefore, it requires strong assumptions in compiling labor data that are meaningful for cross-country comparisons.
To estimate the hours worked for individual economies, I follow the following strategy: I use hours worked for economies, for which the data is available; I assume that the number of hours worked is constant for those economies, where the data on hours worked are not sufficient or unavailable. This assumption is reasonable because observations over 1990-2000 for most economies with well-representative data reveal that the number of hours worked per week changed only slightly during the period. If hours worked data are not available or not well-representative, I use data on fulltime equivalents, assuming a full-time employee in each group of countries works the same number of hours per year. If full-time equivalents data are not available, I use employment data, assuming the average worker in each group of countries work the same number of hours per year.
Labor Input and Labor Quality Conventionally, labor input is measured simply as hours worked. However, the true measure of labor input should capture not only hours worked but also changes in labor
26
quality. The early works attempting to measure labor input with attention to labor quality are Griliches (1960), Denison (1962), and Jorgenson and Griliches (1967). All of these studies construct the labor quality index of labor input by aggregating hours for workers with different characteristics, using hourly wage rates as weights. A comprehensive methodology for estimating labor input is introduced by Jorgenson, Gollop, and Fraumeni (1987). This methodology, in principal, parallels capital services. That is, labor input is measured as the aggregate of labor service flows rendered by all worker categories, which are classified by sex, age, educational attainment, employment class, and industry. For worker category j at time t, labor service flow Lj,t is proportional to hours worked Hj,t: [II-15]
Lj,t = qj Hj,t
where qj is the constant proportionality for worker category j. Without losing generality, qj can be set to unity. At time t, aggregate labor input Lt is defined as the share-weighted aggregate of the labor inputs of all worker categories21: [II-16]
Lt = v j ,t Lj,t
where weights v j ,t are average value shares of labor income in period t, computed using the Tornqvist average method: [II-17]
v j ,t =
W j ,t * L j , t 1 { + 2 (W j ,t * L j ,t ) j
W j ,t 1 * L j ,t 1
(W
j ,t 1
* L j ,t 1 )
}
j
Wj,t in equation [II-17] denotes the wage of worker category j at time t.
21
The number of worker categories classified by Jorgenson, Ho, and Stiroh (2002) for each of the 44 industries is 168 including all combinations of sex (2), age (7 groups), educational attainment (6 classes), employment class (two types).
27
Furthermore, at time t, the aggregate index of labor quality Qt is defined as the ratio between the aggregate labor input Lt and the un-weighted sum of hours worked Ht: [II-18]
Qt = Lt / Ht
Equation [II-18] can be rewritten as [II-19]
Lt = Ht * Qt
That is, the aggregate input can be computed as the product of total hours worked and the labor quality index. To construct the labor quality index, I hypothesize that it is a function of educational attainment, institutional quality, and standard of living. The aggregate labor quality index Qt is estimated as the following: (i) Start with the labor quality index constructed by Jorgenson (2003) for the G7 economies for 1990, 1995, and 2000. (ii) Project the aggregate labor quality index for other 43 economies for 1990, 1995, and 2000, using the following OLS model: Qt = 0 + 1 Educationt + 2 Institution1 + 3 Institution2 + 4 Income1990 + 5 year where Educationt is the educational attainment in year t of the population aged 25 or over from the Barro-Lee education dataset; Institution1 and Institution2 are the institutional qua l i t yi ndi c e sf r om t heWor l dBa nk’ ss e tofg ove r na nc eindicators measured for 1996 ( I ns t i t ui on1=“ Rul eofLa w”a ndI ns t i t ui on2= “ Re g ul a t or yQua l i t y ” ) ;I nc ome 1990i s GDP (measured in PPP) per capita in 1990, which is a proxy for the standard of living; and Year is the dummy for year; the model has a high predictive power (R2 = 0.973) and all the explanatory variables are significant. The model implies that for each country, educational attainment is the key factor behind the change in labor quality index over
28
time because the variables Institution1, Institution2, and Income1990 are constant throughout the period 1990-2000.
II.6.The Sources of Output Growth Equation II-6 indicates that the sources of output growth can be grouped into three main channels: Contribution of capital input, which consists of that of non-ICT (sKn K n ) and ICT capital. The contribution of ICT capital comes from its three main components: hardware (sKc K ), software (sKs Ks ), and telecom (sKt K ). c t Contribution of labor input, which includes that hours worked (sL H) and labor quality (sL Q). Contribution of TFP growth: TF P
The sources of output growth for the 50 economies over the two periods 1990-1995 and 1995-2000 are reported in Tables 1. The output growth rate and its sources for a group (as well as for the whole sample) are share-weighted averages. The weight of each individual economy in a group (or the whole sample) for a period is the average of its shares in GDP (measured in PPP$ in 1999) at the beginning and the end of the period.
29
Table 1: Sources of Output Growth, 1990-1995 vs. 1995-2000
Country
GDP Growth
G7 Economies Canada France Germany Italy Japan UK US Weighted Avg.
1.72 1.06 1.59 1.26 1.39 1.59 2.36 1.84
Period 1990-1995 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
GDP Growth
Period 1995-2000 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
0.26 0.16 0.16 0.11 0.17 0.22 0.39 0.27
0.53 0.50 0.50 0.34 0.97 0.40 0.49 0.57
0.66 0.24 -0.01 0.54 0.51 0.00 1.00 0.62
0.08 0.29 0.28 0.29 0.22 0.64 0.18 0.46
0.18 -0.13 0.66 -0.02 -0.47 0.33 0.30 -0.08
3.59 2.45 1.73 1.87 1.44 2.77 4.10 2.99
0.65 0.33 0.33 0.22 0.36 0.49 0.78 0.56
0.72 0.38 0.32 0.40 0.66 0.53 0.92 0.70
1.24 0.78 -0.03 0.38 0.04 0.91 1.36 0.83
0.12 0.22 0.32 0.28 0.15 0.18 0.22 0.32
0.86 0.75 0.78 0.60 0.24 0.65 0.82 0.58
Non-G7 Industrialized Economies Australia 3.33 0.33 Austria 2.03 0.12 1.48 0.19 Denmark 1.95 0.14 Finland -0.68 0.14 Greece 1.24 0.08 Ireland 4.55 0.23 Israel 6.36 0.31 Netherlands 2.08 0.25 New Zealand 3.00 0.33 Norway 3.62 0.15 Portugal 1.74 0.14 Spain 1.33 0.10 Sweden 0.59 0.18 Switzerland -0.08 0.21 Weighted Avg. 1.90 0.18
0.45 0.70 0.60 0.22 -0.15 0.17 0.53 1.46 0.41 0.11 0.05 0.70 0.72 0.15 0.49 0.49
0.90 0.59 0.22 0.41 -1.77 0.46 1.53 3.95 0.73 1.54 0.78 0.05 -0.40 -0.67 0.41 0.34
0.67 0.33 0.51 0.27 0.21 0.18 0.39 0.35 0.36 0.46 0.42 0.65 0.31 0.54 0.14 0.49
0.98 0.30 -0.04 0.90 0.90 0.35 1.87 0.29 0.34 0.55 2.22 0.20 0.60 0.39 -1.32 0.39
3.92 2.44 2.72 2.63 4.97 3.26 9.22 3.74 3.45 2.27 3.01 3.53 3.69 2.84 1.76 3.44
0.60 0.24 0.37 0.38 0.48 0.20 0.50 0.50 0.54 0.56 0.39 0.39 0.22 0.62 0.45 0.41
0.78 0.65 0.62 0.62 0.02 0.41 1.29 1.26 0.52 0.53 0.45 0.96 0.91 0.17 0.33 0.67
1.28 0.10 0.63 0.24 0.99 0.19 3.03 2.14 0.93 0.98 0.87 0.87 2.08 0.65 0.46 1.13
0.20 0.25 0.19 0.32 0.36 0.30 0.53 0.29 0.20 0.26 0.30 0.45 0.30 0.24 0.23 0.47
1.06 1.19 0.90 1.06 3.11 2.15 3.86 -0.45 1.26 -0.06 0.99 0.87 0.17 1.15 0.29 0.77
30
Table 1: (Continued)
Country
GDP Growth
Developing Asia China Hongkong India Indonesia Korea Malaysia Philippines Singapore Taiwan Thailand Vietnam Weighted Avg. Latin America Argentina Brazil Chile Colombia Mexico Venezuela Weighted Avg.
Period 1990-1995 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
GDP Growth
Period 1995-2000 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
11.38 5.21 5.05 7.57 7.19 9.05 2.14 8.72 6.89 8.30 7.91 8.22
0.17 0.28 0.08 0.09 0.26 0.26 0.10 0.33 0.20 0.10 0.15 0.15
2.15 1.54 1.36 1.69 2.28 2.33 0.68 1.63 1.97 2.32 1.24 1.85
0.98 1.15 1.64 2.17 1.91 2.65 1.99 2.00 1.39 1.61 1.54 1.45
1.28 0.32 0.97 0.99 0.70 0.81 0.36 0.73 0.70 0.78 1.07 0.93
6.81 1.92 1.00 2.63 2.04 2.99 -0.99 4.03 2.64 3.50 3.92 3.84
7.92 3.35 5.57 0.68 4.66 4.57 3.49 6.15 5.61 0.24 6.48 5.89
0.44 0.51 0.18 0.08 0.46 0.48 0.19 0.69 0.42 0.14 0.41 0.34
2.45 1.51 1.70 1.71 1.78 1.92 0.82 1.85 2.17 0.85 2.25 2.01
0.72 1.53 1.65 1.57 1.03 2.08 1.65 1.40 0.66 0.64 1.27 1.09
1.04 0.10 0.88 0.35 0.46 0.35 0.40 0.44 0.48 0.17 0.97 0.87
3.26 -0.30 1.17 -3.02 0.92 -0.25 0.44 1.77 1.87 -1.56 1.58 1.58
6.35 3.09 8.33 4.43 1.52 3.39 3.46
0.10 0.10 0.22 0.14 0.14 0.11 0.12
0.29 0.28 1.64 0.95 0.70 0.14 0.50
-0.48 1.14 2.02 2.35 1.91 2.32 1.32
0.42 0.49 0.59 0.54 0.26 0.15 0.85
6.02 1.08 3.87 0.46 -1.50 0.66 0.67
2.59 2.25 4.44 0.92 5.36 0.61 3.01
0.18 0.31 0.37 0.41 0.18 0.23 0.27
0.33 0.40 1.95 0.48 0.77 0.03 0.54
2.71 0.81 0.66 0.06 2.41 1.53 1.47
0.30 0.41 0.36 0.48 0.43 0.03 0.79
-0.93 0.32 1.11 -0.50 1.57 -1.21 -0.05
31
Country
GDP Growth
Eastern Europe Bulgaria Czech Hungary Poland Romania Russia Slovakia Slovenia Weighted Avg. Other Economies Egypt S. Africa Turkey Weighted Avg.
All Sample
Period 1990-1995 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
GDP Growth
Period 1995-2000 Sources of Growth (% points per annum) Capital Labor ICT Non-ICT Hours Quality TFP
-2.65 -0.97 -2.39 2.17 -2.15 -9.52 -3.03 -0.59 -6.38
0.13 0.22 0.22 0.11 0.04 0.07 0.20 0.15 0.10
-0.74 -0.32 0.03 0.11 -0.84 -0.60 -0.31 -0.89 -0.49
-2.92 0.01 -1.16 -0.68 -0.71 -1.45 -0.37 -0.24 -1.20
-0.15 0.14 0.08 0.32 -0.07 -0.83 -0.03 0.12 0.82
1.02 -1.01 -1.56 2.32 -0.57 -6.71 -2.52 0.27 -5.60
-1.31 0.95 3.93 5.01 -1.60 1.13 4.02 4.25 1.72
0.23 0.43 0.42 0.34 0.10 0.10 0.37 0.32 0.19
-0.84 0.19 0.35 1.12 -0.32 -1.77 0.21 0.14 -0.88
-0.35 -0.41 0.60 -0.09 -0.31 -0.19 -0.32 0.17 -0.16
-0.07 0.07 0.51 0.50 0.02 -0.06 0.35 0.54 0.84
-0.28 0.68 2.06 3.14 -1.09 3.05 3.41 3.09 1.72
3.34 0.86 3.14 2.26
0.10 0.20 0.08 0.13
0.17 -0.06 1.71 0.70
1.59 2.36 2.24 2.17
0.73 0.22 0.70 1.05
0.75 -1.86 -1.58 -1.79
5.30 2.42 3.70 3.54
0.19 0.42 0.22 0.29
0.57 0.23 1.69 0.91
2.84 0.17 1.47 1.25
1.02 -0.51 0.45 0.75
0.68 2.12 -0.12 0.35
2.68
0.21
0.74
0.73
0.64
0.37
3.70
0.44
0.93
0.93
0.55
0.85
(50 Economies)
32
Figure 1: Sources of Output Growth by Group of Economies % 9 8 7 3.84
6
1.58
5 4
0.93
0.87 1.05
3 2 1 0
0.77 1.45
0.58
0.47
0.32 0.46
0.83
0.62
0.70
0.57 0.27 -0.08
0.56
1.09
0.39
1.72
0.34
0.50 0.12
1.85 0.67
0.41
0.15
2.17
1.25
1.47 1.32 0.54 0.27 -0.05
-1
0.55
0.37
0.85
2.01
0.85
0.75
0.79
1.13
0.49 0.34 0.49 0.18
0.35
0.67
0.64
0.93
0.73 0.84
0.82
0.19
0.10 -0.49
0.91 0.70
0.93 0.74
0.29
0.21
0.44
1995-2000
1990-1995
1995-2000
0.13
-0.88
-1.20
-0.16
-1.79
1995-2000
1990-1995
-2 -3 -4 -5.60
-5 -6 -7 -8 -9 1990-1995
1995-2000 G7
1990-1995
1995-2000
1990-1995
Non-G7
1995-2000
Developing Asia
ICT
Non-ICT
1990-1995
1995-2000
Latin America
Labor Hours
1990-1995
Eastern Europe
Labor Quality
Other-3
All Sample
TFP
33
III. ICT as a Source of Economic Growth22 III.1. Global Economic Growth The sample of 50 economies included in this study represents approximately 94% of the global GDP in 200023. The growth dynamics observed for the sample, therefore, well characterize that of the global economy.
24 As shown in Table 1 (the bottom panel), the out putg r owt hoft he“ g l oba le c onomy ”
increased from 2.68% during 1990-1995 to 3.70% during 1995-2000. The same trend was observed for several groups: G7 (from 1.84% to 2.99%), Non-G7 (from 1.90 to 3.44), and Other-3 (from 2.26% to 3.54%); in particular, Eastern Europe turned around its deep recession (GDP fell by 6.38%) during the first period to a moderate growth in the second (1.72%). However, the output growth, declined from 8.22% during 1990-1995 to 5.89* during 1995-2000 for Developing Asia; and from 3.46% to 3.01% for Latin America over the two periods; the main reason for the growth slowdown in these two groups was the financial crisis that initially erupted in East Asia in 1997.
Table 1 and Figure 1 reveal the main drivers of economic growth by group over the two periods, 1990-1995 and 1995-2000.Fort he“ g l oba le c onomy ”a ndt het hr e eg r oups ,G7, Non-G7, and Other-3, all the sources of growth except labor quality significantly enhance
22
Three terms, economic growth, output growth, and GDP growth are used interchangeably in this paper.
23
Measured in US$, using current market exchange rate (computed from WBI, 2002).
24
Th et e r m“ gl oba le c on omy ”i nt h i ss e c t i on ,h e r e a f t e r ,r e f e r st ot h es a mpl eoft h e50e c on omi e s .
34
their contribution to output growth over the two periods. For Eastern Europe, the main drivers of the turnaround in output growth from the first period to the second were TFP and ICT. The slowdown in output growth experienced by Developing Asia and Latin America was mainly caused by the fall in the contribution of TFP. For these two groups, however, there was a powerful surge in ICT contribution from the first to the second period. Thus, for all groups, ICT contribution played an increasingly important role as a source of growth. The following sections will examine this aspect in details.
III.2.Magnitude of ICT Contribution to Output Growth a) A Powerful Surge in the Second Period The contribution of ICT to output growth was positive for all economies in the two periods 1990-1995 and 1995-2000. Figure 2 shows a decisive shift of the distribution of economies by the magnitude of ICT contribution to output growth, with the average surging from 0.17 in the first period to 0.37 percentage points in the second. For nearly two thirds of economies, the magnitude of ICT contribution at least doubled over the two periods. At the group level, the weighted average of the ICT contribution to output growth increased from 0.27 percentage points during 1990-1995 to 0.56 during 19952000 for G7; from 0.18 to 0.41 for Non-G7; from 0.15 to 0.34 for Developing Asia; from 0.12 to 0.27 for Latin America; from 0.10 to 0.19 for Eastern Europe; from 0.13 to 0.29 for Other-3;a ndf r om 0. 21t o0. 29f ort he“ Gl oba lEc onomy ” (Table 1).
35
Figure 2: Distribution of the 50 Economies by the Magnitude of ICT Contribution to Output Growth, 1995-2000 vs. 1990-1995 Period 1990-95
Period 1995-00
density, % of Countries
10
5
0 .05
.1
.17 .2
.3 .37 .4 .5 .6 ICT Contb. to Output Grow th, ppa
.7
.8
b) ICT vs. Non-ICT capital Non-ICT capital remains a major source of output growth for all groups except Eastern Europe. However, unlike ICT capital, the contribution of Non-ICT capital did not show a solid increasing trend across economies; in fact, it decreased notably in Germany, France, Japan, Korea, Malaysia, Thailand, Colombia, and Venezuela (Table 1). ICT and Non-ICT make up the contribution to growth of capital input, of which the share of ICT rose considerably from 1990-1995 to 1995-2000 across groups, from 32% to 45% for G7; 27% to 38% for Non-G7; 7% to 15% for Developing Asia; 19% to 33% for Latin America; 16% to 24% for Other-3; and from 22% to 32% for the “ g l oba le c onomy ”(Figure 3). For the Eastern Europe group, while the contribution of Non-ICT capital was negative and worsening from the first to the second period, the contribution of ICT capital was positive and substantially increasing. 36
Figure 3: Share in Capital Input Contribution to Growth by Group: ICT vs. Non-ICT 100%
55 68
62
67
73 93
45 32
85
81
15
19
38
33
27 7
0%
84
For Eastern Europe, the contributions of capital input to output growth were negative in both periods; hence, the label values are not the shares but the actual contribution of ICT and Non-ICT capital (in percentage points)
0.10
0.19
-0.49
-0.88
16
68
76
78
24
22
32
-100% 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 G7
Non-G7
Developing Asia
Latin America
ICT
Eastern Europe
Other-3
All Sample
Non-ICT
c) The Global Picture of ICT Contribution to Growth The global picture of ICT contribution to output growth is a mix of static and dynamic features: while a number of economies stayed at the top or the bottom by rank in the magnitude of ICT contribution to output growth, many economies appeared to have acted more aggressively in the second period in reaping the benefit of the ICT revolution for growth. The static feature: Table 2A shows that six industrialized nations–the US and Canada (G7); Australia, New Zealand, Netherlands, and Israel (Non-G7)–and two developing economies: Singapore and Hong Kong (Asia-11) remained in the top-10 in both periods by rank in magnitude of ICT contribution to growth; at the same time, six nations–Indonesia, Romania, Russia, India, Argentina, Egypt, and Greece–stayed in the bottom-10 for
37
both periods. Table 2B sharpens this picture by focusing only on the contribution of IT;25 on this ranking, seven economies stayed in the Top-10 in both periods: the US, Canada, Netherlands, Australia, the UK, New Zealand, and Singapore. Figure 4 further illustrates the static feature of the ICT contribution to output growth. The magnitude of ICT contribution to output growth for an economy in the first period highly predicts that in the second period (the 50 economies lie along the upward diagonal line in the graph).
The dynamic feature: The Tables 2A and Figure 4 also reveal the dynamic feature of the global picture of ICT contribution to growth. Malaysia and Korea fell from the Top-10 in the first period to the Middle-30 in the second, while Ireland and Sweden moved from the Middle-30 to the Top-10; Philippines, Mexico, and Thailand descended from the Middel-30 in the first period to the Bottom-10 in the second, while Brazil, Spain, and Turkey went up from the Bottom-10 to the Middle-30 (Table 2A). In Figure 4, a number of economies significantly deviate from the diagonal, either above or below. The economies lying below the diagonal (such as Canada, Sweden, Finland, China, Colombia, and Brazil) obviously outperformed the average player in reaping the benefits of ICT for growth in the second period, while the economies above the diagonal (such as New Zealand, Israel, Mexico, and Indonesia) fell behind the average player in this regard. 25
In many developing countries, investment in traditional fixed line telephone made up a large share of ICT contribution to growth; therefore, excluding telecom from ICT provides a sharper picture of the c on t r i bu t i onoft h e“ n e we c onomy ”one c on omi cg r owt h .
38
Table 2.A: The Matrix of ICT Contribution to Output Growth: 1995-2000 vs. 1990-1995 0.26 0.26
UK Finland Switzerland China Czech Hungary Taiwan S. Africa Vietnam Colombia Norway Portugal Denmark Slovakia Chile Belgium Japan Poland Germany France Slovenia 0.19 Austria 0.18 Venezuela 0.14 Bulgaria Italy
0.22 0.14 0.21 0.17 0.22 0.22 0.20 0.20 0.15 0.14 0.15 0.14 0.14 0.20 0.22 0.19 0.17 0.11 0.16 0.16 0.15 0.12 0.11 0.13 0.11
0.48 US 0.46 Singapore Canada Australia New Zealand Netherlands Hong Kong Israel 0.49 0.48 Sweden 0.45 Ireland 0.44 0.43 0.42 0.42 0.42 0.41 0.41 0.39 0.39 0.38 0.37 0.37 0.37 0.36 0.34 0.33 0.33 0.32 0.24 0.23 0.23 0.22
0.20 Brazil 0.19 Spain 0.18 Turkey 0.18 0.10 0.10 0.08
0.10 0.10 0.08
0.31 0.22 0.22
3 0 d d le M i
P E R I
O D
1 9 9 0 - 1 9 9 5
T o p
1 0
Malaysia Korea
B o t t o m 1 0
Philippines Mexico Thailand
0.10 0.14 0.10
Greece 0.08 Egypt 0.10 Argentina 0.10 India 0.08 Russia 0.07 Romania 0.04 Indonesia 0.09 Bottom 10
PERIOD
Middle 30 1995 - 2000
0.39 0.33 0.26 0.33 0.33 0.25 0.28 0.31
0.78 0.69 0.65 0.60 0.56 0.54 0.51 0.50
0.18 0.23
0.62 0.50
Top 10
Notes: 1) Numbers to the right of each country name are ICT contribution to output growth (p.p.a) in the two periods 1990-95 and 1995-00, respectively; 2) In each cell, countries are sorted in decreasing order of contribution of ICT to output growth during 1995-00
39
Israel Korea Czech
0.21 0.18 0.18
0.37 US 0.34 Canada 0.33 Singapore Netherlands Australia UK New Zealand
Ireland Malaysia Norway S. Africa Denmark Belgium Germany France Hungary Taiwan China Slovakia Hong Kong Slovenia Portugal Poland Japan Austria Chile Brazil Vietnam 0.12 Colombia 0.10 Bulgaria 0.08 Italy
0.13 0.17 0.12 0.14 0.11 0.15 0.13 0.12 0.17 0.09 0.09 0.17 0.15 0.13 0.07 0.09 0.13 0.09 0.11 0.06 0.09 0.08 0.11 0.07
0.35 0.35 Sweden 0.35 Finland 0.34 Switzerland 0.33 0.30 0.29 0.28 0.28 0.27 0.26 0.26 0.26 0.25 0.24 0.23 0.23 0.21 0.20 0.19 0.18 0.17 0.17 0.15
3 0 M i d d le
P E R I O D
1 9 9 0 - 1 9 9 5
T o p
1 0
Table 2.B: The Matrix of IT Contribution to Output Growth: 1995-2000 vs. 1990-1995
B o t t o m 1 0
Mexico Thailand Russia
0.06 0.06 0.06
Turkey 0.03 Egypt 0.05 Argentina 0.04 Greece 0.03 India 0.04 Romania 0.03 Indonesia 0.05 Bottom 10
0.13 Spain 0.06 0.12 Venezuela 0.05 0.12 Philippines 0.06 0.10 0.09 0.07 0.05 Middle 30 PERIOD 1995
0.31 0.20 0.23 0.20 0.24 0.17 0.24
0.69 0.56 0.53 0.46 0.45 0.41 0.38
0.14 0.10 0.17
0.58 0.41 0.40
0.17 0.15 0.14
Top 10 -
2000
Notes: 1) IT includes only computer hardware and software 2) Numbers to the right of each country is IT contributions to output growth (p.p.a) over 90-95 and 95-00 in two periods 1990-95 and 1995-00, respectively; 3) In each cell, countries are sorted in decreasing order of computer contribution to output growth over 1995-00
40
Figure 4: ICT Contribution to Output Growth: 1995-2000 vs. 1990-1995
.4
US
N ew Zealand Aus tralia
Singapore
Is rael
.3 H ongk ong Malay s ia Kor ea
C anada
1990-95, p.p.a
N etherlands
C hile Slov ak ia
.2
C z ec h H ungary Sw itz er lan d S. Afr ican a Taiw
Ireland UK
Belgium Sw eden
.17 Mex ic o Bulgaria Aus tria Italy uela Venez Thailand Philippines Spain Egy pt Arg entina Indones ia India Greec Turek ey R us s ia
.1
C hina J apan Germany Fra nc e Slov enia N orw ay Vietnam D enmark Por tugal Finland C olombia Poland Bra z il
R omania
0 0
.1
.15
.3
.37 .45 1995- 00, p.p.a
.6
.75
.8
ICT Contribution to Output Grow th, 1995-00 vs . 1990-95
41
III.2.The Share of ICT Contribution in the Overall Rate of Output Growth
Not only the magnitude but also the share in overall output growth of ICT contribution rose significantly from 1990-1995 to 1995-2000. Figure 5 shows this trend at the group and world levels. The share of ICT contribution in overall output growth jumped from 14.5% to 18.8% for G7, 9.8% to 12.0% for Non-G7, 1.8% to 5.8% for Developing Asia, 3.4% to 8.8% for Latin America, 5.8% to 8.1% for Other-3,a ndf r om 7. 7% t o12. 0% f ort h e“ g l oba l e c onomy ” ;f ort heEa s t e r nEur opeg r oup,t hes ha r eofI CTc ont r i but i oni ni t sove r a l l output growth was 11.1% for the second period when its output growth turned positive.
Figure 6, illustrates this trend for individual economies with output growth exceeding 1% in both periods.26 The figure shows that most economies were positioned well below the diagonal; that is, the share of ICT in overall output growth rose substantially from the period 1990-1995 to the period 1995-2000. In particular, ICT accounted for a major share in output growth for Japan, New Zealand, and Germany in the second half of the decade.
26
The economies with output growth below 1% are excluded to avoid overstating the share of ICT contribution in growth.
42
Figure 5: The Share of Input Contribution in Overall Ouput Growth 100%
81.2
85.5
18.8
14.5
90.2
88.0
9.8
12.0
0%
98.2
94.2
96.6
1.8
5.8
3.4
88.9
91.2
11.1
8.8
0.10
For Eastern Europe, the output growth was negative in the first period; hence, the label values are not the shares but the actual contribution of ICT and other sources to output growth (in percentage points)
94.2
91.9
92.3
88.0
5.8
8.1
7.7
12.0
-6.47
-100% 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 1990-1995 1995-2000 G7
Non-G7
Developing Asia
Latin America
The Share of ICT Contribution
Eastern Europe
Other-3
All Sample
The Share of Others
Figure 6: ICT Share in Output Growth: 1995-2000 vs. 1990-1995 (Only economies with average output growth exceeding 1 percent in both periods are included) 20
Diagonal US Canada
France
15
UK
1990-95, %
Belgium Netherlands
Japan New Zealand
10
Australia
Mexico
Germany
Italy Portugal Spain
Denmark
Greece Austria Hongkong
Ireland Poland Philippines
5
Israel KoreaSingapore
Egypt India
Taiwan Chile
Malaysia
Norway Brazil
Turkey Vietnam China Argentina
5
10
15 1995-00, %
20
30
Share of ICT Contb. in Output Growth, 1995-00 vs. 1990-95
43
IV. Determinants of ICT Contribution to Output Growth The previous section has revealed the important features of the magnitude of ICT contribution to output growth, especially its large and increasing variation across economies over the two periods, 1990-1995 and 1995-2000. This section investigates the determinants of the variation of ICT contribution to output growth.
IV.1.Factors Influencing the Magnitude of ICT Contribution to Growth
To investigate the determinants of ICT contribution to growth I look at two sets of factors; the factors in the first set underlie the pace of ICT diffusion and those in the second set have proved to be important for economic growth.
With regards to the first set, previous studies have identified that income level, costs of ICT, education, openness, and institutional quality are the most important factors. Income level appears to be a major determinant of ICT diffusion. Pilat and Devlin (OECD, 2004, Section I) point out that firms in countries with higher levels of income and productivity have greater incentive to invest in ICT. Furthermore, the costs of investment in, and use of, ICT are significant. Quibria et al (2002) and Baliamoune (2002) conclude that income level is a major determinant of ICT diffusion. Pohjola (2003) makes a similar conclusion, finding that income per capita is positively associated with computer hardware spending per capita.
44
Education is a key factor influencing the capability to adopt a new and knowledgeintensive technology such as ICT. Caselli and Coleman (2001) find that computer adoption across 43 countries over 1970-1990 strongly depends on the levels of education of their labor force. Lee (2000) points out that the level of secondary education is important for the adoption of both main lines and mobile phone. Quibria et al (2002) reveal that tertiary education has a significant impact on the penetration of PC and Internet use. Kiiski and Pohjola (2002) find that tertiary education has a positive and statistically significant impact on ICT diffusion for a sample of developing and OECD countries. Furthermore, Pohjola (2003) shows that education, which is measured as average years of schooling, has a positive and statistically significant influence on the adoption of ICT among the 50 ICT-spending economies covered by the WITSA. However, there are also the studies that cast doubt on the role of education in promoting ICT diffusion; for example, Hargittai (1999) and Norris (2000) show that education does not have any predictive power in explaining Internet diffusion.
Openness to international trade facilitates and fosters ICT diffusion. Firms and individuals that expose to international competition and have access to global resources have a more pressing demand for, and better access to, investment in ICT assets. Caselli and Coleman (2001) find that computer adoption significantly depends on the extent of manufacturing import. Baliamoune (2002) concludes that openness has some influence on ICT diffusion over 1998-2000. Pohjola (2003), however, finds no supporting evidence for the positive impact of openness on computer hardware spending per capita.
45
Institutional quality plays a crucial role in facilitating ICT diffusion. Moreover, a good economic environment also enhances the certainty of making profits from investments in new technology. Baliamoune (2002) discusses the institution determinant by examining its two related factors, civil liberties and political freedom. The study finds mixed results: these two factors have a strong influence on the penetration of mobile phones and Internet hosts but their impacts are not significant for PC and Internet use.
The second set of factors that have a significant impact on economic growth includes income level, education, health, institutional quality, and openness (Barro, 1997; Sachs and Warner, 1995; Rodrik et al, 2002). Except for income level, which has a negative e f f e c tone c onomi cgr owt hduet ot he“ c onve r ge nc ee f f e c t ”( Ba r r oa ndSa l a -I-Martin, 1995), the other three factors have solid positive impact on growth. In particular, Sachs and Warner (1995) argue that countries that are open to trade experience unconditional convergence to the income levels of the wealthy nations; Rodrik and his coauthors (Rodrik et al, 2002; Rodrik, 2003) provide convincing evidence that the quality of institutions is even more important than openness and other factors as a deep determinant of growth.
The results of the determinants o ICT diffusion and of economic growth from previous studies, suggest that education, openness, and institutional quality are expected to be among the major determinant of the variation in ICT contribution to output growth across economies. In addition to these factors, I expect that English proficiency is also a potential determinant of ICT contribution to growth.
46
IV.2. Model for Examining the Determinants of ICT Contribution to Growth Because the sample (covering the 50 economies) exhausts the population (which includes the 50 largest ICT-spending economies), it is appropriate to use the fixed effect model to isolate the factors underlying the variation in ICT contribution to growth, excluding the fixed effects induced by individual economies and time periods. The model takes the following form:
[IV-1]
it Cit=+ Xi + Di + T + Xi * T+
where Cit is the magnitude of ICT contribution to output growth (measured in pe r c e nt a g epoi nt s )f ore c onomyi( i =1,2,…,5 0)i npe r i odt( t =1f ort hefist period, 1990-1995, and =2 for the second one, 1995-2000); Xi is a set of the observed characteristics of economy i; Di i sadummyf ore c onomyit oc a pt ur et hee c ono my ’ s unobserved characteristics; T is the time period dummy for the second period (T=1 if t=2, and =0 if t=1; this dummy is also named as After95); Xi * T is a set of interaction terms between the set of the observed characteristics of economy i, Xi, and the time it period dummy T; and finally, is a random effect, independently and identically
distributed among economies and time periods (i=1,..,50; t=1, 2).
The set of variables for the observed characteristics X The set of variables representing the observed characteristics of each individual economy includes the following variables:
47
Developed: This variable is a dummy, which takes the value 1 if the economy is among the 22 industrialized economies; otherwise it takes the value 0. The “ de ve l ope d”s t a t usi ss t r ong l yc or r e l a t e dwi t ht hei nc omel e ve l ; therefore, its impact on the ICT contribution to growth is the net of impacts of the opposite effects: on the one hand, the high income level facilitate ICT diffusion, and hence fosters ICT contribution to growth; on the other hand, the convergence effect makes it harder for a high income economy to generate one percentage point of output growth than for a poor economy. Education: this variable is defined as the educational attainment of the population aged 25 or over in 1995 (Barro and Lee, 1996). The Education variable is expected to positively associate with ICT contribution to growth. Institution (Institutional quality): a good measure for this variable is the Rule of Law index produced by the World Bank (WB, 2003). The Rule of Law index captures the effectiveness of an economy in law enforcement, and therefore is a good proxy for the quality of investment climate in an economy. The Rule of Law index for 1996 is used to define the Institution variable. The variable Institution is expected to positively associate with ICT contribution to growth. Openness: There are two typical measures for openness: the Sachs-Warner openness index and the Trade-to-GDP ratio. The Sachs-Warner openness index is a zero-one dummy taking the value 0 if the economy was closed according to any one of the five subjectively-set criteria27. The Sachs-Warner index is a
27
These criteria include (i) tariff rates exceeding 40 percent, (ii) non-tariff barriers covering more than 40 percent of imports, (iii) socialist economic system, (iv) state monopoly of major export, and (v) black market premium exceeding 20 percent during 1970s or 1980s.
48
desirable index for the whole set of economies in the world but is less relevant for the 50 economies investigated in my study because of two reasons: (i) most of the 50 economies investigated in this study a r e“ ope n”a c c or di ngt oSa c hs Warner Openness Index; therefore, the index provides a variable with little variation among economies; (ii) Things have changed considerably in the 1990s i nt he“ c l os e d”e c onomi e swi t hs ubs t a nt i a lt r a del i be r a l i z a t i ona ndt hedecisive t r a ns i t i onofmos t“ s oc i a l i s t ”c ount r i e st oma r ke te c onomi e s ;a sar e s ul t ,i ti sl e s s pl a us i bl et or e t a i nt he s ee c onomi e si nt he“ c l os e d”s t a t us .TheTrade-to-GDP ratio, which is computed as the ratio between the total trade (sum of exports and imports) and GDP. Although this approach still suffers significant drawbacks28, it is simple and objective; furthermore, it has been widely used in studies on economic growth and ICT diffusion.29 My paper uses the second approach, defining the variable openness as the trade-to-GDP ratio averaged for 19902000. English (English fluency): this dummy variable categorizes the economies that use English as the major language.30 The variable takes the value 1 if an economy belongs to this group and 0 otherwise. One may expect that English proficiency is an important determinant of ICT diffusion as well as ICT
28
For example, countries with a very large economy such as the US have a high level of openness but low trade-to-GDP ratio. 29
For example, Rodrik et al (2002), have uses this approach. Baliamoune (2002) and Pohjola (2003)
30
In our sample of 50 countries, this group consists of nine countries: the US, UK, Ireland, Canada, Australia, New Zealand, South Africa, Hong Kong, and Singapore. For other two countries, India and the Philippines, English is also an official language but it is not spoken by a majority of their populations (from Microsoft Encarta Encyclopedia, 2002 and Gunnemark, 1991).
49
contribution to growth. The countries with English as the major language have a superior advantage in adopting ICT and reaping its benefits because they have more effective communications with the US market, the center of the ICT revolution and the major export market for most countries; moreover, English is the dominant language in the Internet.31 It is surprising, however, to note that Pohjola and Kiiski (2001) find that English proficiency has a negative sign in their model of probing the determinants of Internet penetration over 1995-2000. The possible reason, as they explain, is the lack of data in their variable for English proficiency (defined as the percentage of pupils in secondary education learning English): there are only 17 observations in this dataset.
The above observed characteristics of each individual economy are assumed to be stable through the two periods 1990-1995 and 1995-2000. These characteristics will be investigated to assess the significance of the determinants of ICT contribution to growth. The dummies for individual economies and time periods Each individual economy i (i=1, 2,…,50)i sa s s i g ne dadummyva r i a bl eDi, which takes the value 1 if the observation is for economy i and 0 otherwise. The dummy Di is included in the model to capture the fixed characteristics unique to each individual economy. Practically, only 49 economy dummies (Di, for i=1, 2, …,49)a r ene e de da n dt her e ma i ni nge conomy is the base case).
31
According to a survey completed in January 2000 by NEC Research Institute and Inktomi (NCSE, 2000) the Web has more than one billion unique pages, of which nearly 87 percent are in English.
50
The two time periods, 1990-1995 and 1995-2000, are represented by dummy variable T, which takes the value 1 if the observation is for the second period and 0 otherwise. The time dummy variable T captures the unobserved fixed effects of the second period, which are expected to be distinct from those of the first period due to the prominent emergence of the Internet and the accelerated pace of the ICT revolution.Thet i medummyTi sa l s ona me da s“ Af t e r 95” . The interaction terms M In anticipation of possible accelerated effects of the key explanatory variables, including Income, Education, Institution, Openness, and English, I create the interaction terms for each of these variables with the time dummy T (After95) as follows: Developed_After95 = Developed * After95 Education_After95 = Education * After95 Instituion_After95 = Institution * After95 Openness_After95 = Openness * After95 English_After95 = English * After95 One may expect that the coefficients of these interaction terms are positive and some of them could be statistically significant because these five factors, income level, education, institutional quality, openness, and English language, to some extent, seemed to have accelerated impact on ICT diffusion in the second half of the decade. IV.2. Results The data for examining the determinants of ICT contribution to growth is a set of 100 observations of ICT contribution to growth of the 50 economies for the two
51
time periods 1990-1995 and 1995-2000. The regression is based on model IV-1 and uses the robust estimator of variance. Regression results are reported in Table 3.32
Table 3: Determinants of ICT Contribution to Output Growth
Dependent Variable: ICT Contribution to Output Growth Explanatory Variables Coefficient t-statistics
p-value
Developed
0.035
0.81
0.420
Education
0.020**
2.54
0.015
Institution
0.074***
2.93
0.005
Openness
0.031
1.34
0.189
English
0.095***
3.46
0.001
After95
0.118**
2.19
0.034
Developed_After95
-0.022
-0.68
0.500
Education_After95
0.002
0.27
0.789
Institution_After95
0.061**
2.43
0.019
Openness_After95
0.013
0.67
0.508
0.071**
2.34 0.95
0.024
English_After95 R-squared
100 Number of observations Notes: *, **, and *** indicate, respectively, the 10-percent, 5-percent, and 1-percent significance level.
The following findings stand out: 1) Education, Institutional quality, English fluency have solid positive impact on ICT contribution to growth: The coefficient of the variable Education is positive and significant at the 5 percent significance level. The significance of the Education variable suggests that 32
The results for the 49 economy dummies are not reported to save space.
52
educational attainment is important determinant of ICT contribution to growth. The coefficient of the interaction term Edu_After95 is positive but not statistically significant, which suggests that impact of education in the second half of the 1990s was somewhat accelerated but the accelerated impact was not statistically significant. The coefficient of the variable Institution is large and significant at the 1-percent level. This implies that institutional quality plays a very important role in enhancing the ICT contribution to growth. Furthermore, the coefficient of the interaction term Institution_After95 is positive and significant at the 5-percent level, which indicates that the institutional quality has a significant accelerated effect on the ICT contribution to growth. The finding is consistent with the observation that governments in a number of nations have become highly involved in embracing the ICT revolution to promote economic growth. The quality of institutions is an important factor underlying the effectiveness of the formulation and implementation of their ICT national agenda. The coefficient of variable English is positive and significant at the 1-percent level; furthermore, the coefficient of the interaction term English_After1995 is also positive and significant at the 5-percent significance level. The results indicate that English fluency has a solid impact on ICT contribution to growth and its impact was significantly accelerated in the second period. That is, the countries with English fluency have a superior advantage in reaping the benefits of the ICT revolution; especially the Internet.
53
2)Ope nne s sandt he“De v e l ope d”s t at ushav eapos i t i v ebutnotstatistically significant impact on ICT contribution to output growth: The coefficient of variable Openness is positive but not statistically significant (pvalue=19%). The coefficient of the interaction term Openness_After95 is also positive but not statistically significant. The results imply that that openness has some positive impact on ICT contribution to output growth and this impact was accelerated in the second half of the decade. However, this impact is not statistically significant. Regarding this finding, one should bear in mind the drawback of the variable Openness, which does not well capture the openness of very large economies such as the U.S. Therefore, the implication of this finding is limited. The coefficient of the variable “ developed” is positive but not statistically significant; moreover, the coefficient of the interaction term Developed_ After95 is negative (although it is not statistically significant). There are several reasons explaining this finding: (i) the positive effect of high-income on ICT diffusion does notf a rout we i g ht he“ c onv e r g e nc e ”e f f e c ta sdi s c us s e di npr e vi ouss e c t i on s ;a nd( i i ) “ de ve l ope d” s t a t us doe s not pe r f e c t l y de t e r mi ne t he i nc ome l e ve l ;s e ve r a l economies, such as Singapore or Hong Kong, have a higher income level than some developed nations.
3) The distinct effect of the second period, 1995-2000 The coefficient of the variable After95 is positive and significant at the 5-percent level. In addition, the magnitude of the coefficient is much larger than those of other coefficients. This evidence confirms that there was a boost in ICT contribution to
54
growth in the period 1995-2000 across economies. The finding is consistent with the fact that the period 1995-2000 was characterized by the prominent emergence the Internet and rapid progress in the ICT revolution. The Internet and new ICT products have remarkably enhanced the benefits and lowered the costs of investing in ICT.
V. Impacts of ICT on Economic Growth The previous section confirms that ICT contribution accounts for an increasing share in economic growth from the period 1990-1995 to the period 1995-2000 in most economies. These results, however, are not sufficient to judge how significant ICT has impacted the output growth across economies. This section examine if ICT accumulation is a significant determinant of the variation in growth performance across economies and if ICT is superior to Non-ICT in promoting output growth. V.1. ICT as a Determinant of Output Growth
Model Start with the basic Cobb-Douglass production function: [V-1.A]
Yit = AitKcit1 Kncit2 Lit
where Yit is output, Ait - the technology level, Kcit and Kncit - the ICT and non-ICT capital stocks, and Lit is labor stock (hours worked); subscripts i and t indicate that the observation is for country i in year t.
55
Thet e c hnol ogyl e ve lofe a c hi ndi vi dua lc ount r yi saf unc t i onoft hec ount r y ’ sfixed characteristics and the global technology level shared by all countries. Suppose that this function takes the form Ait = A0eλtDi, where A0eλtis the global technology level shared by all countries ( is the pace of global technology progress over time) and Di represents the fixed characteristics of country i. Therefore, equation V-1.A can be rewritten as Yit = A0eλtDi Kcit1 Kncit2 Lit
[V-1.B]
Taking the logarithm for equation V-2 and rearranging the terms lead to equation
[V-2]
ln(Yit)= 1ln(Kcit) + 2ln(Kncit) + ln(Lit)+ t + ln(A0) + ln(Di)
First-differencing equation [V-3] yields an equation for the rate of output growth as the following: [V-3]
dYit= 1dKcit + 2dKncit + dLit+ + it
where dYit=Ln(Yit)-Ln(Yit-1) is the growth rate of output (from year t-1 to year t); dKcit=Ln(Kcit)-Ln(Kcit-1) and dKncit=Ln(Kncit)-Ln(Kncit-1) –the growth rates of ICT and non-ICT capital stocks; dLit=Ln(Lit)-Ln(Lit-1) –the growth rate of hours worked; is the constant term of the technology progress over the period 1990-2000, and it is the random noise, identically and independently distributed among economies and years. Results and Discussions Because the growth of output and the capital and labor inputs may have simultaneous effects on each other, the estimates produced by the OLS regressions tend to be upwardbiased and should be taken with caution. The IV regressions are designed to surmount
56
this simultaneity problem. The instrumental variables used for the IV regressions are the 1-period and 2-period lags of the independent variables and the time dummy.
The OLS and IV regressions are based on model [V-3] and both use the robust estimator of variance to correct for the heteroskedasticity of the variance. Table 4 provides results of the OLS and IV regressions for the whole sample of 50 economies and for the subsample of 22 industrialized economies.
Table 4: ICT Capital as a Determinant of Output Growth Dependent Variable: dY (output growth rate)
OLS Results Coefficient
dKnc dKc dL Constant N R2
0.44*** 0.10*** 0.87*** -0.01
dKnc dKc dL Constant N R2
0.06 0.08*** 0.97*** 0.00
t-statistics
IV Results p-value
Coefficient
t-statistics
p-value
All Sample (50 Economies) 9.64 0.000 0.34*** 3.90 0.000 0.05** 10.02 0.000 0.74*** -2.39 -0.017 0.00 550 0.51
5.91 1.97 4.57 0.88 550 0.49
0.000 0.049 0.000 0.378
Industrialized Group (22 Economies) 0.77 0.439 -0.14 3.27 0.001 0.15*** 10.33 0.000 1.08*** 1.08 0.280 0.00 242 0.57
-1.61 3.58 7.83 0.04 242 0.54
0.110 0.000 0.000 0.966
The following findings stand out: For the entire sample of the 50 economies, the ICT capital variable is significant at the 1 percent level in the OLS regression and 5-percent in the IV regression. The results indicate that ICT is a significant determinant for the variation in output
57
growth across the 50 economies during the last decade (1990-2000). Furthermore, the coefficient of the ICT capital variable, dKc, in the IV regression implies that, on average, a 10 percent increase in the ICT capital stock adds about 0.45 percent points to output growth. The results also show that Non-ICT capital and labor are important determinants of output growth: Non-ICT capital and labor variables are all significant at the 1 percent level in both OLS and IV regressions.
For the subsample that includes only 22 developed economies, the ICT variable is significant at the 1 percent level in both OLS and IV regressions. Moreover, the magnitude of its coefficient is much larger for the subsample than the entire sample (0.153 vs. 0.045). The results indicate that ICT plays a more important role in determining the output growth for the developed economies than for the developing ones. For an average developed economy, a 10 percent increase in the ICT capital stock adds 1.5 percentage points to output growth. Regarding the Non-ICT capital and labor, the results show that labor variable is significant at the 1 percent level in both OLS and IV regressions, while non-ICT capital is not statistically significant in either of the two regressions.
The findings, thus, confirm that the accumulation in ICT capital has a solid causal effect on output growth and this impact is even stronger for the industrialized group.
58
V.2. Impact of ICT on the Quality of Economic Growth
Thet e r m“ qua l i t yofe c onomi cg r owt h”is defined as follows: an economy with a higher quality of growth can achieve a higher output growth for given growth in aggregate capital and labor inputs.
Previous studies have found that ICT has a positive impact on the quality of growth. At the firm level, Brynjolfsson and Hitt (1995, 1996, 2003) find a positive relationship between IT investment and firm productivity levels; firms investing more in IT produce more output per unit of input. Lichtenberg (1995) estimate firm production functions and finds an excess return on IT equipment and labor. Lehr and Lichtenberg (1999) inspects firm-level data among service industries and report evidence that personal computers make a positive and significant contribution to productivity growth At the industry level, Siegel (1997), examining the US manufacturing industries, reveals that computerization has a significant positive effect on productivity growth. Stiroh (2002), investigating the US 57 major industries, confirms a strong link between IT accumulation and productivity growth; in particular, among the major input classes, only IT-capital deepening is significantly associated with future productivity a c c e l e r a t i on.O’ Ma honya ndVe c c hi( 2003) ,i nve s t i g a t i ngi ndus t r yda t af ort heUSa nd the UK, unearth that ICT has a positive and significant long-run impact on TFP.
At the economy level, Pohjola (2000), by examining IT investment in 39 countries over the period of 1980-1995, concludes that net return on IT investment was much larger
59
than net return on non-IT investment; this implies that investment in IT fosters the quality of growth.
While positive assessments of the role of ICT in productivity growth seem to prevail, their opposite view remains challenging. Gordon (2002) claims that “ c omput e rc a pi t a l did not have any kind of magical or extraordinary effect—it earned the same rate of r e t ur na sa nyot he rt y peofc a pi t a l ” .
This section examines the potential impact that accumulation in ICT capital stock per capita may have on the quality of economic growth. In this investigation, the level of ICT capital stock per capita will be used as proxy for the depth of ICT penetration in an economy. Model Le t ’ ss t a r twi t ht heCobb-Douglas production function in its basic form: [V-4.A]
Yit = AitKitLit
where Yit, Ait, Kit, and L are, respectively, the levels of output, technology, aggregate capital and labor inputs of country i in year t. The aggregate capital input Kit is proportional to the aggregate capital stock, which can be computed from the flow of gross fixed capital formation. Labor input is measured in hours worked.
Suppose that the technology level Ait of country i in year t is a function of global technology level A0eλt( whe r eλi st hepa c eofg l oba lt e c hnol ogypr ogr e s sove rt i me shared by all countries), the country-specific effects Di, and a variable representing the
60
depth of ICT penetration such as the ICT capital stock per capita variable, kcit. Let assume that the technology level Ait takes the following form Ait = A0eλtDi kcitγ Equation [V-4.A], therefore, can be rewritten as Yit = A0eλtDi kcitγKitLit
[V-4.B]
Taking the logarithm for equation [V-4B], then, rearranging the terms lead to the following equation: [V-5]
ln(Yit)= ln(Kit) + ln(Lit) +γ l n( kc t + ln(A0) + ln(Di) it) +
First-differencing equation [V-5] yields an equation for output growth rate as the following: [V-6]
dYit= dKit + dLit +γ dkc it+ + it
where dYit = Ln(Yit)-Ln(Yit-1) is the growth rate of output (from year t-1 to year t); dKit = Ln(Kit)-Ln(Kit-1) –the growth rate of aggregate capital input; dLit = Ln(Lit)-Ln(Lit-1) – the growth rate of labor input; dkcit = Ln(kcit)-Ln(kcit-1) –the growth rate of the ICT capital stock per capita level (which is also labeled as “ dkc pe r c a p”t oma kei te a s i e rt o remember), is the constant term,
and it is the random noise, identically and
independently distributed among all economies and years.
Results Both OLS and IV regressions are used to examine the impact of ICT on the quality of growth. As mentioned in section V.1, the IV regression is designed to surmount the simultaneity problem. The instrumental variables used for the IV regressions are the 1-
61
period and 2-period lags of the independent variables and the time dummy. Both regression are based on model [V-6] and use the robust estimator of variance to correct for heteroskedasticity.
Table 5 provides results of the OLS and IV regressions for the entire sample of the 50 economies and for the subsample of 22 developed economies.
Table 5: Impact of ICT Stock per Capita on the Quality of Growth Dependent Variable: dY (output growth rate)
OLS Results Coefficient
dK dL dkcpercap Constant N R2
dK dL dkcpercap Constant N R2
0.44*** 0.90*** 0.10*** -0.10
0.10 0.97*** 0.90*** 0.03
t-statistics
IV Results p-value
Coefficient
All Sample (50 Economies) 9.50 0.000 0.36*** 10.61 0.000 0.76*** 3.87 0.000 0.04 -2.64 -0.009 0.00 550 0.52 Industrialized Group (22 Economies) 1.19 0.235 -0.06 10.51 0.000 1.05*** 3.59 0.000 0.16*** 0.74 0.461 -0.00 242 0.58
t-statistics
p-value
5.92 4.94 1.49 0.95 550 0.49
0.000 0.000 0.136 0.343
-0.68 7.72 3.89 -0.20 242 0.55
0.500 0.000 0.000 0.838
The following findings stand out: For the entire sample of the 50 economies, the ICT stock per capita variable is significant at the 1 percent significance level in the OLS regression but not
62
statistically significant in the IV regression. The OLS result implies that for the entire group of the 50 economies, output growth across economies is strongly associated with the level of ICT stock per capita, controlling for the growth in the aggregate capital and labor inputs. However, this associated effect is not a causal effect because there might be a simultaneous effect between growth in output and in the ICT capital stock per capita. The IV regression result indicates that the coefficient of the ICT stock per capita is positive but not statistically significant (pvalue=0.136). The result implies that the causal effect of the ICT capital stock per capita on the quality of output growth is not solid for the whole sample of the 50 economies.
For the subsample of the 22 developed economies, the ICT stock per capita variable is significant at the 1 percent significance level in both OLS and IV regressions. The results imply that for the developed economies, the ICT capital accumulation per capita has a solid causal effect on the quality of growth; that is, for given growth rates of aggregate capital (dK) and hours worked (dL), an increase in ICT capital stock per capita clearly enhances the growth rate of output.
63
VI. Conclusion This paper examines the contribution of ICT to, and its impact on, global economic growth.
The study finds that ICT contribution to economic growth is a global phenomenon, which is evident not only in developed economies but also in developing ones. The ICT contribution to growth in most economies drastically increased from the period 19901995 to the period 1995-2000, while its variance was also strikingly widened. The key determinants of the variance in ICT contribution to growth across economies are education, institutional quality, openness, and English fluency. Furthermore, the impacts of institutional quality and English fluency are significantly accelerated over time.
The study confirms that ICT has a significant impact on economic growth. First, the accumulation in ICT capital stock is a significant determinant of the variation in output growth across economies. Second, ICT is superior to Non-ICT in enhancing output growth: for given levels of growth in labor and capital inputs, a higher level of ICT capital stock per capita allows a typical economy to achieve a higher output growth rate.
This study can be broadened for a larger country sample, probably of 110-120 economies for which data on ICT penetration are available. In this future study, the data on ICT penetration will be used to estimate investment flows in ICT assets.
64
Bibliography Accenture. 2001. Creating a Development Dynamic; Final Report of the Digital Opportunity Initiative. ”Ac c e nt ur e ,Ma r kl eFounda t i on,UNDP,J ul y2001. ADB. 2003. Key Indicators 2003. Volume 34, Asian Development Bank. Aghion, Philippe, and Peter Howitt. 1998. Endogenous Growth Theory. MIT Press, Cambridge, Mass. And London. Ar k,Ba r tva n,J oha nnaMe l ka ,Na nn oMul de r ,Ma r c e lTi mme r ,a ndGe r a r dYpma .2002.“ I CT Investment and Growth Accounts for the European Union, 1980-2000. ”Final Report on `ICT and Gr owt hAc c ount i ng’f ort heDGEc on omi c sandFi nanc eoft heEur ope anCommi s s i on, Brussels. Ar ms t r ong ,P. ,T. M.Ha r c ha oui ,C.J a c ks on,a ndFTa r kha ni .2002.“ACompa r i s onofCa na da -US Economic Growth in the Information Age, 1981-2000: The Important of Investment in I nf or ma t i on a nd Communi c a t i on Te c hnol og i e s ” ,Ec onomi c Re s e a r c h Pa pe rSe r i e s No.70, February, Zurich. Ar r ow,Ke n ne t h.1962.“ TheEc onomi cI mpl i c a t i onsofLe a r ni ngbyDoi ng ” .Review of Economic Studies 29(1962) Arvanitis. 2004. Information Technology, Workplace Organization, Human Capital, and Firm Pr oduc t i vi t y :Evi de nc ef ort heSwi s sEc onomy , ”i nThe Economic Impact of ICT--Measurement, Evidence, and Implications, pp. 183-211, OECD, Paris, 2004. Atrostic, B.K and S. Nguyen. 2002. Computer Networks and US Manufacturing Plant Productivity, New Evidence from the CNUS Data, Centre for Economics Studies, January 2002. Ba i l e y ,J os e phP.2001.“ Re t a i lSe r vi c e s :Cont i nui ngt heI nt e r ne tSuc c e s s . ”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 173-188, Brookings Institute Press. Washington DC. Baldwin, John R., David Sabourin, and David Smith. 2004.“ Fi r m Pe r f or ma nc ei nt heCa na di a n Food Processing Sector: The Interaction between ICT, Advanced Technology Use and Human Re s our c eCompe t e nc i e s . ”i nThe Economic Impact of ICT, pp. 153-179, OECD, Paris, 2004. Barro R. and Sala-I-Martin, X.1995. Economic growth, McGraw-Hill: New York. Barro, Robert and Jong-Wha Le e .1996.“ I nt e r na t i ona l Measures of Schooling years and Sc hool i ngQua l i t y , ”American Economic Review 86(2): 218-23. Ba r r o,Rob e r t .1991.“ Ec onomi cGr owt hi naCr os s -Se c t i onofCount r i e s . ”Quarterly Journal of Economics, 106, 2(May): 407-433.
65
Barro, Robert. 1997. Determinants of Economic Growth. MIT Press, Cambridge, MA, 1997. Bos wor t hBa r r ya ndJ a c kE.Tr i pl e t t .2 000.“ Wha t ’ sNe wa boutt heNe wEc onomy ?I T,Ec onomi c Gr owt ha ndPr oduc t i vi t y ” .Br ooki ng sI ns t i t ut i on. Br y nj ol f s s on,E.a ndLHi t t .1994.“ I nf or ma t i onTe c hnol ogyas a Factor of Production: The Role ofDi f f e r e n c eAmongFi r ms ” ,MI TSl oa nSc hoolofMa na g e me ntWP371 5-94, August 1994. Br y nj ol f s s on,E.a ndL.Hi t t .1993.“ Ne wEvi de nc eont heReturns on I nf or ma t i onSy s t e ms ” ,MI T Sloan School of Management WP 3571-93, October 1993. Br y nj ol f s s on,E.a ndL.Hi t t .1996.“ Pa r a doxLos s ?Fi r m-level Evidence on the Returns on I nf or ma t i onSy s t e mSpe ndi ng . ”Management Science 42(4): 541-558. Br y nj ol f s s on, E. a nd L. Hi t t .2 0 00. “ Be y ond Comput a t i on: I nf or ma t i on Te c hnol ogy , Org a ni z a t i ona lTr a ns f or ma t i ona ndBus i ne s sPr a c t i c e s . ”Journal of Economic Perspectives, Vol. 14, No. 4, Fall 2000, pp. 23-48. Br y nj ol f s s on,E.a ndL.Hi t t .2003.“ Compu t i ngPr oduc t i vi t y :Fi r m-Le ve lEvi de nc e ” .Pa p e r139, Center for eBusiness at MIT, June 2003. Br y nj ol f s s on,E.a ndSYa ng .1996.“ I nf or ma t i onTe c hnol ogya ndPr oduc t i vi t y :A r e vi e w oft he l i t e r a t ur e ” , Advances in Computers, 43, pp. 179-214 Ca s e l l i ,Fr a nc e s c oa ndCol e ma n,W.J ohnI I .2001.“ Cr os s -Country Technology Diffusion: The Case of Comput e r s . ”American Economic Review, May 2001, 91(2). Cette, G., Y. Kokoglu, and J. Mairesse. 2000.“ TheDi f f us i onofI nf or ma t i ona ndCommuni c a t i on Te c hnol og i e si nFr a nc e .Me a s ur e me n ta ndCont r i but i ont oGr owt ha ndPr oduc t i vi t y . ”Economie and Statistique, No. 339-340. Cha r l e s ,Fi nea ndDa ni e lRa f f .2001.“ Aut omot i veI ndus t r y :I nt e r ne t -Driven Innovation and Ec onomi cPe r f or ma nc e . ”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 62-86, Brookings Institute Press. Washington DC. Chow,Gr e g or yC.1993.“ Ca pi t a lFor ma t i ona ndEc onomi cGr owt hi nChi na ” .Quarterly Journal of Economics. CVIII(3): 809-842, August 1993. Chr i s t e ns e n,La ur i t sR. ;Cummi ng s ,Di a nnea ndJ or g e ns on,Da l eW.1995a .“ Ec onomi cGr owt h, 1947-1973:AnI nt e r na t i ona lCompa r i s on. ”I nDa l eW.J or g e ns on,e d. ,International Comparisons of Economic Growth. Cambridge, MA: MIT Press, pp. 203-96. Chr i s t e ns e n,La ur i t sR. ;Cummi ng s ,Di a nnea ndJ or ge ns on,Da l eW.1995b.“ Re l a t i vePr oduc t i vi t y Levels, 1947-1973:An I nt e r na t i ona lCompa r i s on. ”I nDa l eW.J or g e ns on,e d. ,International Comparisons of Economic Growth. Cambridge, MA: MIT Press, 199a, pp. 203-96.
66
Clarke, George. 2001. Bridging the Digital Divide: How Enterprise Ownership and Foreign Competition Af f e c tI nt e r ne tAc c e s si n Ea s t e r n Eur ope a nd Ce nt r a lAs i a ” .World Bank Development Research Paper. Cl e mons ,Er i ca ndLor i nHi t t .2001.“ Fi na nc i a lSe r vi c e s :Tr a ns pa r e nc y ,Di f f e r e nt i a lPr i c i ng ,a nd Di s i nt e r me di a t i on”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 87-128, Brookings Institute Press. Washington DC. Colecchia, Alessandra and Paul Schreyer. 2001.“ I CTI nve s t me nta ndEc onomi cGr owt hi nt he 1990s: Is the United Sates a Unique Case? A Comparative Study of Nine OECD Countries.” OECD Working Paper No. DSTI/DOC (2001)7. Counc i lofEc onomi cAdvi s e r s .2001.“ Annua lRe por toft heCounc i lofEc onomi cAdvi s e r s . ”The Economic Report of the President, January 2001. Danzon, Patricia and Michael Furukawa. 2001.“ He a l t hc a r e :Compe t i t i ona ndPr oduc t i v i t y . ”I n The Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin. Pp. 189-234. Brookings Institute Press. Washington DC. Da ve r i ,Fr a nc e s c o.2002.“ TheNe w Ec onomy i n Eur ope 1992-2001” .Di s c us s i on Pa pe rN. 2002/70, WIDER, United Nation University. De ut s c heBa nkRe s e a r c h.2003.“ As i a nTi ge r sa f t e rt heI TBoom” ,Economics, No. 40, September 30, 2003. Dewan S. and K. Kraemer. 1998.“ I nt e r na t i ona lDi me ns i on oft he Pr oduc t i vi t y Pa r a dox. ” Communication of ACM, 41(8) pp. 56-62. Dewan S. and K. Kraemer. 2000.“ I nf or ma t i onTe c hnol ogya ndPr oduc t i vi t y :Evi de nc ef r om Country-l e ve lDa t a . ”Ce nt e rf orRe s e a r c honI nf or ma t i onTe c hnol ogy ,Uni ve r s i t yofCa l i f or ni a , Irvine, December 2000. Dol l a r ,Da vi d.1992.“ Out wa r d-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-85. ”Economic Development and Cultural Change. pp. 523-544. Ea s t e r l y ,W.a ndR.Le vi ne ,2000.“ I t ’ sNotFa c t orAc c umul a t i on:St y l i zed Facts and Growth Mode l s . ”World Bank Working Paper. Eggleston, Karen, Robert Jensen, and Richard Zechauser. 2002.“ I nf or ma t i ona ndCommu n i c a t i on Te c hnol og i e s ,Ma r ke t s ,a ndEc onomi cDe ve l opme nt . ”i nThe Global Information Technology Report 200-2002: Readiness for the Networked World, CID, Harvard University. Fine, Charles H., and Daniel M.G. Raff. 2001.“ Aut omot i veI ndus t r y :I nt e r ne t -Driven Innovation a ndEc onomi cPe r f or ma nc e . ”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 62-86, Brookings Institute Press. Washington DC.
67
Gr e t t on,P. ,J .Ga l i ,a nd D.Pa r ha m.2002.“ Upt a kea nd I mpa c t sofI CT i nt heAu s t r a l i a n Ec onomy :Evi de nc ef r om Agg r e ga t e ,Se c t or a la nd Fi r m Le ve l s ” ,pa pe rpr e s e nt e da tOECD Workshop on ICT and Business Performance, Productivity Commission, Canberra, December. Griliches, Zvi and Donald Siegel. 1991.“ Pur c ha s e d Se r vi c e s ,Out s our c i ng ,Comput e r s ,a nd Pr oduc t i vi t yi nMa nuf a c t ur i ng . ”NBERWor ki ngPa pe r#3678,Apr i l1991. Grossman, Ge nea ndEl ha na nHe l pma n.1994.“ Endog e nousI nnova t i oni nt heThe or yofGr owt h. ” Journal of Economic Perspectives, Winter 1994, 8(1), pp. 23-44. Gunnemark, Erik V. 1991. Countries, Peoples, and their Languages: The Geolinguistic Handbook, Dallas, Texas: Summer Institute of Linguistics, Inc., 1991. Ha nna ,Na gy ,Sa ndorBoy s on,a ndSha kunt a l aGuna r a t ne .1996.“ TheEa s tAs i a nMi r a c l ea nd I nf or ma t i onTe c hnol ogy . ”Wor l dBa nkDi s c us s i onPa pe r s ,#326,TheWor l dBa nk,Wa s hi ng t on DC. HBS. 2002.“ Bui l di ngaCl us t e r :El e c t r oni c sa ndI nf or ma t i onTe c hnol ogyi nCos t aRi c a ”Harvard Business School Case, 9-703-422, July 2002. Helpman, E., Editor. 1998. General Purpose Technology and Economic Growth. Cambridge, MA: The MIT Press. Hempell, Thomas and van der Wiel. 2004.“ I CT,I nnova t i on a nd Bus i ne s sPe r f or ma nc ei n Se r vi c e s :Evi de nc ef orGe r ma ny a nd t heNe t he r l a nds , ”i n The Economic Impact of ICT— Measurement, Evidence, and Implications, pp. 131-152, OECD, Paris, 2004. Hubbard, Thomas N. 2003.” I nf or ma t i on,De c i s i ons ,and Productivity: On-Board Computers and Ca pa c i t yUt i l i z a t i oni nTr uc ki ng”The American Economic Review, Vol. 93, no. 4, pp. 1328-53. Jalava, J., and M. Pohjola. 2001.“ Ec onomi cGr owt hi nt heNe w Ec onomy ” .WIDER Discussion Paper 2001/5. Helsinki: UNU/WIDER. Jorgenson, Dale and Stiroh Kevin. 1995.“ Compu t e r sa ndGr owt h. ”Economics of Innovation and New Technology, Vo. 3, pp. 109-115. Jorgenson, Dale and Stiroh Kevin. 1999.“ I nf or ma t i on Te c hnol ogy a nd Gr owt h. ”American Economic Review, May 1999, 89(2), pp. 109-115. Jorgenson, Dale and Stiroh Kevin. 20 0 0.“ Ra i s i ngt heSpe e dLi mi t :USEc onomi cGr owt hi nt he I nf or ma t i onAg e . ”OECD Working Papers, No. 261. Jorgenson, Dale and Zvi Griliches. 1967.“ TheExpl a na t i onofPr oduc t i vi t yCha nge . ”Review of Economic Studies, 34(July), 249-280.
68
Jorgenson, Dale W, Frank M. Gollop, and Barbara M. Fraumi.1987. Productivity and U.S. Economic Growth, Cambridge MA: Harvard University Press. J or g e ns on,Da l eW.1966 .“ TheEmbodi me ntHy pot he s i s . ”Journal of Political Economy, Vol. 74, No. 1, 1-17. J or g e ns on,Da l eW.2001.“ I Ta ndt heU. S.Ec onomy . ”American Economic Review, March 2001, 91(1), pp. 1-32. J or g e ns on,Da l eW.200 3.“ I nf or ma t i onTe c hnol ogya ndt heG7Ec onomi e s . ”World Economics, Vol. 4, No. 4, October-December, 2003, pp. 139-169. Jorgenson, Dale W. and Zivi Griliches. 1967.“ TheExpl a na t i onofPr odu c t i vi t yCha nge . ”Review of Economic Studies, Vol. 34, No. 3, July, 249-283. Jorgenson, Dale W., and Kazuyuki Mot oha s hi .2003.“ Economic Growth of Japan and the United States in the Information Age” , REITI Discussion Paper 03-E-015, July 2003. Jorgenson, Dale W., Mun Ho, and Kevin J. Stiroh. 2002. "Projecting Productivity Growth: Lessons from the U.S. Growth Resurgence." Federal Reserve Bank of Atlanta Economic Review, Third Quarter 2002, p. 1-13. Ka purDe v e s h.2002.“ TheCa us e sa ndCons e que nc e sofI ndi a ’ sI TBoom. ”India Review 1(2), April 2002, pp. 91-110. Ke g e l s ,C. ,M.Va nOve r be ke ,a ndW.Va nZa ndwe g he .2002.“ I CTCont r i but i ont oEc onomi c Performanc ei nBe l g i um:Pr e l i mi na r yEvi de nc e ” ,Wor ki ngPa pe r8-02, Federal Planning Bureau, Brussels, September. Kendrik, John W. 1961. Productivity Trends in the United States, Princeton NJ, Princeton University Press. Kha n,H.a ndM.Sa t os .2002.“ Cont r i but ion of ICT use to Output and Labor Productivity Growth i nCa na da ” , Wor ki ngPa pe r2002-7, Bank of Canada, Ottawa, March. Ki l l e y ,Mi c ha e lT.1999.“ Comput e ra ndGr owt hwi t hCos tofAdj us t me nt :Wi l lt heFut u r eLook Li ket hePa s t ? ”Federal Reserves Board, Finance and Economic Discussion Series Paper, pp. 1996-2036, July 1999. Ki m,J .a nd L. J .La u. 1994.“ TheSour c e sofEc onomi cGr owt h oft heEa s tAs i a n Ne wl y I ndus t r i a l i z e dCount r i e s ” .Journal of the Japanese and International Economies, 8(3), 235-271.
Kim, S.J. 2002. The Digital Economy and the Role of Government: Information Technology and Economic Performance in Korea, Program on Information Resources Policy, Harvard University, January.
69
Klenow, O.J., and A. Rodriguez-Clare. 1997. “ TheNe oc l a s s i c a lRe vival in Growth Economics: Ha sI tGonet ooFa r ? ”NBER Macroeconomics Annual 1997(12), 73-103. Kr a e me rK.a ndJ .De dr i k.1994“ Pa y of f sf r om I nve s t me nti nI nf or ma t i onTe c hnol ogy :Le s s ons from the Asia-Pa c i f i cRe g i on” .World Development, 22(12), pp. 1921-1931. Kr a e me rK.a ndJ .De dr i k.2000.“ Li be r a l i z a t i ona ndt heComput e rI ndus t r y :A Compa r i s onof FourDe ve l opi ngCount r i e s . ”Ce nt e rf o rRe s e a r c honI nf or ma t i onTe c hnol ogya ndOr ga ni z a t i ons , University of California, Irvine. Kr ug ma n,P. ,1994.“ TheMy t hofAs i a ’ sMi r a c l e . ”Foreign Affairs, November/December. Le e ,Fr a nkC.a ndDi r kPi l a t .2001.“ Pr oduc t i vi t yGr owt hi nI CT-Producing and ICT-Using –A Sour c eofGr owt hDi f f e r e nt i a l si nt heOECD? ”OECD Working Paper, No. DSTI/DOC(2001)4 Lee, Il Houng and Youge s hKha t r i .2003.“ I nf or ma t i onTe c hnol ogya ndPr oduc t i vi t yGr owt hi n As i a . ”IMF Working Paper. Lerh, William, and Frank Lichtenberg. 1999.“ I nf or ma t i on Te c hnol ogy a nd I t sI mp a c ton Productivity: Firm-Level Evidence from Government and Private Data Sources 1973-1993” . Canadian Journal of Economics, Vol. 32, No. 2, April, pp. 335-362. Li c ht e nbe r g ,Fr a nk.1995.“ TheOut putCont r i bu t i onsofCo mput e rEqui p me nta ndPe r s onne l :A Fi r m Le ve lAna l y s i s ” .Economics of Innovation and New Technology, Vol. 3, No. 4, pp. 201-217. Luc a s ,He nr yJ r .a ndRi c ha r dSy l l a .2000.“ TheGl oba lI mpa c toft heI nt e r ne t :Wi de n i ngt he Ec onomi cGa pbe t we e nWe a l t hya ndPoorNa t i ons ? ”Smith papers Online, 2000, Robert H. Smith School of Business, Maryland University. Maliranta and Rouvinen. 2004.“ I CTa ndBus i ne s sPr oduc t i vi t y :Fi ni s hMi c r oLe ve lEvi de nc e , ”i n The Economic Impact of ICT--Measurement, Evidence, and Implications, pp. 213-239, OECD, Pa r i s ,2004. ” Mc Af e e ,An dr e w.2001.“ Ma nuf a c t u r i ng :Lowe r i ngBounda r i e s ,I mpr ovi ngPr odc ut i vy . ”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 29-61, Brookings Institute Press. Washington DC. Mc Guc ki ne ta l .1998.“ Adva nc e dTe c hnol ogyUs a g ea ndPr oduc t i vi t yGr owt h” ,Economics of Innovation and New Technology (7), pp.1-26. Milana and Zeli. 2004.“ Pr oduc t i vi t ySl owdowna ndt heRol eofI CT i nI t a l y :A Fi r m Le ve l Ana l y s i s , ”i nThe Economic Impact of ICT--Measurement, Evidence, and Implications, pp. 213239, OECD, Paris, 2004.
70
Miyagawa, T. ,Y.I t o,a ndN.Ha r a da .2002.“ Doe st heI T Re vol ut i onCont r i but et oJ a pa ne s e Ec onomi cGr owt h? ”J CER Di s c us s i onPa pe rNo.75,J a pa n Ce nt e rf orEc onomi cRe s e a r c h, Tokyo. Motahashi. 2001.“ Ec onomi c Ana l y s i s of I nf or ma t i on Ne t wor k Us e :Or ga ni z a t i ona la nd Pr oduc t i vi t yI mpa c tonJ a pa ne s eFi r ms ” ,Re s e a r c ha ndSt a t i s t i c sDe pa r t me nt ,METI ,mi me o. Mun Sung-Ba ea nd I s ha q Na di r i .2002.“ I nf or ma t i on Te c hnol ogy Ext e r na l i t i e s :Empi r i c a l Evi de nc ef r om42U. S.I ndus t r i e s . ”NBER Working Paper Series #9272, October 2002. Na g a r a j a n Anur a dha ,Enr i queCa ne s s a ,Wi l lMi t c he l l ,a nd C. C Whi t eI I I .2001.“ Tr uc ki ng I ndus t r y :Ch a l l e ng e st oKe e pPa c e . ”I nThe Economic Payoff from the Internet Revolution, edited by Robert Litan and Alice Rivlin, pp. 129-171, Brookings Institute Press. Washington DC. Nasscom (National Association of Software and Service Companies) website (www.nasscom.org), October 10, 2003. Na udha usWi l l i a m D.2 001.“ Pr oduc t i vi t yGr owt ha ndTheNe w Ec onomy ” .NBER Working Paper No. 8096, January 2001. Ne l s on,R. R.a ndH.Pa c k.1997.“ TheAs i a nMi r a c l ea ndMode r nGr owt hThe or y ” .World Bank Working Paper. O’ Ma hony ,Ma r ya ndMi c he l aVe c c hi .2003.“ I sThe r eAnI CTI mpa c tonTFP?AHe t e r o ge ne ous Dy na mi cPa ne lAppr oa c h” .NIESR, June 2003. OECD. 2000. OECD Information Technology Outlook, OECD, Paris. OECD. 2001. Science, Technology and Industry Outlook, OECD, Paris. OECD. 2001a. Measuring Productivity. OECD Manual: Measurement of Aggregate and IndustryLevel Productivity Growth, OECD, Paris. OECD. 2001b. Measuring Capital – OECD Manual: Measurement of Capital Stocks, Consumption of Fixed Capital and Capital Services, OECD, Paris. OECD. 2002a. OECD Information Technology Outlook, OECD, Paris. OECD. 2002b. Purchasing Power Parities and Real Expenditures; 1999 Benchmark Year, OECD, Paris. OECD. 2003. ICT and Economic Growth—Evidence from OECD Countries, Industries, and Firms, OECD, Paris. OECD. 2004. The Economic Impact of ICT. OECD, Paris.
71
Oliner, S., and D. Sichel. 2001.“ TheRe s ur g e nc eofGr owt hi nt heLa t e1990s :i sI nf or ma t i on Te c hnol ogyt heSt or y ? ”Journal of Economic Perspectives, (14/4): pp. 3-22. Oulton, Nick (2002), 'ICT and productivity growth in the UK', Oxford Review of Economic Policy, Vol. 18, pp. 363-379. Oyelaran-Oy e y i nka ,Ba nj ia ndCa t he r i neNy a kiAde y a .2002.“ I nt e r ne tAc c e s si nAf r i c a :An Empi r i c a lExpl or a t i on” .INTECH Discussion Paper Series #2002-5. The United Nations University. Pa r ha m,D. ,P.Robe r t s ,a ndH.Sun.2 001.“ I nf or ma t i onTe c hnol ogya ndAus t r a l i a ’ sPr oduc t i vi t y Sur g e ” ,St a f fRe s e a r c hpa pe r ,Pr oduc t i vi t yCommi s s i on,Aus I nf o,Ca nbe r r a . Paul Gretton, Jyothi Gali, and Dean Parham. 2004.“ TheEf f e c t sofI CTsa ndCompl e me nt a r y I nnova t i onsonAus t r a l i a nPr oduc t i vi t yGr owt h, ”in The Economic Impact of ICT--Measurement, Evidence, and Implications, pp. 105-130, OECD, Paris, 2004. Pe r ki ns ,Dwi g ht .2000.“ Fi na nc i a la ndI ndus t r i a lPol i c yofChi naa ndVi e t na m” :i nSt i g l i t za nd Yusuf (ed.), Rethinking the East Asian Miracle, Washington, D.C., New York: Oxford University Press. Pi l a tDi r k ,a nd Fr a nk Le e .2001.“ Pr oduc t i vi t y Gr owt hi nI CT-Producing and ICT-Using Industries: A Source of Growth Differentials in the OECD? STI Working Papers 2001/4, OECD. Pilat, Dirk and Anita Wolfl. 2004.“ I CT Pr oduc t i ona ndI CT Us e :Wha tr ol ei nAgg r e g a t e Pr oduc t i vi t y Gr owt h? ”i n The Economic Impact of ICT--Measurement, Evidence, and Implications, pp. 85-104, OECD, Paris, 2004. Pohj ol aM.2000.“ I nf or ma t i onTe c h n ol ogya ndEc onomi cGr owt h:A Cr os s -Count r yAna l y s i s ” . The United Nations University Working Paper No. 73, January 2000. Porter, Michael. 1998. On Competition, Boston: Harvard Business School Press, October 1998. Qi a ng Chr i s t i ne s ZW. ,a nd Al e xa nde r Pi t t . 2003. “ Cont r i but i on of I nf or ma t i on and Communi c a t i onTe c hnol og i e st oGr owt h” .World Bank Working Paper No. 24, December 2003. Roa c h,St e phe n.1998.“ Nopr oduc t i v i t yBoom f orWor ke r s ” .Issues in Science and Technology, XIV(4) Summer 1998, pp. 49-56. Rodrik, Dani, Arvind Subramanian, and Fr a nc e s c oTr e bbi .2002.“ I ns t i t ut i on:ThePr i ma c yof I ns t i t ut i onsove rGe og r a phya ndI nt e g r a t i oni nEc onomi cDe ve l opme nt . ”NBER Wor ki ngPa pe r No. 9305. Rodr i k,Da ni .2003.“ I nt r oduc t i on. ”I nDa ni e lRodr i ke d. ,In Search of Prosperity: Analytic Narratives on Economic Growth. NJ: Princeton University Press.
72
Roe g e rWe r ne r .2001.“ TheCont r i but i onofI nf or ma t i onTe c hnol og i e st oGr owt hi nEur opea nd t heUS:AMa c r oe c onomi cAna l y s i s ” .European Commission Economic Papers, No. 147, January 2001. Roller, Lars-He ndr i ka ndLe ona r dWa ve r ma n.2001.“ Te l e c ommuni c a t i onsI nf r a s t r uc t ur ea nd Ec onomi cDe ve l opme nt ” ,The American Economic Review, September 2001, pp. 909-923. Rome r ,P.1993.“ I de aGa psa ndObj e c tGa psi nEc onomi cDe ve l opme nt . ”Journal of Monetary Economics, 32, 543-573. Rome r ,Pa ul .1990.“ Endog e nousTe c h ni c a lCha ng e . ”Journal of Political Economy 98 (October, Part 2): S71-S102. Ros t ow,W.1995.“ Le t t e r st ot heEdi t or :TheMy t hofAs i a ’ sMi r a c l e ” ,Foreign Affairs, 74(1), 183-184. RWI, and R. Gordon.2002.“ Ne w Ec onomy —AnAs s e s s me ntf r om aGe r ma nVi e w Poi nt ” . Research from a research project commissioned by the Ministry of Economics and Technology. Berlin. Sa c hs ,J e f f r e ya nd Andr e w Wa r ne r .1995.“ Ec onomi c Re f or ma nd t he Pr oc e s sofGl oba l Integrat i on” .Brookings Paper on Economic Activity 1(1995): 1-118. Sc hr e y e r ,Pa ul .2000.“ TheCont r i but i on ofI nf or ma t i on a nd Commun i c a t i on Te c hnol ogyt o Out putGr owt h:ASt udyoft heG7Count r i e s . ”STI Working Paper 2000/2. Schumpeter, Joseph. 1949. The Theory of Economic Development. 3rd edition. Harvard University Press, Cambridge, MA. Se nha dj i ,Abde l ha k.2000. ” Sour c e so fEc onomi cGr owt h:An Ext e ns i veGr owt h Ac c ount i ng Exe r c i s e . ”IMF Staff Papers. Vol. 47, No. 1, 2000 Si e g e l ,Do n a l d.1997.“ TheI mpa c tof Computers on Manufacturing Productivity Growth: A Multiple-I ndi c a t or s ,Mul t i pl eCa us e sAppr oa c h. ”Review of Economics and Statistics, pp. 68-78. Si mon,J .a ndS.Wa r dr op.2002.“ Aus t r a l i a nUs eofI nf or ma t i onTe c hnol ogya ndI t sCont r i but i on t o Gr owt h” , Research Discussion Paper RDP2002-02, Reserve Bank of Australia, Sydney, January. Sol ow,Robe r t .1956.“ ACont r i bt ui ont ot heThe or yofEc onomi cGr owt h. ”Quarterly Journal of Economics 70, No. 1: 65-94. Solvell Orjan, Goran Lindqvist, Christian Ketels. 2003. The Cluster Initiative Greenbook, Ivory Tower AB: http://www.ivorytower.se/greenbook/dlgrbk.htm.
73
St i r oh,Ke vi n.2002.“ Ar eSpi l l ove r sDr i vi ngt heNe w Ec onomy ? ”The Review of Income and Wealth, No. 1, Volume 48, March 2002, pp. 33-58. Stonema n,Pa ula nd Pa ulDi e de r e n.1994.“ Te c hnol ogy Di f f us i on a nd Publ i cPol i c y ” .The Economic Journal, 104(425), July 1994, pp. 918-930 Ta l l on,Pa ula ndKr a e me rKe nne t h.1999.“ TheI mpa c tofTe c hnol ogyonI r e l a nd’ sEc onomi c Growth and Development: Lessons forDe ve l opi ng ” .Ce nt e rf orRe s e a r c h on I nf or ma t i on Technology and Organizations, Graduate School of Management, University of California, 1999. UN. 2002. World Economic and Social Survey-Trends and Policies in the World Economy, United Nations, New York, 2002. Va nde rWi e l ,H.2001.“ Doe sI CTBoos tDut c hPr oduc t i vi t yGr owt h” ,CPBDoc ume ntNo.016, CPB Netherlands Bureau of Economic Policy Analysis, December. Vi j s e l a a r ,Foc c oa ndRona l dAl be r s .2002.“ Ne w Te c hnol og i e sa ndPr oduc t i vi t yGr owt hi nt he Euro Ar e a . ”Working Paper No. 122, European Central Bank. WDI. 2002. World Bank Development Indicators, CD-ROM, 2002. Whe l a n Ka r l .2000.“ Comput e r ,Obs ol e s c e nc e ,a nd Pr oduc t i vi t y ” ,Federal Reserve Board, Finance and Economics Discussion Series, January 2000. Wi e l d,va nde rH.2000.“ I CTI mpor t a n tf orGr owt h” .CPB Report 2000/2. WITSA. 1998, 2000, 2002. Digital Planet, World Information Technology and Services, 1998, 2000, 2002. Woo,Thy eWi ng .1997.“ Chi ne s eEc onomi cGr owt h”i nMi c ha l eFouqui na ndFrancoise Lemoine (ed.), The Chinese Economy, Economica, London, 1998. Young ,Al wy n.1992.“ ATa l eofTwoCi t i e s :Fa c t orAc c umul a t i ona ndTe c hni c a lCha ng ei nHong Konga ndSi ng a por e ” ,NBER Macroeconomics Annual 1992, Cambridge, MA: The MIT Press, 13-54. Young ,Al wy n.1995.“ TheTy r a nnyofNumbe r s :Conf r ont i ngt heSt a t i s t i c a lRe a l i t i e soft heEa s t As i a nGr owt hExpe r i e nc e ” .The Quarterly Journal of Economics, 110(3), Aug., 1995, 641-680.
74
APPENDICES Appendix A: The Country Sample: Selected Indicators for Individual Economies (Data of 2000) Population (millions)
Argentina Australia Austria Belgium Brazil Bulgaria Canada Chile China Colombia Czech Denmark Egypt Finland France Germany Greece Hong Kong Hungary India Indonesia Ireland Israel Italy Japan
37 19.2 8.1 10.3 170.4 8.2 30.8 15.2 1,262.50 42.3 10.3 5.3 64 5.2 58.9 82.2 10.6 6.8 10 1,015.90 210.4 3.8 6.2 57.7 126.9
GDP per Capita US$
PPP$
7,933 23,838 32,763 30,830 4,624 1,503 22,541 5,354 824 2,290 5,311 38,522 1,226 32,024 29,811 32,623 13,105 24,218 5,425 459 994 27,741 17,067 20,885 44,830
12,377 25,693 26,765 27,178 7,625 5,710 27,840 9,417 3,976 6,248 13,991 27,627 3,635 24,996 24,223 25,103 16,501 25,153 12,416 2,358 3,043 29,866 20,131 23,626 26,755
GDP Growth Rate 1990-95
6.3 3.3 2.0 1.5 3.1 -2.6 1.7 8.3 11.4 4.4 -1.0 2.0 3.3 -0.7 1.1 1.6 1.2 5.2 -2.4 5.1 7.6 4.5 6.4 1.3 1.4
1995-00
2.6 3.9 2.4 2.7 2.3 -1.3 3.6 4.4 7.9 0.9 0.9 2.6 5.3 5.0 2.5 1.7 3.3 3.3 3.9 5.6 0.7 9.2 3.7 1.9 1.4
Total ICT Spending as % of GDP
4.1 9.5 7.3 8 8.3 4.1 8.6 7.8 5.4 12.1 9.2 9.2 2.3 7.8 8.7 7.9 6.2 8.8 9.5 3.9 2.2 9.2 7.8 5.7 8.2
ICT Penetration per 1000 Inhabitant Fixed line telephone
213 525 467 498 182 350 677 221 112 169 378 720 86 550 579 611 532 583 372 32 31 420 482 474 586
Mobile Phone
163 447 762 525 136 90 285 222 66 53 424 631 21 720 493 586 557 809 302 4 17 658 702 737 526
PC
Internet Users
51 465 277 345 44 44 390 82 16 35 122 432 22 396 304 336 71 351 85 5 10 359 254 180 315
68 344 259 227 29 53 413 167 18 21 97 365 7 372 144 292 95 383 148 5 10 207 204 229 371
75
Appendix 1: (Continued) Population (millions)
Korea, Rep. 47.3 Malaysia 23.3 Mexico 98 Netherlands 15.9 New Zealand 3.8 Norway 4.5 Philippines 75.6 Poland 38.7 Portugal 10 Romania 22.4 Russia 145.6 Singapore 4 Slovakia 5.4 Slovenia 2 South Africa 42.8 Spain 39.5 Sweden 8.9 Switzerland 7.2 Taiwan 22.1 Thailand 60.7 Turkey 65.3 UK 59.7 United States 281.6 Venezuela 24.2 Vietnam 78.5 Sources: WBDI (2002), ITU.
GDP per Capita US$
PPP$
13,062 4,797 3,819 30,967 17,548 37,954 1,167 4,223 12,794 1,460 2,455 28,230 4,160 11,659 3,985 17,798 31,206 46,737 15,446 2,805 3,134 21,667 31,996 3,300 356
17,380 9,068 9,023 25,657 20,070 29,918 3,971 9,051 17,290 6,423 8,377 23,356 11,243 17,367 9,401 19,472 24,277 28,769 20,552 6,402 6,974 23,509 34,142 5,794 1,996
GDP Growth Rate 1990-95
7.2 9.1 1.5 2.1 3.0 3.6 2.1 2.2 1.7 -2.2 -9.5 8.7 -3.0 -0.6 0.9 1.3 0.6 -0.1 6.9 8.3 3.1 1.6 2.4 3.4 7.9
1995-00
4.7 4.6 5.4 3.5 2.3 3.0 3.5 5.0 3.5 -1.6 1.1 6.2 4.0 4.2 2.4 3.7 2.8 1.8 5.6 0.2 3.7 2.8 4.1 0.6 6.5
Total ICT Spending as % of GDP
6.6 6.8 3.2 9.6 13.7 6.8 3.9 6.1 7.1 2.3 3.6 8.6 9.7 8.2 5.3 5.2 10.4 10.4 5.6 3.6 4.8 9.2 8.2 3.9 6.4
ICT Penetration per 1000 Inhabitant Fixed line telephone
464 199 125 618 500 532 40 282 430 175 218 484 314 386 114 421 682 727 567 92 280 589 700 108 32
Mobile Phone
567 213 142 670 563 751 84 174 665 112 22 684 205 612 190 609 717 644 802 50 246 727 398 218 10
PC
Internet Users
238 103 51 394 360 491 19 69 299 32 63 483 137 276 62 143 507 500 223 24 38 338 585 46 9
403 159 28 245 217 490 27 72 250 36 21 299 120 151 56 137 456 297 281 38 31 301 339 39 3
76
Appendix B. Estimating Investment Flows B.I. WITSA/IDC ICT Spending Data The most usable data provided by the WITSA/IDC source is related to ICT spending. Figure A.II.1 describes the structure of WITSA/ICT data on ICT spending. The total spending on ICT is broken down into three main components: IT Spending, Telecommunications Spending, and Office Equipment Spending.
The IT Spending
component is composed of External IT Spending and Internal IT Spending. The External IT Spending consists of spending on IT hardware, IT software, and IT services, which are purchased by businesses, households, schools, or government agencies from external vendors. IT hardware includes servers, personal computers, workstations, data communication equipment; IT software comprises the purchases of all software products and external customization of computer program; IT services are IT-related services provided to a firm by an external agent or corporation.
The Internal IT Spending covers all IT-related expenses that cannot be attributed to an external vendor. The Telecommunications Spending component brings together expenditures by businesses, households, schools, and government agencies on not only telecommunication equipment but also telecommunication services. The Office Equipment Spending component consists of expenditures on typewriters, calculators, copiers, and other office equipment.
77
Figure B.1: Structure of WITSA/IDC Data on ICT Spending
Total ICT Spending
Telecommunications Spending
External IT Spending
Spending on Hardware
Spending on Software
IT Spending
Office Equipment Spending
Internal IT Spending
Spending on IT Services
For our purpose of estimating ICT investment flows, the WITSA data for the External IT Spending component, which includes purchase of IT Hardware, IT Software, and Telecommunications from external vendors, are the most relevant information. However, the WITSA/IDC data on ICT expenditures have the following limitations: 1) WITSA keeps track of the data only for 1992 onwards, while measuring the contribution of ICT to growth during 1990-2000 requires the ICT investment flows not only for 1990-2000 but also for many years before that period.; and 2) the spending data are aggregate and their breakdowns by purchaser group (business sector vs. non-business sector) and by type of expenditure (investment vs. non-investment) are not available.
78
To overcome the two limitations of the WITSA/IDC dataset, some significant projecting and estimating works are needed. For projecting backward the WITSA expenditure series, an OLS model with a high predictive power can be used. For estimating the business investment flows into a type of ICT asset from its WITSA/IDC expenditure data, it makes sense to resort to the pattern observed for the U.S. case, where data on business sector investment in ICT have been well documented by the Bureau of Economic Analysis (BEA) for many years. Schreyer (2000) and Daveri (2002) use the U.S. pattern to calibrate (but not in the same way33) ICT investment flows from the WITSA/IDC expenditure data, while Lee and Khatri (2003) estimate the ICT investment flows by just approximating them with the WITSA expenditure flows.
To further enhance the accuracy of the estimated investment flows in the ICT assets, my study introduces a finer estimating approach, which uses a combination of the US and global ICT market pattern and actual ICT penetration in individual economies. The details are presented in the following two sections.
B.II. Projecting backward WITSA ICT Expenditure Flows For an ICT asset type c (c could be hardware, software, or telecommunication equipment), its WITSA spending data are observed for the 50 economies under investigation over 10 years (1992-2001). These data allow one to project backwards the
33
Schreyer (2000) assumes Investment = Expenditure for hardware, =0.3 * Expenditure for telecom, but he bypasses software in his study; Daveri (2002) assumes Investment = 0.57* Expenditure for hardware, = 2.05*Expenditure for software, and = 0.33*Expenditure for telecom based on the average ratios observed for the U.S. during the 1990s.
79
spending data for the ICT asset c for earlier years, namely from 1960 to 1991. My study employs the following OLS log-log backward projection model:
[B.II-1]
ln(Eci t-1) = 0 + 1ln(Ec i t) + 2 ln(y i t-1)+ i t
where Eci t represents expenditure on ICT asset c (subscripts i and t indicate, respectively, country i and year t), yi t is GDP per capita, i t is the error term. The model specifies that, for a country i, spending on ICT asset c in year t-1 can be deduced from GDP per capita in that year and the spending on the asset c in period t. Model II-1 has a high predictive power for each ICT asset type (for predicting expenditure in 1991, R2 equals 0.98 for hardware, 0.99 for software, and 0.99 for telecom). The model, therefore, is used to project the expenditure series on the three types of ICT asset (hardware, software, and telecom) for the years prior to 1992 for the 50 countries/economics (appendix I elaborates this backward projecting process).
B.III. Examining the U.S. Pattern
In order to identify a systematic difference between the WITSA/IDC expenditure and the actual investment data flows for an ICT vintage, one can examine the ratio between the two flows for the U.S. case, where the data on ICT investment have been well recorded by the Board of Economic Analysis (BEA)34 for many years. Tables II.1.A depicts the ratios between WITSA/IDC expenditure and the BEA investment for the three ICT assets, hardware, software, and telecommunication equipment over 22 years from 1980 to
34
Business sector investment series in ICT are provided on the BEA website (www.BEA.org).
80
2001 (the WITSA/IDC expenditure data for 1980-1991 are predicted values). The results reveal the following pattern:
(i) The Consistency: The ratio of investment/spending for each type of ICT good is fairly consistent over time. There are only some notable changes, especially for hardware, between the two periods 1980-1989 and 1990-2001. The fact that the share of business investment in total expenditure on each ICT vintage, especially computer hardware, was higher during 1980-1990 than 1991-2001 is understandable because the penetration of ICT in the household sector relative to the business sector was less intensive in the earlier period than in the later. Within each of these two periods, the ratio for an ICT asset type (hardware, software, telecommunication equipment) is consistent. As a result, the 95 percent Confidence Intervals for the three ratios make up narrow ranges: (0.70, 0.82) over 1981-1990 and (0.55, 0.59) over 1991-2001 for hardware; (2.20, 2.30) and (2.01, 2.13) for software; and (0.32, 0.34) and (0.29, 0.37) for telecommunication equipment in the two periods, respectively. Because ICT products are nearly identical across countries, it is reasonable to expect that the consistency of the investment-expenditure ratios holds across countries over the two periods 1980-1989 and 1990-2001.
(ii) The Mean: The mean of the ratio of investment/expenditure is 0.76 over 1981-1990 and 0.57 over 1990-2001 for hardware, 2.25 and 2.07 for software in the two periods, respectively, and 0.33 for telecom in both periods. Using these mean values allows one to estimate investment flows into ICT assets for the U.S. with a high degree of accuracy (Table II.1.B).
81
Table B.1: Relationship between Spending and Investment by Type of ICT Assets: US Patterns Spending (WITSA, $Billions)
Year HW
SW
Investment (BEA, $Billions) TEL
HW
SW
Investment/Spending Ratio TEL
SW
TEL
26.67
6.12
87.97
17.10
12.90
29.00
0.64
2.11
0.33
1982
29.17
7.03
93.82
18.90
15.40
31.10
0.65
2.19
0.33
1983
31.94
8.10
100.14
23.90
18.00
31.90
0.75
2.22
0.32
1984
34.94
9.33
106.75
31.60
22.10
36.60
0.90
2.37
0.34
1985
38.20
10.75
113.55
33.70
25.60
39.90
0.88
2.38
0.35
1986
41.75
12.40
120.65
33.40
27.80
42.10
0.80
2.24
0.35
1987
45.61
14.30
128.07
35.80
31.40
42.10
0.78
2.20
0.33
1988
49.81
16.52
135.82
38.00
36.70
46.70
0.76
2.22
0.34
1989
54.37
19.09
143.86
43.10
44.40
46.90
0.79
2.33
0.33
59.34
22.09
152.24
38.60
50.20
47.50
1990
Average for Period 1981-1990
Mean Margin of Error (=0.05, 10 observations for 10 years) 95 percent Confidence Interval
0.65
2.27
0.31
0.76
2.25
0.33
0.06
0.05
0.01
(0.70, 0.82)
(2.20, 2.30)
(.32, .34)
1991
64.77
25.59
161.03
37.70
56.60
45.70
0.58
2.21
0.28
1992
70.74
29.72
170.40
43.60
60.80
47.80
0.62
2.05
0.28
1993
80.97
33.02
181.33
47.20
69.40
48.20
0.58
2.10
0.27
1994
89.79
37.78
195.17
51.30
75.50
54.70
0.57
2.00
0.28
1995
105.67
40.67
205.58
64.60
83.50
60.00
0.61
2.05
0.29
1996
128.87
46.80
209.59
70.90
95.10
65.60
0.55
2.03
0.31
1997
138.61
54.01
220.07
79.60
116.50
73.70
0.57
2.16
0.33
1998
159.48
65.25
231.07
84.20
140.10
81.20
0.53
2.15
0.35
1999
169.19
75.01
242.62
90.40
162.50
93.70
0.53
2.17
0.39
2000
165.47
90.97
252.33
93.30
179.40
116.60
0.56
1.97
0.46
136.05
96.56
265.95
74.20
180.40
90.60
2001
Average for Period 1991-2001
HW
1981
Mean Margin of Error (=0.05, 11 observations for 11 years)
0.55
1.87
0.34
0.57
2.07
0.33
0.02
0.06
0.04
(0.55, 0.59)
(2.01, 2.13)
(0.29, 0.37)
95` percent Confidence Interval
82
(iii) The Compatibility between the U.S. pattern and the Global Market: The information provided by the International Telecommunication Union on the global telecommunication market over 1991-2001 allows one to compare the U.S. and global pattern of the investment-to-spending ratio. Because investment includes equipment, installation, and other related services, it is reasonable to assume that the investment is about 1.3-1.4 times higher than the market value of the equipment at the global market level. Table B.2 shows the ratio between equipment and total telecom market over 1991-2001. It is interesting to note that this ratio is also consistent over the period and its adjusted value (multiplied by 1.4 to take into account the cost of additional investment) is very similar to the U.S. pattern. The investment-spending ratio at the global market level has a mean of 0.33, which is the same as in the U.S. and its 95 percent confidence interval is (0.32, 0.34). This observation implies that the U.S. pattern is relevant for the world global ICT market. Table B.2: Global Telecom Market
Year 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Market Revenue (current price, US$ Billions) Invest/Spending Ratio Total Market Equipment Services Equip/Total Ajusted (T) (E) (S) (E/T) (E/T)*1.4 523 120 403 0.23 0.32 580 132 448 0.23 0.32 605 135 470 0.22 0.31 675 158 517 0.23 0.33 779 183 596 0.23 0.33 885 213 672 0.24 0.34 946 234 712 0.25 0.35 1015 248 767 0.24 0.34 1123 269 854 0.24 0.34 1210 290 920 0.24 0.34 1232 264 968 0.21 0.30 Mean 0.24 0.33 SE 0.006 0.008 95% Confidence Interval (0.23, 0.24) (0.32, 0.34)
Source: International Telecommunication Union, 2001
83
B.IV. Estimating ICT Investment Flows
The consistency of the investment-to-expenditure ratios revealed from the U.S. pattern is expected to hold across countries because the ICT assets are nearly identical all over the world. This feature implies that it is plausible to use a constant ratio to estimate the investment flow into an ICT asset from its WITSA/IDC expenditure for each of the two time periods, 1981-1990 and 1991-2001. However, it is less plausible to assume that the mean values of the ratios observed for the U.S. are similar across economies. One of the notable reasons is the dissimilarity in the market efficiency and purchasing power parity of US$ for ICT goods across countries. A careful calibration based on the ICT penetration provides better estimates for the ratios in each individual economy. The investment in each of the three ICT asset types now can be simply computed from the WITSA spending series as Ic,a,t = c,a,t*Sc,a,t, where Ic,a,t, c,a,t, and Sc,a,t are, respectively, investment, the estimated investment-to-spending ratio and WITSA/IDC spending in year t for country c, and ICT asset a.
84
