The Impact of Broadband on the Economy: Research to Date and Policy Issues

Transcript

1 r L egu a tory & market environment International Telecommunication Union L 2012 Telecommunication Development Bureau apri Place des Nations impact of broadband CH-1211 Geneva 20 Switzerland on the economy www.itu.int S b roadband Series roadband Serie b APRIL 2012 Printed in Switzerland Telecommunication Development Sector impact of broadband on the economy Geneva, 2012 04/2012

2

3 The Impact of Broadband on the Economy: Research to Date and Policy Issues April 2012

4 This study on the impact of broadband on the economy was prepared by Dr. Raul Katz, Director, Business Strategy Research, at the Columbia Institute for Tele -Information (CITI) at Columbia University, under the direction of the BDT Regulatory and Market Environment Division (RME). The author would like to acknowledge the support of Javier Avila, Giacomo Meille and Julian Katz-Samuels, all researchers at the Columbia Institute for Tele-Information, and Fernando Callorda, consultant at Telecom Advisory Services, LLC. ITU wishes to express thanks to Jim Holmes, Incy te Consulting and Denis Villalobos from ICE, Costa Rica for their comments and advice. on broadband that are available online and free of charge This report is part of a new series of ITU reports www.itu.int/broadband . at the ITU Universe of Broadband portal: Please consider the environment before printing this report.   ITU 2012 All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU.

5 Impact of broadband on the economy Table of Contents Page iii ... Preface ... ... Foreword ... v 1 Introduction ... v ... 3 2 Economic Impact of Broadband: A Review of the Literature... 2.1 4 Contribution to economic growth ... Impact on pro 2.2 8 ductivity ... creation ... 10 2.3 Impact on job 10 2.3.1 Broadband construction effects and their counter-cyclical importance ... Broadband positive externalit ies on job creation ... 12 2.3.2 Creation of consumer surplus ... 14 2.4 2.5 Impact on firm efficiency ... 1 5 Conclusion ... 17 2.6 ... loped Countries: Case Studies ... 18 3 Economic Impact of Broadband in Deve United States: employment creation as a resu lt of the economic stimulus programme . 18 3.1 3.2 Germany: The impact of the national broadband plan on employment and economic growth ... 23 Economic Impact in Developing Countries: Case Studies ... 29 4 4.1 Latin America: Contribution to re gional economic growth ... 30 4.1.1 Multivariate Regressi on Analysis: ... 30 4.1.2 32 Case studi es ... Assessment of Broadband Economic States ... 39 Impact in Arab 4.2 Multivariate Regressi on Analysis: ... 4.2.1 39 General Impact Model ... 43 4.2.2 Case Studi es ... 44 4.2.3 ... 50 Asia Pacific ... 4.3 India: the impact of broadband on employment and economic growth ... 50 4.3.1 Malaysia: The contribution of broa 4.3.2 53 dband to economic growth ... 4.3.3 China: The relation between broa dband and economic growth ... 54 4.3.4 dband to employment growth... 56 Indonesia: The contribution of broa 8 5 5 Analysis of Case Study Results ... 6 Estimation of Broadband Gaps a nd Investment Requirement ... 61 6.1 Methodology ... . 61 6.2 The National Broadband Pl an in Germany ... 62 The National Broadband Pl an in Brazil ... 65 6.3 6.4 Conclusion ... ... 68 i

6 Impact of broadband on the economy Page 7 The Role of Public Policy and Regulation in Boosting the Development of Broadband ... 68 68 National broadband planning as a tool ... 7.1 level ... 70 7.1.1 Creating awareness at the highest national 71 Coordinating policies from different private and government entities ... 7.1.2 Developing state policies that go beyond electoral cycles ... 72 7.1.3 72 7.1.4 Building ownership and accountability of the executive branch ... Competition policies to stimulate 73 7.2 infrastructure investment ... Role of government intervention in promoting broadband deployment ... 77 7.3 Stimulating innovation in app lications and services ... 81 7.4 7.5 83 Stimulating broadband demand ... option: the broadband demand gap ... 83 7.5.1. Inhibitors of broadband ad Relevant policies aimed at addressing the broadband demand gap ... 86 7.5.2 Addressing taxation as a barrier to broadband adoption ... 89 7.6 Conclusion ... ... 92 8 8.1 broadband economic impact ... 92 The nature of the evidence of 8.2 The need to emphasize data gathering to refine impact measurement ... 93 Disaggregated data for ICT, broadband and economic indicators ... 94 8.2.1 8.2.2 Quarterly da ta ... 94 8.2.3 Range of broadband do wnload speed ... 94 8.2.4 95 Data on wireless broa dband Internet ... Data measuring: th 95 e demand gap ... 8.2.5 Variables for income endogeneity ... 95 8.2.6 The policy kit for stimulating broadband deployment and adoption ... 96 8.3 aphy ... ... 96 Bibliogr Measuring Broadband Economic Impact ... 101 Appendix A Methodologies and Data Utilized in A.1 Input/output analysis to measure mult ipliers of broadband deployment ... 101 Methodology ... 101 A.1.1 Data utilized for input/o utput analys A.1.2 103 is ... A.1.3 Advantages and disadvantages of input/output an alysis ... 104 A.2 externalities of broadband ... 104 Econometric analysis to measure 105 A.2.1 Methodology ... A.2.2 Data utilized for econometric analys is ... 106 A.2.3 Advantages and disadvantages of econometric analysis ... 107 A.3 Measuring consumer surplus of broadband ... 108 A.3.1 Data utilized in measuring the consumer surplus of broadband ... 108 A.3.2 Methodology ... 108 A.3.3 Advantages and di sadvantages ... 109 ii

7 Impact of broadband on the economy Page Appendix B Variables Utilized in Econometric An alyses ... 110 omic Impact Model in Latin America and the Appendix C Dataset utilized for the Broadband Econ Caribbean ... 116 ... Appendix D Dataset utilized for the Broadband Econ omic Impact Model in Arab States ... 119 iii

8

9 Impact of broadband on the economy Preface The past twenty years has been an extraordinary time for the development of information and ile miracle’ we have brought the benefits of ICTs communication technologies (ICTs) – and with the ‘mob within reach of virtually all the world’s people. ITU has been in the forefront of this transformational ascent and is today committed to continue to driving po sitive change in the sector and beyond. It is now time to make the next step, and to ensure that everyone – wherever they live, and whatever their circumstances – has access to the benefits of broadband. This is not just about delivering connectivity for connectivity’s sake – or even about giving people access to the undoubted benefits of social communications. It is about leveraging the power of broadband technologies – and especially mobile technologies – to make the world a better place. In 2010, ITU, in conjunction with UNESCO, launched the Broadband Commission for Digital Development – to encourage governments to implement national broadband plans and to increase access to broadband applications and services. The Commission is co-chaired by President Paul Kagame of Rwanda and Carlos Slim, President of the Carlos Slim Foundation. We have around 60 Broadband Commissioners – all top- level leaders in their field – representing government s, industry, academia and international agencies. At the Broadband Leadership Summit held in October 2011 in Geneva, the Broadband Commission recognized broadband as a critical modern infrastructure contributing to economic growth and set four clear, new targets for making broadband policy unive rsal and for boosting affordability and broadband uptake. Out-of-the-box models that promote competition, innovation and market growth are now needed to make the broadband opportunity reachable for all world citizens. At ITU, the United Nations specialized agency for ICTs and telecommunications, we are committed to playing a leading role in the development of the digital economy through extending the benefits of advances in broadband and embracing the opportunities it unleashes. The three ITU sectors – Radiocommunications, Standardization and Develo pment – are working together to meet these challenges and our collective success will be a key fact or in ensuring the provision of equitable broadband access throughout the world. The ITU Broadband Reports are one contribution towards this commitment. Dr. Hamadoun I. Touré Secretary-General, ITU v

10

11 Impact of broadband on the economy Foreword st Century, and I believe its transformative power as an Broadband has become a key priority of the 21 enabler for economic and social growth makes it an essential tool for empowering people, creating an environment that nurtures the technological and service innovation, and triggering positive change in business processes as well as in society as a whole. Increased adoption and use of broadband in the next decade and beyond will be driven by the extent to which broadband-supported services and applications are not only made available to, but are also relevant and affordable for consumers. And while the benefits of broadband-enabled future are manifest, the broadband revolution has raised up new issues and challenges. In light of these developments, ITU launches a new series of ITU Broadband Reports. The first reports in the series launched in 2012 focus on cutting edge po licy, regulatory and economic aspects of broadband. Other related areas and themes will be covered by subsequent reports including market analysis, broadband infrastructure and implementation, and broa dband-enabled applications. In addition, a series of case studies will complement the resources already made available by ITU to all its many different types of readers, but especially to ICT regulators and policy-makers. This new series of reports is important for a number of reasons. First of all, the reports will focus on nd developing countries alike. Secondly, the various topical issues of special interest for developed a reports build on ITU’s recognized expertise in th e area augmented by regular feedback from its Membership. Last but not least, this series is import ant because it provides a meaningful contribution to the work of the Broadband Commission for Digital Development. The findings of the ITU Broadband Reports will trace paths towards the timely achievement of the ambitious but achievable goals set recently by the Commission as well as provide concrete guidelines. As broadband is a field that’s growing very fast, we need to constantly build knowledge for our economies and societies to thrive and evolve into the future. For these reasons, I am proud to inaugurate this first series of the ITU Broadband Reports and look forward to furthering ITU’s work on the dynamic and exciting broadband ecosystem. Brahima Sanou Director, ITU Telecommunication Development Bureau vii

12

13 Impact of broadband on the economy 1 Introduction that enables high-speed transfer of data, is The diffusion of broadband, defined as the technology inextricably linked to the emergence of the Internet. Wh ile at its initial stages the Internet was primarily 1 accessed through dial-up means , consumer and enterprise demand prompted the development of eeds. As a result, starting around the mid-1990s, technologies that facilitated access at higher sp telecommunications and cable TV companies began offering services that significantly enhanced the experience of Internet use. Investment and adoption soared around the world. By 2009, there were 2 1.8 billion Internet users and 471 million broadband subscriptions . Between 2004 and 2010, telecommunications and cable TV companies in the United States invested over USD 97.7 billion in 3 to industrialized nations. In some . Broadband capital is not only restricted broadband deployment increasing exponentially. Chinese companies have emerging countries broadband investment is also 4 invested USD 7.44 billion in broadband since 2009 , while Malaysian operators invested USD 1.6 billion 5 . since 2009 With these amounts of capital being dedicated to th e technology, it is natural that policy makers and researchers in the social sciences have begun to analyse the economic and social impact of broadband. In researching the economic contribution of information fact, social scientists and policy makers had been and communication technologies for quite a while. The first analyses of the impact of fixed telephone 6 . Ever since, density on economic growth were conducted in the mid-1970s by World Bank researchers enhancements both in the quality of data and so phistication of econometric tools have yielded continuous improvement in tackling the question of economic impact of telecommunications. searchers. First, its deployment has proceeded at Broadband, however, represents a new challenge for re an incredibly fast pace. Within 12 years, broadband has been adopted by over 62 per cent of households 7 in the United States, 80 per cent in the Netherlands and 95.9 per cent in Korea (ITU, 2010; OECD, 2010) . Consequently, the length of time series data of broa dband adoption is considerably shorter than for voice telecommunications. Second, only the countries that have understood early on its economic potential have proceeded to collect statistics at the beginning of the diffusion process. Third, since broadband is an access technology for data communications, it only has an economic effect in combination with the adoption of information technology, and the implementation of organizational and process changes in enterprises. In sum, because broadband has been deployed in such a short time span and it is an enabler of remote information technology access, it has repr esented a substantial research challenge. The primary challenge, though, remains the lack of disaggregated datasets that allow to quantitatively establish the conditions under which broadband has an economic effect. These methodological challenges rendered the broa dband policy making process quite complex. It is difficult to ascertain precisely if broadband contributes to economic growth or it is deployed as a result of growing development. This problem risks repeating the debate started when economists started looking at the impact of computing. As expected, the original results were not conclusive. Robert Solow, the Nobel Economics laureate from MIT, concluded at the time “you can see the computer age everywhere 1 Dial up technology refers to Internet access over conventional voice telephone lines at speeds that do not exceed the 56 Kbps. 2 Source: Internet World Statistics (September 2009). In June 2010, the number of Internet users had reached 1.96 million. 3 See Atkinson and Schultz (2010). 4 Source: Sinopac (2009). Taiwan Research. 5 See AM Research (2010). Telecommunications: CAPEX risk from escalating competition in broadband. 6 See Jipp (1963). 7 In addition, in the last five years, the combination of wireless technology and broadband service is taking service d to the individual user. adoption from the househol 1

14 Impact of broadband on the economy 8 but in the production statistics” . His conclusion kicked off a sceptical body of research and theory. In particular, Paul Krugman, another Nobel laurate, st ated in the early 1990s that “either the technology 9 isn’t all it's cracked up to be, or we haven’t yet seen the impact of the new technology on the economy” , while Robert Gordon concluded that computers made only a small contribution to productivity because 10 . “there is something wrong with computers” Luckily enough, the availability of larger data sets at the beginning of the 21st century allowed researchers to more precisely estimate the effects of computing. This led to the development of a new theory based on growth accounting economics that could not only pinpoint the economic impact of information 11 technology, but also identify differential effects by region of the world . For example, a study relating labour productivity growth on ICT investments on an industry level concluded that the faster productivity growth in the US compared to EU countries can be attributed to a larger employment share in the ICT 12 . No one doubts today that producing sectors and a faster growth in industries that intensively use ICT computing in particular and ICT in general have si gnificantly contributed to economic growth in the industrialized world during the 1990s and 2000s. The evidence on broadband is not quite conclusive, however. As detailed above, the study of the economic effects of broadband presents several methodological challenges. Research has confronted these challenges by proceeding along three avenues. In the first place, macro-economic research 13 grounded on the Harvard economist Robert Barro's endogenous technical change model has analysed the aggregate impact of broadband on economic developm ent. In this case the guiding question is what is the contribution of broadband to GDP growth, productivity and employment? The second avenue has researched the impact of broadband from the microe conomic perspective. It is conducted at the firm level and emphasizes the contribution of broadband to business process efficiency and sales growth. The key issue here is to understand the return on broadband and IT investment at the firm and sector level. ion from a qualitative perspective, choosing the case The third school of thought tackles this last quest study as its primary analytical tool. Nevertheless, the evidence accrued by these three bodies of research is beginning to support the impact. However, when hypothesis that broadband has an important economic comparing findings across research, a number of caveats need to be raised. First, broadband exhibits a higher contribution to economic growth in countries that have a higher ad option of the technology (this could be labelled the 14 ). Second, broadband has a stronger productivity impact in “critical mass” or “return to scale” theory” sectors with high transaction costs, such as financial se rvices, or high labour intensity, such as tourism and lodging. Third, in less developed regions, as po stulated in economic theory, broadband enables the adoption of more efficient business processes and lead s to capital-labour substitution and, therefore loss of jobs (this could be labelled the “productivity shoc k theory”). Fourth, the impact of broadband on small and medium enterprises takes longer to materialize due to the need to restructure the firms' processes and labour organization in order to gain from adopting the technology (this is called “accumulation of intangible capital”). Finally, the economic impa ct of broadband is higher when promotion of the technology is combined with stimulus of innovative businesses that are tied to new applications. In other words, the impact of broadband is neither automatic nor homogeneous across the economic system. This 8 See Solow, 1987. 9 See Krugman, 1993. 10 See Gordon, 1998. 11 See Jorgenson et al., 2006a, and van der Ark et al., 2002. 12 See van der Ark et al., 2003. 13 See Barro, 1991. 14 According to this theory, driven by network effects, the economic impact of broadband increases exponentially with the r references in Sections 2.2 and 8.1. penetration of the technology. See furthe 2

15 Impact of broadband on the economy emphasizes the importance of implementing public policies not only in the areas of telecommunications regulation, but also in education, economic development and planning, science and technology, and others. The purpose of this study is threefold. On one hand, it presents the evidence generated by the different bodies of theory regarding the economic impact of broadband. The purpose is not only to summarize but also to present the complexities and conditions under which broadband has an impact. On the other hand, it reviews the results of research the author has conducted across the world measuring the impact of broadband on economic growth and employment crea tion. In this context, it presents a methodology for calculating the investment necessary to implemen t national broadband plans. Finally, it outlines the public policy implications, which can stimulate deployment and maximize the impact of the technology. 2 Economic Impact of Broadband: A Review of the Literature The economic impact of broadband manifests itself through four types of effects (see Figure 1). The first effect results from the construction of broadband networks. In a way similar to any infrastructure project, the deployment of broadband networks crea tes jobs and acts over the economy by means of multipliers. The second effect results from the “spill- over” externalities, which impact both enterprises and consumers. The adoption of broadband within firms leads to a multifactor productivity gain, which in turn contributes to growth of GDP. On the other hand, residential adoption drives an increase in household real income as a function of a multiplie r. Beyond these direct benefits, which contribute to GDP growth, residential users receive a benefit in terms of consumer surplus, defined as the difference between what they would be willing to pay for broadband service and its price. This last parameter, while not being captured in the GDP statistics, can be significant, insofar that it represents benefits in terms of enhanced access to information, entertainment and public services. Figure 1: Broadband economic impact Consumer Residential surplus Penetration Direct Household benefits income Enterprise Broadband penetration deployment Total Factor productivity Contribution Investment in to GDP growth infrastructure deployment Source: Author Research aimed at generating hard evidence regarding the economic impact of broadband is fairly recent. The results of the research and the evidence generated so far fall into five areas: 1. Contribution to economic growth (“positive externalities”). 2. Contribution to productivity gains. 3. Contribution to employment and output of br oadband deployment (“countercyclical effect”). 4. Creation of consumer surplus. 5. Improvement of firm efficiencies. 3

16 Impact of broadband on the economy This section presents the research conducted to date on the economic impact of broadband. In reviewing the literature, it will become apparent that there is no single approach to assess broadband's economic ed so far (input / output analysis, econometric contribution. Each methodology that has been utiliz modelling, measurement of consumer surplus, and microeconomics case studies) will be reviewed and the robustness of the evidence generated will be assessed. In reviewing the methodologies, it will also become apparent that the overarching condition guiding the selection of one approach over another is driven primarily by data availability. 2.1 Contribution to economic growth Broadband technology is a contributor to economic gr owth at several levels. First, the deployment of broadband technology across business enterprises improv es productivity by facilitating the adoption of inventory optimization, and streamlining of supply more efficient business processes (e.g., marketing, chains). Second, extensive deployment of broa dband accelerates innovation by introducing new consumer applications and services (e.g., new form s of commerce and financial intermediation). Third, broadband leads to a more efficient functional deploy ment of enterprises by maximizing their reach to labour pools, access to raw materials, and consumers, (e.g., outsourcing of services, virtual call centres.) Research aimed at generating hard evidence regarding the economic impact of broadband is fairly recent. The review of the research indicates that there are multiple approaches to estimate the economic impact of broadband, ranging from highly sophisticated econometric techniques to qualitative micro-level case studies. Not all approaches are suitable to all situations. The choice of analytical techniques will be driven by the availability of data and ty pe of effect to be analysed. The study of the impact of broadband on economic growth covers numerous aspects, ranging from its aggregate impact on GDP growth, to the different ial impact of broadband by industrial sector, the increase of exports, and changes in intermediate de mand and import substitution. While the research on d its positive impact, it has also yielded results the contribution of broadband to GDP growth has confirme e analyses have primarily focused on OECD countries that vary widely. Constrained by data availability, th (generally Western Europe and North America) a nd states in the United States (see Table 1). Table 1 – Research results of br oadband Impact on GDP growth Country Authors – Institution Data Effect Not statistically significant results United States Crandall et al. (2007) – 48 States of US for the Brookings Institution period 2003-2005 Thompson and Garbacz 46 US States during the A 10% increase in broadband (2008) – Ohio University period 2001-2005 penetration is associated with 3.6% increase in efficiency OECD Czernich et al. (2009) – A 10% increase in broadband 25 OECD countries between 1996 and 2007 University of Munich penetration raises per-capita GDP growth by 0.9-1.5 percentage points Koutroumpis (2009) – 2002-2007 for 22 OECD An increase in broadband Imperial College countries penetration of 10% yields 0.25% increase in GDP growth High Income et al. (2009) – World 1980-2002 for 66 high Qiang 10% increase in broadband Bank Economies penetration yielded an additional income countries 1.21 percentage points of GDP growth Low & Middle 10 % increase in broadband 1980-2002 for the Qiang et al. (2009) – World Bank remaining 120 countries income penetration yielded an additional economies (low and middle income) 1.38 in GDP growth Source: Author 4

17 Impact of broadband on the economy lude that broadband penetration has an impact on GDP As the data in Table 1 indicates, most studies conc ribution appears to vary widely, from 0.25 to growth. However, one observes that such a cont 15 1.38 per cent for every increase in 10 per cent of penetration . Explanations for this variance are manifold. Clearly, some of the discrepancies come from the usage of different datasets as well as model specifications. However, in some cases differences may be due to methodological shortfalls. For one, at very high leve ls of data aggregation, such as country data, the econometric models do not account for the wide discrepancy between regions that are caused by fixed effects. For example, a large portion of the variance in the study by Qiang et al. (2009) is explained by dummy variables for Africa and Latin America (nearly ten times as much as the estimate given by Barro (1991) in the original formulation of the model). This probably suggests an over-estimation of impact of broadband on GDP growth. It also justifies the need to rely on the differentiation of fixed effects and to conduct the analysis at lower levels of aggregation su ch as states and, where data is available, even counties or administrative departments. Many of the problems identified stem from data av ailability, since researchers lack a host of useful variables and must work at high levels of aggregatio n. However, despite the degree of discrepancies, the research consistently concludes that broadband has a significant positive effect on GDP growth. In addition to measuring the aggregate economic impa ct at the macro level, research on the economic impact of broadband has focused on the specific processes that underlie this effect. So far two questions have been studied in detail: 1. Does the economic impact of broadband increase with penetration and can we pinpoint a saturation threshold when decreasing returns to penetration exist? 2. What explains the lagged effect of broadband on the economy? A critical element of the evolving theoretical framework of network externalities of broadband is the impact infrastructure penetration levels may have on output. Is there a linear relationship between broadband adoption and economic growth? Or are we in the presence of a more complex causality effect? The “critical mass” findings of research of the impact of telecommunications on the economy, indicates that the impact of broadband on economic growth may only become significant once the et al. (2006) contend that the adoption of the platform achieves high penetration levels. However, Gillett relation between penetration and economic impact should not be linear “because broadband will be adopted (...) first by those who get the greatest benefit (while) late adopters (...) will realize a lesser benefit” (pp. 10). With both points of view in mind, it would appear that the strength of the relationship is highest once the technology has achieved a certain critical mass but before it reaches saturation (see Figure 2). Theoretically, it appears that there is a non-linear (or inverted U shape) relationship between broadband penetration and output. At low levels of broadband pe netration, we believe the impact of broadband on the economy is minimal due to the “critical mass” conc ept. According to the “return to scale” theory, the impact of telecommunications infrastructure on the economic output is maximized once the infrastructure reaches a critical mass point, generally associated with levels of penetration of developed countries. As a result, we initially observe increasi ng returns on growth (see Roeller and Waverman, 2001; Shiu and Lam, 2008). While Roeller and Waverman (2001) associate “critical mass” with near universal voice telephony penetration, we are starting to identify this phenomenon for broadband as well. 15 Or .36% if we make the standard assumption that 1% increa se in productivity or efficiency results in 1% increase in GDP in Thompson and Garbacz (2008). 5

18 Impact of broadband on the economy Figure 2: Impact of broadband on output over diffusion process Saturation point + Critical Mass threshold INCREASE EMPLOYMENT - - + BROADBAND PENETRATION Source: Adapted from Katz (2008a). The implication of this finding for developing countries is significant. Research points to the fact that in impact, broadband needs to reach high levels of order to achieve an important level of economic penetration. For example, Koutroumpis (2009) foun d that for OECD countries the contribution of broadband to OECD economic growth increased with penetration (see Figure 3). pact of broadband on GDP growth Figure 3: OECD: Percentage of im 0.025 0.0045 0.004 High penetration 0.02 0.0035 •Denmark, Norway, Netherlands, 0.003 Sweden, 0.015 0.0025 Switzerland •Average 0.002 growth contribution to GDP 0.01 growth: 0.023 0.0015 0.001 0.005 Cluster average impact on growth Country Average % Impact of BB on 0.0005 0 0 Low penetration % 3% 2 4% 7% 3 30% 25% 24% 2 22% 21% 17% 16% 1 •Greece, Portugal, Italy, Broadband Penetration (2007) New Zealand, Austria, Hungary, Spain, Ireland Medium penetration Source: adapted from Koutrompis (2009) •Average contribution to •Germany, France, Japan, Belgium, UK, GDP growth: 0.008 Australia, US, Canada, Luxemburg •Average contribution to GDP growth: 0.014 Source: OECD As seen above, according to Koutroumpis' research , in countries with low broadband penetration (under 20%), an increase of 1 per cent in broadband adoption contributes to 0.008 per cent of GDP growth, while in countries with medium penetration (between 20% and 30%), the effect is of 0.014 per cent and in 6

19 Impact of broadband on the economy the impact of 1 per cent adoption reaches 0.023. The countries with penetration higher than 30 per cent, quite significant. Unless emerging economies do not implication of this finding for developing countries is strive to dramatically increase their penetration of broadband, the economic impact of the technology will be quite limited. At the other end of the penetration process, some authors have already pointed out a potential 16 .They find that beyond a certain adoption level (not specified, as of yet), the effect of “saturation” effect broadband on the economy tends to diminish. For example, Atkinson at al. (2009) point out that network externalities decline with the build out of networks and the maturation of technology over time. There is evidence that supports this argument. It has been demo nstrated in diffusion theory that early technology adopters are generally those who can elicit the higher returns of a given innovation. Conversely, network externalities would tend to diminish over time because those effects would not be as strong for late adopters. 17 added dummy variables to account for 10 per To test the saturation hypothesis, Czernich (2009) et al. cent and 20 per cent broadband penetration to th eir models. They found that 10 per cent broadband penetration has a significant impact on GDP per capit a: between 0.9 and 1.5 percentage points. Similarly, in their study of the state of Kentucky, Shideler et al. (2007) estimated that employment growth is highest around the mean level of broadband saturation at the county level, driven by the diminishing returns to scale of the infrastructure. According to this, a cr itical amount of broadband infrastructure may be needed to sizably increase employment, but once a community is completely built out, additional broadband infrastructure will not further affect employment growth. The saturation evidence still needs to be carefully test ed particularly in terms of what the optimal point is beyond which broadband exhibits decreasing economic returns. For example, in a study conducted in Germany by this author (discussed later), it was not possible to identify a saturation point for broadband 18 . Furthermore, even if that were to be found confirming evidence of saturation with regard penetration to contribution to GDP or employment creation, that would not put into question the need to achieve universal broadband in terms of the social benefits it yields to end users. Most of the statistical research on the economic impa ct on GDP growth is performed using regressions of cross-lagged indicators (in other words, an increase in broadband deployment in year one is found to have an impact two or more years later). This approach is common in the assessment of economic impact of infrastructure (given that no deployment has an immediate economic impact.) However, the premise underlying the lagged effects assumption comprises a more complicated process of broadband adoption. Management science has studied how technology is adopted by individual firms and how it impacts firm productivity. First of all, purchasing ICT is not the only requirement for improving productivity. In fact, both management and economics literature have shown that it is necessary to modify business practices in order for information technology impact firm effi ciency. Accordingly, independently from the pace at and productivity is driven by what has been called which ICT is being adopted, the impact on efficiency 19 . “accumulation of intangible capital” This effect that has been studied for ICT exists in the case of broadband as well. Technology adoption is only the first step in the assimilation of busine ss processes that yield improvement in productivity. To sum up, in order to fully increase efficiency and output, the adoption of information and communication technologies by enterprises require s the introduction of a number of processes and organizational changes. These changes, as well as training and other cultural factors, (such as 16 Gillett et. al, 2006. 17 Op. cit. Above. 18 See Katz et al . (2010a). 19 See Basu & Fernald (2006). 7

20 Impact of broadband on the economy organizational transformation), are referred to as the entrepreneurial spirit, willingness to take risks in an accumulation of intangible capital. Broadband does not in itself have an economic impact. It represents an enabler for the adoption of e-business processes that result in increased efficiency (such as streamlined access to raw materials and management of the supply ch ain, or better market access). Intangible capital accumulation and the adoption of e-business proces ses delay the full economic impact of broadband. Lagged effects are neither uniform nor permanent. They are most marked at the start of broadband deployment. It stands to reason that once firms have undergone the transformation required to enable the full impact of broadband, further deployment of the technology should have an immediate impact. Finally, van der Ark et al. (2002) and Gulton et al. (2003) note that institutiona l variables such as labour market regulation could also have a significant impa ct on models that link broadband and productivity. The public policy implications of th is effect cannot be understated. To achieve full economic benefit of broadband deployment, governments need to emphasize the implementation of training programmes and, in the case of SMEs, offer consulting servic es that help firms capture the full benefit of the 20 technology . 2.2 Impact on productivity It is logical to assume that productivity of information workers, defined as the portion of the economically active population whose working function is to proc ess information (administrative employees, managers, teachers, journalists) depends directly on the investment in ICT capital (and particularly broadband). The 21 have, in fact, concluded that the larger the per cent of the workforce studies conducted by this author dedicated to information generation and processing is, the higher the proportion of capital stocks invested in the acquisition of ICT infrastructure (see Figure 4). Figure 4: Information workers and ICT investment 40% Info workers = 0.6123 IT capital - 0.0733 35% 2 R = 0.6403 30% 25% 20% 15% Fixed capital 10% 5% IT Capital as a Percentage of Total 0% 70% 30% 20% 10% 40% 0% 60% 50% Information Workers as a percent of the economically active population Data for information workforce was derived from ILO statistics while IT Capital was sourced from Kaplan Note: (2001). Source: Adapted from Katz (2009b). 20 For additional details, see section 7.5. 21 See Katz, 2009b. 8

21 Impact of broadband on the economy t indicate the existence of a direct relationship Figure 1 and the corresponding regression coefficien and IT capital investment in a given economy: as existing between the amount of information workers expected, the larger the proportion of information workers in a given the economy, the more capital is invested in information technology. between ICT and productivity? In his economics How can one theoretically explain the relationship dissertation at Harvard University (1982), Charles Jonscher raised the hypothesis that if we can measure the micro-economic impact of ICT on firm productivity, then we should also be able to link the growth in informational occupations and the adoption of technology to improve their productivity at the macroeconomic level. This is what is depicted in Figure 5. Figure 5: Causality model: ICT innovation and diffusion is driven by the growth of information workforce Increasing complexity ECONOMIC of DEVELOPMENT production processes Reduction of WORKFORCE At some point, the uncertainty in SPECIALIZATION information information workforce becomes handling a bottleneck in the GROWTH OF system of INFORMATION WORKFORCE production Productivity NEED TO ADOPT ICT increase (first TO INCREASE effect) PRODUCTIVITY OF INFORMATION WORKERS Productivity increase (second effect) Source: Adapted from Katz (2009c). According to this causality framework, economic grow th logically leads to increasing complex production processes. In turn, complexity in production processes results in increasing the functional complexity within firms (e.g. more inputs to be combined, more steps to be scheduled in a timely manner, more als and with buyers of the end product). The first interactions occurring with suppliers of raw materi response of economic organizations to this effect is the creation of “information workers”—labourers ation for purposes of organizing the production of whose primary function is the manipulation of inform goods. At some point, however, information-processing workers become a bottleneck in the economic system. They cannot grow forever because this process reduces the overall availability of resources in other occupations. Furthermore, when information wo rkers become a large proportion of the workforce, ttleneck itself. In other words, there is a limit to the complexity of information processing becomes a bo the possibility of manually storing, transferring and processing the growing amounts of information. This is where information and communication technologies come in. Their development and adoption is aimed at increasing the productivity of information worker s and addressing this bottleneck. The availability of computing and communications allows firms (and their information workers) to be more productive in specific component performing this important their manipulation of information. Broadband is a productivity enhancement. For example, research on the impact of broadband on productivity has successfully identified positive (2009) determined the economic effect of broadband on the GDP effects. For example, Waverman et al. of 15 OECD nations for the time period of 1980 to 2007. These included 14 European countries and the et al. United States. By relying on an augmented production function derived from Waverman (2005), the authors specified two models: a production function and a hedonic function for ICT capital stocks. 9

22 Impact of broadband on the economy Broadband impact on the productivity of the more developed nations in the sample was found to 22 . In other words, Waverman estimated be 0.0013 and was statistically significant at the 5 per cent level that for every 1 per cent increase in broadband penetr ation in high and medium impact income countries, productivity grows by 0.13 per cent. In another do cument, the authors commented upon the productivity ly low ICT penetration (Greec e, Italy, Portugal, Spain effect in the countries of their sample with relative and Belgium.). They found that broadband impact on productivity was nil, which indicated the high 23 . In other words, for broadband to have an impact on adoption costs, and critical mass thresholds 24 productivity, the ICT eco-system has to be sufficiently developed . It would appear, therefore, that in developed countries with high broadband penetratio n, the technology has an impact on aggregate productivity levels. 2.3 Impact on job creation This section will review the evidence regarding the impact of broadband in terms of job creation. Differences will be made between the research focused on measuring the impact of broadband deployment programmes (e.g. counter-cyclical impact of broadband network construction) and the spill- over effect that broadband can have in terms of ge the economy once it is nerating employment across being deployed. As in the prior chapter, a section will focus on specific effects, such a differential impact by industrial sector and/or regions. 2.3.1 Broadband construction effects and their counter-cyclical importance Broadband network construction affects employment in three ways. In the first place, network construction requires the creation of direct jobs (s uch as telecommunications technicians, construction workers, and manufacturers of the required telecomm unications equipment) to build the facility. In addition, the creation of direct jobs has an impact on indirect employment (such as upstream buying and selling between metal and electrical equipment manufacturing sectors). Finally, the household spending based on the income generated from the direct and indirect jobs creates induced employment. et al. Four national studies have estimated the impact of network construction on job creation: Crandall (2003), Atkinson (2009), Liebenau et al. (2009), and in prior research carried out by the author (Katz et al. et al. , 2008). They all relied on input-output matrices and assumed a given amount of capital investment: USD 63 billion (needed to reach ubiquitous broadband service in the United States) for Crandall et al. (2003), CHF 13 billion for Katz et al. (2008b) (to build a national multi-fibre network for Switzerland), USD 10 billion for Atkinson (2009) (as a US broadband stimulus) and USD 7.5 billion for Liebenau et al. et al. (2009) (needed to complete broadband deployment in the United Kingdom) (see Table 2). Since these studies were triggered by the considerat ion of countercyclical plans devised to face the economic crisis, they tend to focus primarily on gaug ing the ability of broadband jobs to create jobs. All studies calculate multipliers, which measure the total employment change throughout the economy resulting from the deployment of a broadband network. Multipliers are of two types. Type I multipliers measure the direct and indirect effects (direct plus i ndirect divided by the direct effect), while Type II multipliers measure Type I effects plus induced effect s (direct plus indirect plus induced divided by the direct effect). Cognizant that multipliers from one ge ographic region cannot be applied to another, it is useful to observe the summary results for the multip liers of the four input-output studies (see Table 3): 22 The original regression yielded a coe fficient of 0.0027 for the 2/3 more developed countries in the sample and negative effect for the lower third. A negative effect did not make se the effect for the lower third nse so the authors constrained to zero. At that point the coefficien t for the full sample moved to 0.0013. 23 See Waverman, 2009. 24 For example, Waverman et al. estimated that in the United States broadband penetration contributed approximately to 0.26% per annum to productivity growth, resulting in 11 additional cents per hour worked (or USD 29 billion per year). 10

23 Impact of broadband on the economy Table 2 – Broadband impact on job creation Authors – Objective Results Country Institution (*) • Creation of 140,000 jobs per year Estimate the employment impact United States et al. Crandall over ten years of broadband deployment aimed (2003) – at increasing household adoption Brookings Total jobs: 1.2 million (including • from 60% to 95%, requiring an Institution 546,000 for construction and investment of USD 63.6 billion 665,000 indirect) Atkinson et al. Estimate the impact of a USD • Total jobs: 180,000 jobs-year 10 billion investment in (including 64,000 direct and 116,000 (2009) – ITIF broadband deployment indirect and induced Switzerland Katz Total jobs: 114,000 over four years Estimate the impact of deploying a • et al. national broadband network (2008b) – CITI (including 83,000 direct and 31,000 indirect) requiring an investment of CHF 13 billion United Total jobs: 211,000 jobs-year Liebenau et al. • Estimate the impact of investing Kingdom USD 7.5 billion to achieve the (including 76,500 direct and 134,500 (2009) – LSE ” Plan target of the “Digital Britain indirect and induced) (*) Note: ITIF: Information Technology and Innovation Foundation Columbia Institute for Tele-Information CITI: LSE: London School of Economics Source: Author. Table 3 – Employment multiplier effects of studies relying on input-output analysis Studies Type I Type II Country et al. (2003) N.A. 2.17 United States Crandall et al. (2009) N.A. 3.60 Atkinson et al. 1.83 3.42 (2009) Katz Katz et al. (2008) 1.38 N.A. Switzerland N.A. Liebenau et al. (2009) 2.76 United Kingdom Germany Katz (2010) 1.45 1.92 et al. et al. et al. Note: Crandall (2009) do not differentiate between indirect and induced effects, (2003) and Atkinson therefore we cannot calculate Type I multipliers; Katz el (2008) did not calculate Type II multiplier because induced effects were not estimated . Source: Compiled by the author. According to the sector interrelationships depicted above, a European economy appears to have lower indirect effects than the US. Furthermore, the decomposition also indicates that a relatively important job creation induced effect occurs as a result of household spending based on the income earned from the direct and indirect effects. While input-output tables are a reliable tool for predic ting investment impact, two words of caution need to be given. First, input-output tables are static models reflecting the interrelationship between economic sectors at a certain point in time. Since those interactions may change, the matrices may lead us to overestimate or underestimate the impact of networ k construction. For example, if the electronic equipment industry is outsourcing jobs overseas at a fast pace, the employment impact of broadband deployment will diminish over time and part of the counter-cyclical investment will “leak” overseas. Second, it is critical to break down employment effe cts at the three levels estimated by the input-output oadband deployment. Having said that, all these effects table in order to gauge the true direct impact of br 11

24 Impact of broadband on the economy have been codified and therefore, with the caveat of the static nature of input-output tables, we believe that the results are quite reliable. 2.3.2 Broadband positive exte rnalities on job creation Beyond the employment and output impact of networ k construction, researchers have also studied the impact of network externalities on employment variously categorized as “innovation”, or “network 25 effects” . The study of network externalities resulting from broadband penetration has led to the identification of numerous effects: New and innovative applications and services, such as telemedicine, Internet search, • 26 . e-commerce, online education and social networking 27 • New forms of commerce and financial intermediation . 28 . • Mass customization of products 29 • Reduction of excess inventories and optimization of supply chains . 30 • Business revenue growth . 31 Growth in service industries • Most of the research regarding the impact of broadband externalities on employment has been conducted using US data. There are two types of studies of these effects: regr ession analyses and top down multipliers. The first ones attempt to identify the macro-economic variables that can impact 32 employment , while the second ones rely on top-down network effect multipliers. (2007), et al. Among the econometric studies of employment impact, are Gillett et al. (2006), Crandall Shideler et al. (2007) and Thompson and Garbacz (2008). The evidence regarding broadband employment externalities appears to be quite conclusive (see Table 4). Again, the impact of broadband on employment creati on appears to be positive. However, as the data indicates, the impact on employment growth varies widely, from 0.2 per cent to 5.32 per cent for every increase in 1 per cent of penetration. There are several explanations for this variance. As Crandall indicated, the overestimation of employment creation in his study is due to employment and migratory trends, which existed at the time and biased the sample data. In the case of Gillett et al. (2006), researchers should be careful about analysing local effects because zip codes are small enough areas that cross-zip code commuting might throw off estimates on the effect of broadband. For example, increased wages from broadband adoption in one zip code would probably raise rent levels in neighbouring zip codes prompting some migration effects. Finally, the wide range of effects in the case of Shideler et al. (2007) is explained by the divergent effects among industry sectors. 25 et al. , 2009. See Atkinson 26 Op. cit. 27 Op. cit. 28 Op. cit. 29 Op. cit. 30 , 2006. et al. , 2002; Gillett et al. See Varian 31 See Crandall et al. (2007). 32 t differentiate between construction and spill-over effects. In general, studies based on regression analysis do no 12

25 Impact of broadband on the economy ct on employment in the United States Table 4 – Research results of broadband impa Data Effect Authors – Institution Crandall et al. (2007) – 48 states for the period For every 1% point increase in broadband penetration 2003-2005 Brookings Institution in a state, employment is projected to increase by 0.2 to 0.3% per year “assuming the economy is not already at 'full employment'” Thompson and Garbacz 46 states during the period Positive employment generation effect varying by (2009) – Ohio 2001-2005 industry University Gillett et al. (2006) – Broadband availability increases employment by 1.5% Zip codes for the period 1999-2002 MIT Shideler et al. (2007) – Disaggregated county data An increase in broadband penetration of 1% contributes Connected Nation for state of Kentucky for to total employment growth ranging from 0.14% to 2003-4 5.32% depending on the industry Source: Author. Beyond regression studies, “network effect” multipliers have been used to assess the impact of broadband on job creation in a top down fashion. Within this group, key studies are Pociask (2002), Atkinson (2009) studies relied (2009) and Liebenau et al. (2009). Pociask (2002) and Atkinson et al. et al. is applied to the network construction employment on an estimated “network effect” multiplier, which estimates. For example, Pociask relied on two multiplier estimates (an IT multiplier of 1.5 to 2.0 attributed ed to Microsoft) and calculated an average of 4.1. to a think tank and another multiplier of 6.7, attribut (2009) derived a multiplier of 1.17 from Crandall et al. (2003). Though the top- et al. Similarly, Atkinson ct, it does not have a strong theoretical basis. down approach allows estimation of the broadband impa Network effects are not built on interrelationships between sectors. They refer to the impact of the technology on productivity, employment and innovation by industrial sector. The methodological implications of these studies are that in order to properly measure the contribution datasets that include time series for employment of broadband to job creation, it is advisable to have level, broadband penetration, and related human capita egated level, such as l statistics at a disaggr 33 . counties, departments, or administrative district Like the relationship between broadband and GDP gr owth, the contribution of broadband to employment is also conditioned by a number of special effects. Studies have particularly focused on two specific questions: 1. Does the impact on employment differ according to industry sector? 2. Is there a decreasing return in employment generation linked to broadband penetration? As with GDP, the spill-over employment effects of broadband are not uniform across sectors. According to Crandall et al. (2007), the job creation impact of broadband tends to be concentrated in service industries, (e.g., financial services, education, health care, etc.) although the authors also identified a positive effect in manufacturing. In another study, Shideler et al. (2007) found that, for the state of Kentucky, county employment was positively related to broadband adoption in the following sectors. The only sector where a negative relationship was found with the deployment of broadband (0.34% – 39.68%) was the accommodations and food services industry . This may result from a particularly strong capital/labour substitution process taking place, whereby productivity gains from broadband adoption yields reduced employment. Similarly, Thompson and Garbacz (2008) conclude that, for certain industries, 33 and Chile included in Sections 3.2 and 4.3. See examples in case studies of Germany 13

26 Impact of broadband on the economy 34 “there may be a substitution effect between broadband and employment” . It should therefore be considered that the productivity impact of broadba nd can cause capital-labour substitution and may result in a net reduction in employment. in the case of rural econ This particular effect has been analysed by this author omies of the United States. In particular, it was found that, within rural counties, broadband penetration contributes to job creation in financial services, wholesale trade, and health sectors. This is the result of enterprise relocation enabled by broadband, which benefits primarily urban communities in the periphery of metropolitan areas (Katz et al. 2010d). In summary, research is starting to pinpoint different employment effects by industry sector. Broadband may simultaneously cause labour creation triggered by innovation in services and a productivity effect in labour intensive sectors. Nevertheless, we still lack a robust explanation of the precise effects by sector n that the sectoral composition varies by regional and the specific drivers in each case. However, give economies, the deployment of broadband should not have a uniform impact across a national territory. To sum up, some researchers have found a decreasing impact of broadband on employment. While Gillett et al. (2006) observed that the magnitude of impact of broadband on employment increases over time, they also found that the positive impact of broadband on employment tends to diminish as penetration increases. This finding may support the existence of a saturation effect. Coincidentally, Shideler et al. (2007) also found a negative statistically significant relationship between broadband saturation and employment generation. This would indicate that at a certain point of broadband deployment, the capability of the technology to have a positive contribution to job creation starts to diminish. Creation of consumer surplus 2.4 There are some specific economic effects of broadba nd that are not necessarily captured by economic growth or employment creation. This is the case of consumer surplus, which has also been found to be affected by the positive externalities of broadband. Consumer surplus is defined as the amount that consumers benefit from purchasing a product for a pric e that is less than what they would be willing to pay. To compare the gains in consumer surplus by an inve stment one has to compare the initial (before the investment) consumer surplus with the consumer surplus at the end of the investment. During an investment period consumer surplus may change because of two reasons. The first one is an outward- shift of the demand curve and the second is a price reduction. The shift of the demand curve can occur because of the broader penetration of high speed broadband. The price reduction is a result of productivity gains and competition. In the case of deployment of high-speed broadband infrastructures, competition becomes effective at the applications layer. This development is responsible for an increase in consumer surplus in future periods compared to former periods. Consumer surplus is the utility gain by consumers due to prices that are lower than their reservation prices. In Figure 6 the consumer surplus is the area between the demand curve and the market price. Consumers gain utility because they can purchase a produc t at a lower price than they are willing to pay. The larger the area under the curve is, the more utility that consumers derive. The price reduction may result from productivity ga ins and competition. More competition and market saturation force producers to reduce prices. These two developments are responsible for increases in consumer surplus. As indicated in Figure 6, the dark grey area represents the initial consumer surplus at t = 0. The shift of the demand curve at t = 1 results in an additional consumer surplus (light grey area). The whole consumer surplus in period 1 is the sum of the dark and light grey areas. 34 et al. (2006). This effect was also mentioned by Gillett 14

27 Impact of broadband on the economy The estimation of consumer surplus resulting from broadband penetration is important, but this economic benefit is not captured by GDP. This approach has been utilized by Crandall and Jackson (2003) to estimate the US consumer surplus derived from new services like shopping, entertainment and telemedicine enabled by broadband infrastructure. Si ) relied on regression milarly, Lee and Lee (2006 techniques to estimate the consumer surplus for the Korean telecommunications market. Figure 6: Consumer surplus Source: Author Greenstein and McDevitt (2009) estimated the consumer surplus generated by broadband adoption in the United States. In their analysis for the period between 1999 and 2006, the authors determined that in 2006 the consumer surplus generated by broadband re presented USD 7.5 billion (or 27 per cent of the total USD 28 billion in broadband surplus). This was calculated on the basis of what users would be willing to pay to adopt broadband and substitute narrowband access. Consumer surplus can also be conceptualized in terms of the benefits that broadband represents to the end user. The variables driving willingness to pay include the rapid and efficient acce ss to information, savings in transportation for conducting transactions, and benefits in health and entertainment. The authors also recently estimated the surplus generated as a result of broadband adoption in Canada, United Kingdom, Spain, Mexico, Brazil and China (Gre enstein & McDevitt, 2010). In this case, due to the e benefit derived from price declines, which necessarily data limitations, they restricted their analysis to th underestimates its total impact. Nevertheless, the researchers determined that for 2009, the total Brazilian broadband surplus represented USD 7.03 billion, of which 22 per cent should be considered to l surplus is USD 2.30 billion, and the consumer portion be consumer driven. In the case of Mexico, the tota was 8 per cent. In general terms, the authors conc luded that the total broadband surplus is directly related to broadband penetration. 2.5 Impact on firm efficiency Converging with the aggregate macro-economic resear ch, the microeconomic analysis of the impact of broadband has helped understand the multiple effects that the technology has on firm performance. Microeconomic research has yielded the following estimates of firm productivity enhancement (see Table 5). In addition to the impact on productivity, other microeconomic studies have focused on the impact of broadband technology on business expansion, product innovation and new business creation. With ed the impact of broadband access on exports of regards to business expansion, Clarke (2008) studi 15

28 Impact of broadband on the economy in the manufacturing sector firms with Internet manufacturing and service firms. The author found that access enabled by broadband generate 6 per cent more foreign sales than the rest. In the service sector, broadband enabled firms generate between 7.5 per cent and 10 per cent more sales. s been found to have a positive impact on the In addition to increasing exports, broadband ha development of new businesses. This results from the network effects of connectivity. When a large enough number of households are connected to br oadband, the incentive to develop new businesses around information search, advertising and electronic commerce increases. Table 5 – Broadband-induced productivity improvement Study Industrial Sector Share of Informational E Business Impact on Firm Productivity activities that involve external parties 5% Manufacturing Atrostic and Nguyen (2006) 25% ~ ~ Services Rincón-Aznar et al. (2006) 50% 10% ~ ~ Information Fornefeld et al. (2008) 20% 100% ~ Source: Fornefeld et al. (2008). utilized to estimate the impact of broadband on job The results of the microeconomic research have been creation. Fornefeld , (2008) identified three ways that broadband impacts employment: first, the et al. introduction of new applications and services causes acceleration of innovation; second, the adoption of more efficient business processes enabled by broadband increases productivity; and third, the ability to process information and provide services remotely makes it possible to attract employment from other regions through outsourcing. These three effects act simultaneously, whereby the productivity effect and by the innovation effect and gain of outsourced potential loss of jobs due to outsourcing are neutralized jobs from other regions. Figure 7: Network effects of broadband on employment Source: Adapted from a model originally developed by MICUS in a report for the European Commission (see Fornefeld et al., 2008) According to Fornefeld et al. (2008), the negative effect of broadband productivity is compensated by the increase in the rate of innovation and services, ther eby resulting in the creation of new jobs. The third effect may induce two countervailing trends. On the one hand, a region that increases its broadband penetration can attract employment displaced from ot her regions by leveraging the ability to relocate functions remotely. On the other hand, by increasing broadband penetration, the same region can lose 16

29 Impact of broadband on the economy jobs by virtue of the outsourcing effect. While we are gaining a better understanding of these combined “network effects”, the research is still at its initial st ages of quantifying the combined impact. The study by Fornefeld (2008) is probably the first attempt to build a causality chain. It applies ratios derived from et al. micro-economic research to estimate the combined impact of all effects. 2.6 Conclusion A review of the research on the economic impact of broadband indicates multiple effects. First and foremost, the evidence is fairly conclusive about the contribution of broadband to GDP growth. While the amount of this contribution varies, the discrepancies ca n be related to different datasets as well as model specifications. Secondly, broadband has been found to have an impact on the productivity at the firm level. Evidence generated both at the micro-economic and macro-economic level appears to confirm this effect. In addition, research has been successful in id entifying the existence of a critical mass, indicating the existence of increasing economic returns of br oadband penetration. On the other hand, consistent with the research at the ICT level, broadband economic impact could be mediated by a lag effect, indicating that adoption does not automatically tr anslate into growth but that it would require the accumulation of intangible capital, defined as the ch anges in business processes and firm culture that lead to assimilation of improved business processes. of network construction Thirdly, broadband does contribute to employment gr owth, both as a result programmes and following spill-over impacts on the rest of the economy. While the deployment programmes are, as expected concentrated in the construction and telecommunications sectors, the transaction costs (financial services, education, and impact of externalities are greater in sectors with high health care). Finally, beyond economic growth and job creation, broadband has a positive effect in consumer surplus in terms of benefits to the end user that is not captured in the GDP statistics. These include efficient access to information, savings in transportation and benefits in health and entertainment, and can be measured in terms of the difference between consumers' willingness to pay for the broadband service and actual prices. In addition, the review of the literature confirms the existence of multiple methodological approaches aimed at measuring such an impact. Input-output anal ysis has proven to be highly reliable tool to estimate the counter-cyclical impact of broadband co nstruction programmes. Econometric analysis, while limited by data availability, has been proven effectiv e in identifying the spill-over effects of broadband on the rest of the economy. These results have been confirmed by micro-economic research that has concluded outlining the impact of broadband in fost ering efficiencies and value added opportunities at the firm level. As the review indicates, most of the research has so far been conducted in developed nations, either the United States or Western Europe. The challenge go ing forward is to search for similar effects in developing countries, where data availability remains an even larger challenge. In particular, this issue is quite relevant for African countries. In the search for available information to conduct the quantitative case studies presented below, this author encountered a lot of obstacles in identifying adequate datasets for nations in Africa. 17

30 Impact of broadband on the economy Economic Impact of Broadband in De veloped Countries: Case Studies 3 The following section presents research conducted by the author primarily at the aggregate macro- economic level. It aims to validate the existence of economic impact of broadband on GDP growth and job creation. It also includes the following case studies: • United States: employment creation triggered by the Broadband Technology Opportunities 35 Program . 36 • Germany: impact of the National Broadband Plan on economic growth and job creation . 37 utilized in these two cases are the following: The methodologies • Input/output analysis was utilized for estimating the employment and GDP impact of broadband construction (in the United States and Germany). • commissioned to Fornefeld et al (2008) by the Micro-economic estimates, as utilized in the study ng the contribution to employment resulting European Commission, was relied upon for estimati from broadband externalities. • Econometric modelling, as utilized in several studies (including those conducted by authors affiliated by the World Bank, the Brookings Instit ution, and MIT) was relied upon in the case of broadband externalities impact on GDP and employment in Germany. 3.1 as a result of the economic stimulus United States: employment creation programme In the last few years, spurred by the economic crisis, many governments around the world have implemented programmes aimed at deploying broadband in order to stimulate employment (see Table 6). In 2009, this author attempted to estimate the jobs that could be generated as a result of the grants to be disbursed by the broadband provisions of the conf erence report on the American Recovery and study differentiated between jobs generated through Reinvestment Act, published February 13, 2009. The to unserved/underserved areas, and employment created capital spending in the form of grants allocated as a result of network externalities caused by the deployment of such an infrastructure. The study found that approximately 127,000 jobs could be created over a four year period from network 38 construction. According to the analysis , the investment of USD 6.390 billion will generate 37,283 direct jobs over the course of the stimulus programme (estima ted to be four years). In addition, based on a Type 39 . The split across sectors is I employment multiplier of 1.83, the bill could indirectly generate 31,046 jobs presented in Table 7. 35 This analysis is based on prior research contained in Katz, R., and Suter, S. (2009a). Estimating the economic impact of the broadband stimulus plan . Columbia Institute for Tele-Information Working Paper presented at a conference on “Spending the Broadband Stimulus: Maximizing the benefits an ld at the National Press d monitoring performance” he Club in Washington, D.C. on February 19, 2009. 36 This analysis is based on prior resear ch contained in Katz, R. L., Vaterlaus, S., Zenhäusern, P., Suter, S. (2010a). The impact of broadband on jobs and the German economy. Intereconomics , January-February, Volume 45, Number 1 , 26-34. The results were originally presented at the Confeder ation of German Industries in Berlin on June 17, 2009. 37 See review of methodologies utilized in Appendix A. 38 An estimate of funds dedicated pr imarily to broadband deployment, as opposed to ancillary activities such as broadband mapping. 39 For methodology, see Katz et al (2009a). 18

31 Impact of broadband on the economy Table 6 – Counter-cyclical government programmes Country Broadband Focus United States Launched the USD 7.2 billion broadband st imulus programme focused on providing service to unserved and underserved areas Government plans to spend AUD 11 billion of to Australia tal AUD 43 billion required for construction of the National Broadband Network Germany Government has announced a National Br oadband Strategy with the objective to have nationwide capable broadband access (1 Mbps ) no later than the end of 2010 and provide 75 per cent of German households access to a broadband connection of at least 50 Mbps by 2014 (estimated investment: EUR36 billion) Broadband government promotion provides financia l incentives to municipalities to fund 2/3 of Sweden total NGN investment (EUR864 million) Portugal Government announced a EUR 800million credit line for the roll-out of NGAN. This is part of an the first step in a EUR 2.18billion plan to boost the country's economy Ireland The government will invest EUR 322 million in a National Broadband Scheme aimed at completing country coverage Canada Has relied on four programmes to promote broadband development resulting in an overall investment of CAD 300 million Finland Government funds one-third of the NGN project cost (USD 130.73 million) ,12 million investment to boost fibre over the next five years New Zealand Government funds USD 458 Source: Author. Table 7 – United States: Type I employment effects of Broadband Stimulus Bill Sectors Jobs created 4,242 Direct Employment Electronic equipment Construction 26,218 Communications 6,823 Subtotal 37,283 9,167 Distribution Indirect Employment Other market/non-market services 8,841 1,536 Transportation Electronic engineering 959 1,839 Metal products 8,704 Other Subtotal 31,046 68,329 Total Type I Employment 1.87 ct)/direct employment Type I multiplier (Direct+Indire Source: Adapted from Katz et al. (2009a). 046) and induced jobs (59,500) yielded a total The combination of direct (37,283), indirect (31, period. The average annual employment generation employment impact of 127,800 jobs over a four-year effect is 31,950 jobs per year. 19

32 Impact of broadband on the economy In addition, to network construction, the investment in broadband would trigger new jobs as a result of spill-over effects on the rest of the economy. The calculation of spill-over effects was performed by selecting those states in the US where the per cent of residential households which have access to at least 40 There are 18 states one broadband supplier (that is to say telco or cable, primarily) is 93 per cent or less. that lag the national average broadband penetration significantly: while broadband in those states has been adopted by 47 per cent of households (or 21 per cent of the population), the US average is 62 per cent (or 25 per cent of the population). The assump tion utilized to estimate the employment network effects of the stimulus programme was that it would deploy enough lines to allow these eighteen states 41 . 000 subscribers would be added to the existing base to reach the national average, meaning that 3,928, As reviewed above, the estimation of network effect s needs to be done stepwise by accounting for jobs that will be gained as well as those that could be lost. Network effects driven job gains in the targeted regions result from three combined trends: innovation and the creation of new services, attraction of jobs (from either other US regions or overseas), and productivity enhancement. es sector was estimated by applying the ratio of The impact of innovation on the professional servic 42 productivity gains to the creation of new employment. Then this effect is applied to the economy of the targeted states as a whole. As a result, th e following effect is predicted (see Table 8). Table 8 – United States: Jobs gained due to the innovation effect resulting from increased broadband penetration 2009 2010 2011 2012 Total 2008 Total employment 30,123,300 Employment likely to be affected 25,165,000 by outsourcing trend Growth rate in broadband 9% 8% 7% 6% penetration 175,000 Jobs gained by creation of new 33,000 55,000 47,000 40,000 business services Jobs gained as a result of new 64,000 55,000 46,000 38,000 203,000 economic activity Total jobs gained 378,000 119,000 102,000 86,000 71,000 Source: Adapted from Katz et al. (2009a). The impact of broadband on outsourcing operates in two directions: broadband can facilitate the attraction of new jobs and it can enable the relocation of others in regions other than the one being targeted. Assuming the same rate of broadband penetration as utilized in the calculation of the innovation effect, the gains and losses due to enhanced outsourcing can be estimated (see Table 9). 40 See FCC Table 14 of HSPD1207. We have been informed th at Table 14 actually overesti mates the accessibility percent by approximately 2% because cable TV operators tend not to report accurate deployment numbers. 41 There is a large gap between households served by at least one broadband technology (average 89%) and broadband penetration (47%).If the ratio househol ds served/adopted (1.90) remains, th e capacity to serve 7,463,200 additional households should be deployed in order to increase the su bscriber base by 3,928,000. This is well within the bounds of the total grants of the program. 42 improvement takes place. functions where productivity Innovation is assumed to occur in the sectors and 20

33 Impact of broadband on the economy st due to accelerated outsourcing resulting Table 9 – United States: Jobs gained and lo from increased broadband penetration 2008 2009 2010 2011 2012 Total Total employment 30,123,300 Employment likely to be 25,165,000 affected by outsourcing trend Growth rate in broadband 9% 8% 7% 6% penetration Jobs gained 49,000 44,000 38,000 164,000 33,000 Jobs lost 82,000 73,000 64,000 55,000 274,000 Net (33,000) (29,000) (26,000) (22,000) (110,000) Source: Adapted from Katz et al. (2009a). The resulting net number of jobs lost due to outsourcing (110,000) overestimates the economic disadvantage of the targeted 18 states. Their position in a ranking of salary differentials and cost of living indicates that they tend to be in the bottom quartile of the distribution. This would lessen the impact of a negative outsourcing trend. However, it is importan t to consider the impact that broadband may have in given the fact that broadband may also cause job the potential displacement of jobs. Furthermore, growth in the targeted area, one might assume that pro-active employment relocation policies could noted that job outsourcing could take place within increase the number of jobs created. It should be national markets, such as the entire US territory, in which case jobs lost in one state will be gained in another one. Additionally, outsourcing gains could occur internationally. By tapping into new labour pools and lowering labour costs, broadband could help the Unit ed States regain some of the jobs originally lost. As a result of the uncertainty regarding the amount of jobs that be gained or lost, it is prudent to build two additional scenarios (an optimistic and a mid-course one) to be considered with the one derived above, which is considered to be pessimistic (see Table 10). Table 10 – United States: Alternative scenarios regarding outsourcing impact 2009 2010 2011 2012 Total (33,000) (29,000) (26,000) (22,000) (110,000) Pessimistic Scenario Mid-course scenario 8,000 7,500 6,000 5,500 27,000 Optimistic scenario 44,000 38,000 33,000 164,000 49,000 Source: Adapted from Katz et al. (2009a). The optimistic scenario assumes that the eighteen states' comparative advantage regarding factor costing, combined with labour retention policies, are sufficient to cancel out the trend toward job displacement. The mid-course scenario represents the mid-point between the pessimistic and optimistic. In addition, increased adoption of broadband has an impact on productivity because it is an enabler of more efficient business processes. To calculate the productivity impact, the methodology derived by Fornefeld et al. (2008), which is based on empirical firm-level study of sectoral productivity improvements resulting from adopting online services, was applied. By differentiating the productivity impact in es (20%), and the rest of the service sector (10%) manufacturing (5%), professional and information servic 21

34 Impact of broadband on the economy and applying these ratios to sectoral employment, the jobs that could be lost as a result of broadband 43 . diffusion was calculated Assuming that the stimulus programme causes the sa me rate of broadband penetration in the targeted areas, we were able to calculate the jobs lost due to the increased adoption of more efficient processes, enabled by broadband (see Table 11). As Table 10 indicates, the productivity effect resulting from increased broadband penetration could result in 266,000 jobs lost over four years. through the analysis reviewed above results in the The compilation of all employment effects calculated following numbers: Table 11 – United States: Jobs lost due to productivity improvement resulting from increased broadband penetration Total 2010 2011 2012 2009 2008 25,661,000 Manufacturing and Services Employment 3,860,000 Professional and Information Services 8% 7% 6% Growth Rate in BB penetration 9% 48% 52% 56% 59% 62% Broadband penetration Jobs lost in professional and 19,000 17,000 15,000 13,000 64,000 information services Jobs Lost in other sectors 54,000 47,000 40,000 202,000 61,000 266,000 Total jobs lost 53,000 80,000 71,000 62,000 Source: Adapted from Katz et al. (2009a). impact of the broadband stimulus plan Table 12 – United States: Total employment 2009 2010 2012 Total 2011 Direct effects 9,325 9,325 9,325 9,325 37,300 Indirect effects 7,750 7,750 7,750 7,750 31,000 Induced effects 14,875 14,875 14,875 14,875 59,500 Network effects (optimistic) 87,000 74,000 62,000 50,000 273,000 46,000 Network effects (mid-estimate) 30,000 37,500 22,500 136,000 Network effects(pessimistic) 5,000 (2,000) (5,000) (1,000) 1,000 Total (optimistic) 105,950 93,950 81,950 400,800 118,950 Total (mid-estimate) 77,950 69,450 61,950 54,450 263,800 Total (pessimistic) 36,950 32,950 29,950 26,950 126,800 Source: Adapted from Katz et al. (2009a). 43 et al. (2009a). See methodology in Appendix B.2 of Katz 22

35 Impact of broadband on the economy These estimates allow drawing the following conclu sions for the network construction job creation: from the stimulus programme has moderate • The deployment of broadband accesses resulting direct employment effects (37,300 jobs over a four year period). • Indirect and induced multipliers are important, gene rating a total of 127,800 jobs over four years. pe II multipliers is close to Atkinson • (2009), the While, on aggregate, the estimate of Ty et al. other sectors of the economy based on calculation of indirect effects (impact in interrelationships) is more conservative, and the induced effects (derived from household spending) appear to be a bit more optimistic. This is because the estimation of induced effects is based on sector-specific multipliers rather than aggregates. The great part of the difference between the projections for the number of total network • et al. construction jobs and those of Atkinson (2009), (229,475 versus 127,800), is due to the difference in the absolute size of the stimul us initially assumed (USD 10 billion versus USD6.4 billion). Therefore, it is considered that the estimates for jobs created as a result of network construction • are quite robust. When moving to the estimation of network externalit ies, one should observe the wide range of network effects in this study. The estimates of network effects range from close to nil to a much more optimistic scenario than Atkinson et al. (2009). As a result, the broadband stimulus could either lead to no et al. externalities or the creation of up to 273,000 jobs in four years. This number exceeds Atkinson (2009) estimates for a USD 10 b illion programme (268,480 jobs). jobs as a result of outsourcing, a high level of In summary, by ranging the potential success in attracting uncertainty is introduced in the final estimate. Since increased broadband penetration has an impact on productivity and outsourcing (which can result in job destruction), unless the innovation and in-sourcing programmes are effective in promoting growth and job creation, any network effects can be significantly eroded. As a result, in order to be successful the br oadband stimulus programme needs to be coordinated with other employment generation initiatives. Germany: The impact of the nation 3.2 al broadband plan on employment and economic growth This case study quantifies the macroeconomic impact of investment in broadband technology on employment and output of Germany’s economy. Two sequential investment scenarios are analysed: the first is based on the national broadband strategy anno unced by the German Government in 2009. It aims to provide 75 per cent of German households with access to a broadband connection of at least 50Mbps by 2014. The second scenario (labelled “ultra-broadband” and covering 2015-2020) defines the investment required to provide to 50 per cent of households with at least 100 Mbps, and another 30 per cent with 50 Mbps by 2020. construction of broadband infrastructure required to In order to estimate the impact resulting from the meet the strategy targets, the investment is broken down in three primary sectors of the economy that receive the lion’s share of benefits: manufacturing of electronic equipment, construction and telecommunications. Fulfilling the 2014 objectives of the national broadba nd strategy is estimated to generate 304,000 jobs 44 over five years (between 2010 and 2014). Following the breakdown of the construction effects, 158,000 jobs will be created in equipment manufacturing, cons truction and telecommunicati ons. We estimate that 44 Given the static nature of I/O-matrix it is not possible to project job creation over time. This could be done, however, if yearly investment data is available. 23

36 Impact of broadband on the economy job creation will be apportioned among the sectors as follows: construction will benefit the most with 125,000 jobs, followed by telecommunications (28, 400) and electronics equipment manufacturing (4,700). onships are 71,000. The key sectors benefited from the Total indirect jobs generated by sector interrelati indirect effects are distribution (10,700), other se rvices (17,000) and metal products (3,200). Finally, , will result in 75,000 induced jobs. Based on these household spending generated directly and indirectly estimates, the Type I multiplier for em ployment is 1.45 and Type II is 1.92. Additionally, the implementation of the expect ed ultra-broadband evolution will generate 237,000 incremental jobs between 2015 and 2020. This figure is distributed similarly to the one above: it ed jobs. As expected, multipliers comprises 123,000 in direct jobs, 55, 000 indirect jobs and 59,000 in induc will be similar, the Type I multiplier for employment is 1.45 and Type II is 1.93. Table 13 presents all the employment effects reviewed above: The sector impact of direct and indirect effects is included in Table 14. As Table 14 indicates, the labour intensive nature of broadband deployment causes significant creation of construction jobs. Despite the high-technology nature of the ultimate product, broadband is to be seen as economically meaningful as conventional infrastructure investment such as roads and bridges. pact of broadband network construction Table 13 – Germany: Total employment im 2014 National broadband Type of Impact 2020 Ultra-broadband Total evolution strategy 158,000 123,000 281,000 Direct effect Indirect effect 71,000 55,000 126,000 75,000 Induced effect 134,000 59,000 Total 304,000 237,000 541,000 Type I multiplier 1.45 1.45 1.92 Type II multiplier 1.93 Source: Adapted from Katz et al. (2010a). direct and indirect job creation Table 14 – Germany: Sector impact of 2014 National broadband 2020 Ultra-broadband Total strategy evolution Construction 125,000 99,000 224,000 28,400 Telecommunications 49,400 21,000 Other services 17,000 13,000 30,000 Distribution 10,700 8,400 19,100 Metal products 4,800 3,700 8,500 Electronics equipment 4,700 8,100 3,400 3,200 2,500 Electrical equipment 5,700 Financial services 3,000 2,000 5,000 Other 32,200 25,000 57,200 Total 178,000 407,000 229,000 Source: Adapted from Katz et al. (2010a). In addition to estimating employment effects, industr ial output and GDP impact were also calculated. The investment required to meet the targets of th e 2014 Broadband strategy, (EUR 20,243 million), will generate additional production totalling EUR 52,324 m illion. This means that for each Euro invested in broadband deployment, EUR 2.58 will be generated in output. Of this, EUR 4,146 million (8 per cent of 24

37 Impact of broadband on the economy total output) will be based on imported goods. This in dicates a relatively low level of output “leakage” to other national economies. Of the remaining production, EUR 18,733 million would be additional GDP ployment will trigger 0.93 per cent in additional (+0.15%). Again, each Euro invested in broadband de value added, or incremental GDP. Table 15 compiles the economic impact of the 2014 and 2020 targets: Table 15 – Germany: Industrial output of broadband construction (in EUR millions) 2020 Ultra-broadband 2014 National broadband Total evolution strategy 35,933 Investment 20,243 15,690 52,324 Total additional production 93,073 40,749 Domestic 48,178 37,609 • 85,787 Additional Value added 18,733 14,631 33,364 Intermediate outputs 29,466 22,978 52,444 7,294 • Imported 4,146 3,148 Source: Adapted from Katz et al. (2010a). To sum up, the incremental GDP growth achieved by investing in broadband deployment would amount to EUR 33,364 million (which represents +0.12 per cent of the German GDP). This amount does not include the additional impact to be achieved once the network construction is completed. The author also found that the network externalities of broadband, (that is to say, the positive effects in employment and economic output resulting from enhanced productivity, innovation and value chain e analysis of these effects examined the relation decomposition), are significant throughout Germany. Th between broadband penetration and economic growth and job creation. It has found that the economic ar after deployment and tends to diminish over time stimulus impact of broadband is highest in the first ye (see Figure 8). Figure 8: Conceptual depiction of broadband externalities HI Economic Impact LO T+2 T+3 T+1 T+4 Increase in BB penetration Source: Author me series between 2000 and 2006 indicate with high Results of the regression analysis for national ti significance levels that there is a strong impact of broadband penetration on GDP growth, which tends to diminish over time. On the other hand, results regarding the impact of broadband penetration on . Therefore, they do not allow us to indicate the employment creation carry a low level of significance existence of causality with certainty. 25

38 Impact of broadband on the economy In addition, by splitting the national territory into two groups, counties with 2008 average broadband penetration of 31 per cent of population and count ies with average broadband penetration of 24.8 per ork effects of broadband varies by region. In high cent, the analysis determined that the type of netw of the technology is very high both on GDP and broadband penetrated counties the short-term impact employment, but it declines over time. This “supply shock” is believed to occur because the economy can immediately utilize the new deployed technology. Furthermore, the fact that employment and GDP grow in parallel indicates that broadband has a significant impact on innovation and business growth, thereby overcoming any employment reduction resulting from productivity effects. On the other hand, in counties with low broadband penetration the impact on GDP of broadband penetration is lower than in high penetrated areas in the short term, but “catches up” to comparable levels over time. The impact of broadband on employment is slightly negative in the initial years. This indicates that the impact of broadband in low penetration areas is more complex than in the high penetration areas. The increase in broadband penetration in low penetrated areas takes longer to materialize economic growth because these economies require a longer period of time to develop and fully utilize the technology. However, after three years the level of impact of broadband in low penetrated regions is as high as in the more developed areas. Negative initial employment growth appears to indicate that the productivity increase resulting from the in troduction of new technology is the most important network effect to begin with. However, once the economy develops, the other network effects 45 . (innovation and value chain recomposition) start to pl ay a more important role, resulting in job creation Therefore broadband deployment in low penetrated areas will likely generate high stable economic growth (“catch up” effect) combined capital/labour s ubstitution, which initially limits employment growth al fashion a comparison of impact in both regions. (“productivity” effect). Figure 9 presents in conceptu 46 Figure 9: Conceptual view of comparative broadband regional effects Low Broadband Penetration Landkreise High Broadband Penetration Landkreise GDP GDP HI HI Employment Employment Economic Impact Economic Impact LO LO T+2 T+3 T+3 T+2 T+4 T+4 T+1 T+1 Increase in BB Increase in penetration BB penetration • • High economic growth initially, High stable economic growth (“catch up” effect) diminishing over time (“supply shock” • Capital/labor substitution limits effect) New Economic Growth (innovation, employment growth (“productivity • effect”) new services) Source: Adapted from Katz et al. (2010a). 45 This said, the available data sets do not enab le us to test this last point at this time. 46 Only effects up to t + 3 are estimated. 26

39 Impact of broadband on the economy the impact of broadband on economic growth and These differentiated effects were used to estimate at broadband penetration in advanced areas would employment are estimated. It was stipulated th increase from 31 per cent in 2008 to 45.9 per cent in 2014, while the low penetration areas will increase from 24.8 per cent to 37.4 per cent. This trend is largely driven by the coverage of “white spots” and an 47 . Together, these improvements amount to an incremental improvement of service in “grey spots” 48 . increase in penetration of approximately 25 per cent in both regions between 2008 and 2011. 49 The per cent increase was inputted in the regressi on models specified for the time series 2000-2006. The regression models estimate an incremental annual GDP growth rate of 0.61 percentage points for low penetrated counties and 0.64 percentage points for high-penetrated counties, which over three years amounts to 1.93 per cent and 1.82 per cent respectively. These incremental percentage point increases were applied to the GDP of both regions (estimated to be EUR 1,698 billion for high penetrated counties and EUR 791 billion for low penetrated counties). Th is represents an incremental GDP of EUR 32,809 million for high-penetrated counties and EUR 14,375 millio n for low penetrated counties. In sum, the total incremental GDP is EUR 47,184 million (+0.62%) in three years. Following the same methodology, it is estimated th at 162,000 jobs will be created. More developed 50 broadband areas are expected to gain 132,000 and the low penetrated regions 30,000. The differentials across regions are driven by the divergent effects discussed above. As discussed throughout the review of the study projections, the estimates were generated for several years and dependent on stages of network deployme nt. For example, the projected 541,000 jobs due to network construction do not occur all in one year but over a ten-year period. To understand the yearly impact of the estimates, a table that displays the yearly impact over time is included (see Table 16). 47 ice, while “grey spots” are the areas with uneven coverage. "White spots" are defined as areas lacking wireless serv 48 anslate infrastructure deployme nt programs into increased A cautionary note should be made that in order to tr complemented with very targeted demand promotion broadband penetration, network construction should be programs (be they community aggregation programs such as the ones of the Dutch govern ment, tax deductions such as the ones implemented in Sweden, and, potentially, subsidies) that stimulate adopters to sign up for service. 49 g_gdp_03_06 = β 1*gdp_pc_2000 + β 2*g_pop_00_06 + β 3*g_bbpen_02_03 g_emp_03_06 = 1*gdp_pc_2000 + β 2*g_pop_00_06 + β 3*g_bbpen_02_03 β 50 While it is not possible to determine, as in the case of network construction, what type of sectors would be mostly impacted by network externalities, experience indicates th at higher developed areas will generate knowledge-intensive occupations such as R&D and product development, while less developed regions will attract low-end information intensive jobs, such as virtual call centers. 27

40 Impact of broadband on the economy and economic impact per annum Table 16 – Germany: Employment Source: Adapted from Katz et al. (2010a). Based on these figures, the net employment effects over two time periods were calculated (see Table 17). Table 17 – Germany: Employment effect over time (in thousands) Network Construction Network externalities Total Direct Indirect Induced Total 2010-14 158 71 75 304 103 407 561 2015-20 123 55 59 237 324 968 Total 281 126 134 541 427 Source: Adapted from Katz et al. (2010a). Between 2010 and 2014 407,000 jobs will be created and between 2015 and 2020 561,000 jobs will be 51 created . The growth in impact over time is due to the fact that between 2015 and 2020, additional ultra- broadband lines are deployed as part of the second phase of the national strategy. To sum up, the national broadband strategy and the expected evolution of ultra-broadband through 2020 will have a significant impact on jobs and GDP of estimated that a total the German economy. It is a total of 968,000 incremental jobs, of which 541,000 investment of nearly EUR 36 billion will generate will be derived from the network construction required to meet the stipulated targets. An additional 427,000 will be generated after the network is deploy ed due to enhanced innovation and new business creation. From an incremental economic growth standpoint, network construction will yield additional value added of EUR 33.4 billion, while network externalities will result in an additional EUR 137.5 billion. 51 Three remarks are noteworthy: First, it is obvious that, while total projections have been split evenly over time, one would expect yearly projections to vary. For example, more jobs are generated in the beginning of network deployment than in the back-end. This would require further refinement of these projection s that takes into account construction plans. Second, as mentioned above, in order to avoid double counting with respec t to network construction effects in the first years and to assure a conservative calculation some of the totals have been reduced. Third, we assume that the riod with effects through 2020 following regression model is capable of projecti ng economic impact over a three year pe suite. 28

41 Impact of broadband on the economy Economic Impact in Developing Countries: Case Studies 4 The following section compiles research conducted by the author primarily at the aggregate macro- economic level. It aims to validate the existence of economic impact of broadband on GDP growth and job creation in developing countries. It includes the following case studies: 52 Latin America and the Caribbean: contribution of broadband to GDP growth • . 53 . • Brazil: the impact of broadband on employment and economic growth 54 • Chile: the impact of broadband on employment and economic growth . • Dominican Republic: the contribution of broadband to employment growth. • Arab States: impact of broadband on economic growth. • Saudi Arabia: the impact of broadband on employment. • India: the impact of broadband on employment and economic growth. • Malaysia: the impact of broadband on economic growth. • China: the relationship between broadband deployment and economic growth. • Indonesia: the contribution of broadband to employment growth. As it can be seen, in the case of Latin America, first a regional model was constructed to identify broadband contribution to the sub-continent. This was followed for specific modelling exercises for Brazil, Chile and the Dominican Republic. In the case of the Arab States, a general model was developed, followed by a quantitative case study on Saudi Arab ia, and qualitative studies based on descriptive statistics on Jordan, United Arab Emirates and Qatar. In the case of Asia, only case studies were developed for a selected set of countries. In all of the case studies mentioned in the list ab ove, the analysis was conducted based on econometric modelling similar to the one utilized for the studies conducted by researchers affiliated with the World Bank, MIT, and the Brookings Institution. In applying this approach in developing countries, the study is the lack of sufficiently large time series and confronted with major methodological problems due to sample sizes and sufficiently disaggregated indicators . For example, in the case of developed countries, Katz (2010a) were able to rely on over 424 observations regarding economic, control and broadband et al. variables for German counties. This situation is rarely the case in the emerging world. This is why the results that will be presented in the next section need to be prefaced with a word of caution. Sample sizes timates. In five of the twelve models presented lower than 30 observations limit the robustness of es below (economic growth for Latin America, economic growth and employment creation models for Chile, employment creation model for the Dominican Republic, and economic growth for the Arab States), the number of observations exceeded this threshold. Secondly, the lack of time series does not allow the construction of models based on simultaneous equations, such as the one developed by Koutroumpis (2009). As a result, it is difficult to control for 52 This analysis was originally published in Katz, R.L. “The contribution of broadba nd to economic development”, in Jordan, V., Galperin, H. and Peres, W. (2010). Fast-tracking the digital revolution: Broadband for Latin America and the Caribbean , published by the Economic Commission for Latin America and the Caribbean (ECLAC) and DIRSI. 53 This analysis is based on prior rese arch contained in Katz, R. L. (2010b). La banda ancha: un objetivo irrenunciable para Brasil presented at the 54o Panel Telebrasil. Guaruja, on August 18, 2010. 54 This analysis was originally published in Katz, R.L. “The contribution of broadba nd to economic development”, in Jordan, V., Galperin, H. and Peres, W. (2010). Fast-tracking the digital revolution: Broadband for Latin America and the America and the Caribbean (ECLAC) and DIRSI. Caribbean , published by the Economic Commission for Latin 29

42 Impact of broadband on the economy 55 lied upon for this issue was a cross-lagged regression . endogeneity. The only approach that could be re The authors recognize that this methodology has some limitations regarding endogeneity control. to GDP growth increases wi th penetration, it is Thirdly, recognizing that the contribution of broadband pertinent to assert that at adoption levels as low as the ones pervading the emerging world up to 2008 56 . the economic effects could be quite limited 4.1 Latin America: Contribution to regional economic growth In a prior paper, this author presented a simple regression model linking Latin American broadband 57 penetration and economic development . In this case, this author attempted to advance the research by 58 . This model has been used developing a multi-variate equation based on the endogenous growth model by several authors to assess the impact of broa dband and other telecommunications technologies on a 59 country's economic growth . 4.1.1 Multivariate Regression Analysis: Given the lack of available time series data regarding broadband penetration in Latin American 60 , the author chose to conduct a cross-sectional an alysis with pooled data for the period 2004- countries 2009, relying on ordinary least squares with robust e rrors. Two problems needed to be addressed with this type of analysis. The first proble m is that the constant term in a li near regression does not capture the on is to rely on panel data, which allows control potential differences among countries. One possible soluti for the country idiosyncratic factors. However, the lim ited availability of data prevented the author from relying on this approach. Yet, by including variables such as technology (e.g. broadband) and the level of openness of the economy, it is expected that problems linked to the omitted variables be mitigated. The second problem has to do with the endogeneity between GDP per capita and broadband penetration. Ideally, the author would have liked to tackle this problem by relying on an approach similar to Koutroumpis (2009), who implemented a simultaneous equations model that endogeneizes the decision to deploy broadband as a function of GDP per capita, pricing, competition and regulation. However, given the lack of time series data on pricing and competit ion for Latin American countries, it was impossible to use this approach. Therefore, to control for this problem the analysis was based on the lag of the broadband penetration variable. 55 See definition of cross-lagge d regression methodologies in Appendix A, Section A.2. 56 In that vein, it was only at the begi nning of the current century, after more than forty years of intense adoption of computers, that growth accounting econ e impact of ICT on US productivity omists were able to finally show th (Jorgenson et al. , 2007). 57 See Katz, 2009d. 58 See Barro, 1991. 59 See Qiang & Rossotto, 2009; Crandall et al ., 2007; Garbaz et al. , 2008. 60 Broadband penetration data for the majority of Lati n American and Caribbean countries after 2003 is available (25 countries included in the sample). 30

43 Impact of broadband on the economy The following variables were used (see Table 18): Table 18 – Latin America: Variables utilized to measure broadband impact on economic growth Source Rationale Data set Type of variable GDP (2004-6) and GDP World Bank and Central Banks Dependent variable Economic growth (2007-9) Control measure for Level of economic World Bank GDP per capita for 2003 and 2006 starting point of growth development Average Investment / GDP Control for World Bank Control measure for Investment for (2001-3) and (2004-6) differences in investment levels Population growth Population growth for Control measure for World Bank (2004-6) and (2007-9) differences in population size Control measure for Human Capital Tertiary education ( 2002) Unesco, Earthtrends, University differences in human of West Indies, Euromonitor, capital Government of the Commonwealth of Dominica Globalization Average globalization index Control measure for Dreher et al. (2008). Measuring NY: Springer. globalization. for (2001-3) and (2004-6) differences in level of openness (economic, social and political) Independent variable ITU and National Regulatory Fixed Broadband Broadband Agencies penetration growth penetration (2003-4) and (2005-6) growth Source: Author. The dataset relied upon is included in Appendix C. The model results are as follows: Table 19 - Latin America: Broadband impact on economic growth Standard T-statistic P>[t] 95% Conf. interval GDP growth Coefficient error Broadband penetration growth 1.98 0.054 –.0002942 .0320372 .0158715 .0080104 –.0471624 .1689699 –0.28 0.782 –.3881575 .2938328 Average Investment / GDP –.4469177 1.40418 –0.32 0.752 –3.280668 2.386832 Population growth 1.93 .2139614 .1108325 0.060 Human capital –.0097076 .4376304 Level of development –.0006957 –3.85 0.000 –.0010602 –.0003313 .0001806 Globalization .1929498 –0.34 0.737 –.4546908 .324086 –.0653024 Constant 13.02883 12.04659 1.08 0.286 –11.28217 37.33982 49 Number of observations F(6,42) 7.18 Prob>F 0.0000 R2 0.3814 Root MSE 7.024 Source: Author. 31

44 Impact of broadband on the economy The coefficient of broadband penetration was positive and significant. According to this, an increment of 1 per cent on the penetration of broadband services could generate an additional 0.0158 to the GDP 61 . This econometric estimate was used to calculate the contribution of broadband to growth of the region the growth of GDP in Latin America and the Caribbean. Based on IMF projections, the economic growth in the region between 2009 and 2010 will be 3.4 per cent, resulting in an aggregate GDP of USD 3,925 billion. The model specified above indicates that the elasticity of broadband with regard to GDP growth is 62 . In light of the possibility of sample bias and given the lack of time series, it is advisable 0.0158per cent to define a reasonable elasticity range. The author believes that an appropriate choice would span from the estimate derived in this study to that deri ved by Koutroumpis (2009) for OECD countries with broadband penetration lower than 20 per cent: 0.008 pe r cent. Thus, by relying on both estimates and considering the growth of broadband in the region, (37 per cent per annum between 2007 and 2009), we USD 6.7 billion and 14.3 billion. This impact includes conclude that the technology contributed between direct effects from the telecommunications industry and indirect spill-overs. 4.1.2 Case studies d above which proves in the aggregate that broadband In addition to the cross-sectional analysis presente has a contribution to economic growth, research also should focus on the specific effects by country. It is expected that, given the particularities of each nati on, the economic impact of broadband might be of a different nature. The following set of case studies provides a first view of broadband’s impact at the country level. 4.1.2.1 Brazil: The impact of broadband on economic growth and employment azil, it was attempted to replicate the analysis To estimate the economic impact of broadband in Br conducted at the country level for Germany to an emerging country. However, the lack of disaggregated data prevented the specification of a model with robust estimates due to the limited number of observations. As a result, the model built to esti mate the impact of broadband on the Brazilian GDP growth relied on a database for the 27 states of Br azil. It contained the following information (see Table 20): Table 20 – Brazil: Variables used to measur e the impact of broadband on GDP growth Source Series Variable Observations 63 Economic growth IBGE Dependent variable Regional GDP per capita (2006-7) Control variable used to determine the GDP per capita (2002) Level of economic IBGE development starting point of development Human capital Illiteracy rate 2002 IBGE Control variable to determine differences in human capital Interstate trade costs Cost index for inter-state Newton de Castro Control variable to determine differences in cost of transportation of (2004) trade costs goods Costs to create a new Lima Chagas Control variable to determine Business creation business (2009) environmental differences in the creation of new businesses 61 Alternative specifications were used including primary and secondary education enrollment rates as proxy for human capital, but only the model including tertiary education resulted significant. 62 ic crises, since the model was run for the period 2004-9. This is indicative for a period without major econom 63 Instituto Brasileiro de Geografia e Estadistica. 32

45 Impact of broadband on the economy Variable Observations Series Source Average of Gini Change in economic Control variable to determine IBGE differences in income distribution inequality coefficient (2004-5) Growth of household Household Survey Independent variable Fixed Broadband broad.penetration (IBGE) Δ penetration (2005-6) Source: Adapted from Katz (2010b). The methodology used to calculate the penetratio n of broadband by state follows the methodology 64 ) (2010) proposed by IPEA ( Instituto de Pesquisa Economica Aplicada . The state-wise penetration was of Households administered by the IBGE ( Instituto calculated using data from the National Survey 65 . Each response in the survey was Brasileiro de Geografia e Estatística ) for the years 2005 and 2008 linked to only one household, thus eliminating the possibility of over-estimation. Finally, the compound annual growth rate was calculated to construct the broadband penetration growth variable. The results are as follows: Table 21 – Brazil: Impact of broadband on GDP growth Coefficient GDP Growth (2006-7) T-statistic P>[t] Standard error Economic development 0.0002883 –2.57 0.018** –0.0007415 –0.4950848 0.2323575 –2.13 0.046** Human capital 0.641 Interstate trade costs –0.0004711 0.0009957 –0.47 New business creation –0.0009246 –0.13 0.899 0.0072004 32.67246 0.71 0.488 Change in economic inequality 46.25561 Δ (2005-6) 0.0082117 0.0500811 0.16 0.871 Broadband penetration 19.50307 Constant 10.06483 0.611 0.52 Number of observations 27 F(6,20) 5.84 Prob>F 0.0012 2 R 0.2880 4.4716 Root MSE Source: Adapted from Katz (2010b). variables with appropriate signs are the 2002 State GDP As the results indicate, the statistically significant and illiteracy rate. In accordance with Barro’s economic growth theory, the 2002 State GDP is negative, 64 The 2005 and 2008 National Household Survey “Pesquisa Nacional por Amostra de Domicilios" contains a question regarding access to broadband in households. The survey ha s two sections, individuals and households, the “Broadband Access” question is included in the latter. Because identificati on variables are included in the first section of the survey, individuals, this question and the identification variables could be matched quantitatively. Finally, using the factors of expansion located on the first section of the survey, broadba nd penetration was calculated for each state. This is the same methodology used by the IPEA in its research study “Análise e recomendações para as políticas públicas de massificação de acesso à internet em banda larga”, published on April 26 2010. 65 The survey relies on a national sample; to expand the survey the coefficients provided by the IBGE were used. The question used to determine the existence of a broadband connection in a household is: “Have you accessed to the household during the last three months?”. internet using a broadband connection in your 33

46 Impact of broadband on the economy which would indicate a conditional convergence toward a single growth path. The illiteracy rate is also negative, which confirms the importance of human capital in economic development. The coefficient for significant, which demonstrates that its impact is broadband penetration rate of change is not statistically not uniform across all Brazilian states. However, we assu me our estimation is valid because it is consistent with the results of similar studies (e.g. Koutroumpis, 2009). Thus, while recognizing the model limitations, we expect that when controlling for education level and departing point of economic development in the case of Brazil, an increase of 1 per cent in the ra te of penetration of broadband could contribute 0.008 percentage points to GDP growth. In addition to estimating the broadband impact on GD P growth, we studied the impact of the technology on Brazilian employment. In this case, a cross-sectional sample similar to that one utilized by Katz et al (2010a) for Germany was constructed, where the dependent variable was the rate of change of unemployment. The following information was compiled for the 27 Brazilian states: Table 22 – Brazil: Variables utilized to estima te the impact of broadband on job creation Variable Series Source Observations Change of Change in unemployment IBGE Dependent variable rate (2006-7) unemployment rate Level of development GDP per capita (2003) IBGE Control variable to determine the point of departure of state economic growth Growth in household Fixed Broadband Household survey Independent variable broadband penetration Δ (2005-6) (IBGE) penetration Control variable to differentiate the IBGE Human capital Years of schooling level of human capital by state Population growth Population growth IBGE Control variable to differentiate the (2006-7) level of population growth by state Source: Adapted from Katz (2010b). The results of the model are as follows (see Table 23). Table 23 – Brazil: Impact of broadband on Job creation Unemployment Rate Coefficient Standard error T-statistic P>[t] Level of state economic –0.0449243 0.0259892 –1.73 0.098 development –1.94 Growth in household broadband 0.066 –0.0069189 0.003575 penetration 0.256 0.1095254 0.0940011 1.17 Human capital Population growth 0.2009585 0.1213108 1.66 0.112 .5035225 Constant –0.1925308 0.706 –0.38 Number of observations 27 F(4,22) 3.76 Prob>F 0.0178 2 R 0.4058 0.27016 Root MSE Source: Adapted from Katz (2010b). 34

47 Impact of broadband on the economy ling years and population growth is not statistically It is counterintuitive that the difference in schoo vel of unemployment across regions. However, the significant in explaining the differences in the le relationship between the rate of change of unemployment rate and the rate of change in broadband penetration is significant and with the expected negative sign. According to the model results, a change of 10 per cent in broadband penetration could reduce th e unemployment rate by 0.06 percentage points. 66 , which Brazilian unemployment in 2009 was 7.7 per cent and that by 2010 it would fall to 7.4 per cent would imply a reduction of 3.89 per cent. According to our estimates, if broadband penetration were to grow by 20 per cent, the impact on the rate of change of unemployment would be a further 0.138. Thus we would expect such a deployment of broadband to result in a reduction of unemployment from the original 3.89 per cent to 4.03 per cent. 4.1.2.2 Chile: The impact of broadband on employment and economic growth The availability of an extensive database of quarterly data for Chile allowed the development of a panel of time series data for each of Chile's administrative regions. This data base was constructed by compiling olitan areas due to the lack of quarterly data) from data for each of Chile's regions (except for the metrop 2001 until the fourth quarter of 2009. The data set contains the following information: Table 24 – Chile: Variables used to estima te the broadband impact on job creation Variable Series Source Observations Employment Quarterly employment rate Regional Institutes of Dependent variable (2002-9) Statistics Level of economic Quarterly index of economic Regional Institutes of Control variable of activity (2001-9) activity by region Statistics economic development Independent variable Growth in broadband in fixed Quarterly Δ Subtel broadband penetration penetration (2002-9) Human capital Schooling years (population Employment Survey, INE Control variable to determine differences in 15 years old and older) human capital Dominant industrial Contribution of the mining Central Bank of Chile Variable to control for sectors regional specialization in and financial sector to regional GDP (2002-2008) dominant economic activities Dynamic sectors Contribution of the Central Bank of Chile Variable to control for agricultural and trade sector regional specialization in dynamic economic to regional GDP (2002-2008) activities Source: Author A model including level of economic activity and broadband penetration was specified. In addition, an alternative model was proposed aimed to study possibl e effects of human capital and specialization on the level of employment. According to methodology used, other specific characteristics of each region that could have an impact on the labour market are controlled by the fixed effects of the panel data. Thus, the model results are as follows (see Table 25). 66 According to a consensus forecast compiled by EMIS. 35

48 Impact of broadband on the economy The economic activity variable remained unaltered between models 1 and 2; in both cases it was 67 and positive in both significant and positive. The coefficient of broadband penetration is significant cients of broadband impact in both specifications specifications. The small variation between the coeffi suggests the robustness of the following conclusion: an increase in 1 percentage point in broadband penetration would contribute nearly 0.18 pe rcentage points to the employment rate. Table 25 – Chile: Broadband impact on job creation Model 2 Employment rate Model 1 Coefficient t-statistic t-statistic Coefficient Level of economic activity 0.000353 5.90 0.000353 5.72 by region Growth in broadband 0.18118 2.56 0.1774 3.85 penetration –0.0042 –1.87 Human capital Dominant sectors –0.00133 –1.66 Dynamic sectors 0.001743 1.27 Constant 0.8682527 109.03 0.913817 25.95 Number of observations 324 Number of observations 276 F(2,310) 60.89 F(5,259) 20.78 Prob>F 0.0000 0.0000 Prob>F 0.2820 R2 0.2863 R2 33.89 F(11,259) 24.41 F(11,310) 0.0000 Prob >F 0.0000 Prob>F Source: Author . An interesting result on the second specification wa s the coefficient of human capital, which resulted significant at the 10 per cent level and with a negative si gn. Contreras at al (2008) argues that this result is mostly explained by the impressive increment in the years of schooling of the population in one 68 generation and the entrance of women to the labour force. Today, the Chilean workforce comprises 6,500,000 indi viduals, which result in 93 per cent employment rate. Of those, it is estimated that broadband depl oyment, which reached a penetration of 9.78 per cent, contributed in 1.76 percentage points to the employ ment rate, which amounts to the creation of 114,426 direct and indirect jobs. In addition to estimating the impact of broadband on job creation in Chile, we attempted to estimate its contribution to the country's GDP growth. For this pu rpose, a database for the 13 regions of Chile was built, which contained the following information (see Table 26): 67 It was significant at the 1 per cent level on the first specification and at the 5 per cent level on the second one. 68 labor force. As more women According to Contreras, et al (2008), women were educat ed but were not part of the entered the labor force the average schooling years of the population increased but also increased the number of women unemployed looking for jobs. 36

49 Impact of broadband on the economy Table 26 – Chile: Variables used to measure the impact of broadband on GDP growth Source Observations Variable Series Dependent variable Economic growth Average annual growth of GDP per Central Bank of Chile capita (2003-4; 2005-06;2007-08) Regional GDP (2000;2001;2002) Level of Variable to control for Central Bank of Chile point of departure in development economic development Human capital Percentage of population with Casen Survey Variable to control for the some level of tertiary education level of human capital (2003-2006) Region population as a per cent of National Institutes of Population size Variable to control for Statistics – Chile country population (2002) level of density Regional population growth (2002- National Institutes of Variable to control for Population growth Statistics – Chile differences in rate of 2008) population growth Contribution of the agricultural and Central Bank of Chile Economic dynamics Variable to control for trade sector to regional GDP (2002) regional specialization in dynamic economic activities Urban density Percentage of people living in National Institutes of Variable to control for Statistics – Chile level of urbanization urban centres by region (2002) Independent variable Growth of Under-Secretariat of in fixed broadband penetration Δ Telecommunications (2002-3: 2004-05; 2006-07) broadband penetration (Subtel) Source: Author. The results are as follows: Table 27 – Chile: Contribution of broadband to GDP growth Dependent variable: GDP P>[t] Coefficient Standard error T-statistic Growth (2006-7) –1.35E–06 7.67E–07 –1.76 0.088 Level of development Human capital 1.62164 0.942361 1.72 0.095 Population size 0.529056 0.292309 1.81 0.08 0.068 Population growth 2.068497 1.091899 1.89 Economic dynamics –0.00809 –0.1 0.921 0.080571 –0.05128 0.080926 Urban density 0.531 –0.63 Broadband penetration growth 0.00927 0.003716 2.49 0.018 Constant –2.08091 6.80365 –0.31 0.762 39 Number of observations F (7,31) 2.98 Prob>F 0.0165 2 R 0.4021 Source: Author. 37

50 Impact of broadband on the economy Broadband penetration was found to be statistically sign ificant and with the expected sign in terms of contributing to GDP growth in Chile. According to the coefficient of this variable, a 10 per cent increase in penetration will result in an increase of 0.09 percentage points in regional GDP of Chile's regions. The r the starting point of development. The coefficient initial regional GDP variable was included to control fo sign is consistent with the conditional converge nce theory. According to the model, human capital measured as the percentage of people with tertiary education is extremely important, once again corroborating the results obtained for Latin America. In addition, two variables were included to determine the effect that population growth could have on ntrols for the density effect. According to this, broadband penetration. The population size variable co population size in a certain region could result in attractive markets, which would attract firms to the given region, therefore driving economic growth. The co efficient in this case was statistically significant containing the right sign. On the other hand, acco rding to the traditional development theory, the population growth variable was included. In this case, once again, the coefficient was positive and statistically significant. Dominican Republic: The contribution of broadband to employment growth 4.1.2.3 To study the contribution of broadband to employme nt in the Dominican Republic, a database for 32 provinces was constructed (see Table 28). Table 28 – Dominican Republic: Variables used to estimate the broadband impact on employment growth Type of Variable Source Rationale Data Set Change in Difference in unemployment Dependent variable National Statistical Office rate between 2009 and 2008 unemployment Population growth rate 2009- National Statistical Office Control unemployment Population growth 2008 by population growth Indotel Average fixed broadband Effect of broadband Broadband penetration 2008-2009 penetration penetration Change in number of Change in number of National Statistical Office Control of sources of establishments establishments 2008-9 new jobs National Statistical Office Share of private investment in Construction value Relationship between construction sector of Province. Investment and unemployment 2008-2009 Share of square metres of National Statistical Office Relationship between Construction area Investment and private constructions of province 2008-2009 unemployment Source: Author The model results are as follows (see Table 29). Table 29 – Dominican Republic: Broadband impact on job creation Coef. Growth in unemployment t P>t 95% Conf.Interval Std. Err. Population growth 0.72442 0.24939 2.90 0.0070 0.21180 1.23704 –0.29529 –0.02211 Change in broadband 0.0350 –0.56846 0.13290 –2.22 penetration Change in number of –0.14959 0.04728 –3.16 0.0040 –0.24678 –0.05241 establishments 0.39469 0.99443 Value of construction 2009 0.69456 0.14588 4.76 0.0000 –0.38015 Change in construction 2008-9 –0.64299 0.12787 –5.03 0.0000 –0.90583 1.51111 –0.02477 Constant 0.74317 0.37360 1.99 0.0570 38

51 Impact of broadband on the economy 32 Number of Obs F(5,26) 12.70 Prob>F 0.0000 R2 0.4175 Source: Author oadband on unemployment. According to Table 40, an The results show an extremely high impact of br increase in broadband penetration of 1 per cent would diminish unemployment in 0.29 percentage points. For example, if the unemployment rate were to be 14 per cent, an increase of 1 per cent in broadband penetration would contribute to a reduction of unemployment to 13.71 per cent. The other variables that indirectly affect unemployment are, as expected, number of establishments intensity of the construction sector in a specific area in 2009. Therefore, between 2008 and 2009, and the lishment, investment in construction and broadband a combination of increase in the number of estab yields a positive effect in terms of job creation. It is considered that the contribution of broadband rela tive to the other two variables is too high. Part of in broadband has taken place in Santo Domingo, the this is due to the fact that the largest increase ne the impact of broadband it would be necessary to capital, and Altagracia, a tourism hub. Ideally, to refi include in the model specification a variable related to the intensity of the tourism sector. However, while data exists for three regions, Altagracia, Puerto Plata and Cibao Norte, data for the rest of the country is not captured. Therefore, it is impossible to introduc e a variable measuring the intensity of tourism. As a result, the case confirms the contribution of broadband to job creation, although the range of impact might be overestimated. ic Impact in Arab States 4.2 Assessment of Broadband Econom The purpose of this section is to estimate the economic impact of broadband on a selected group of Arab States. The universe of analysis comprises the following countries: • Algeria • Saudi Arabia • Kuwait • Bahrain • Lebanon • Syria • Libya • Djibouti • Tunisia • Egypt • Morocco • United Arab Emirates • Iraq • Oman • Yemen • Jordan • Qatar Two methodologies were utilized. In the first place, a multivariate regression model was built to determine the impact of broadband adoption on the GDP growth of a cross-section of Arab countries. Secondly, case studies supported by descriptive statis tics were developed for Jordan, Qatar, the United Arab Emirates and Saudi Arabia. 4.2.1 Multivariate Regression Analysis: 69 This analysis is based on assessing the impact of broadband penetration growth between 2004 and 2010 on growth of the GDP, measured in constant 2005 pric es. For that purpose, a model based on a classical 69 In this analysis, given that the da taset extended through 2010, it was co nsidered necessary to include mobile broadband connections. 39

52 Impact of broadband on the economy production function, complemented with control vari ables to account for country specific factors, was specified. The following variables were utilized (see Table 30): omic impact of broadband in Arab States Table 30 – Variables used to measure the econ Observations Variable Sources Series World Bank and Dependent variable Per capita GDP (in constant 2005 Economic growth IMF prices) growth between 2004-10 Broadband Broadband penetration growth Independent variable ITU 2004-10 (including fixed wired and penetration growth fixed wireless broadband connections) World Bank Measure to determine size of Control for Population growth 2004-10 population population growth Control for Secondary school enrollment rate Measure to determine the World Bank 2000 human capital importance of human capital World Bank Measure to determine point of GDP per capita in year 2000 Control for level of development departure in economic growth Measure to control for the World Bank and US Annual change in price of oil barrel Control for Department of changes in oil impact of changes in oil prices in as a per cent of annual oil exports 2004-10 prices the gross product of Arab States Energy Gross capital formation (as per World Bank Measure to determine the Capital cent of GDP) 2004-10 importance of fixed capital investment Measure to determine the Dreher et al. (2008) Control for level Average globalization index 2004-8 of globalization (estimated 2009-10) different level of economic, social and political integration Control for bank World Bank Measure to determine the Annual domestic credit provided by credit the banking sector (as per cent of importance of financial markets GDP) 2004-10 Source: Author. The regression model was built on time series from 2004 (which includes the observations corresponding to 2010. For this purpose, the model considers only to the growth of variables between 2003 and 2004) 70 those annual observations for countries where broadband penetration was higher than 1 per cent . As a result, the sample considered in the regression model included the following years and countries (see Table 31). Table 31 – Arab States: Observations with a Broadband Penetration Rate Higher than 1 per cent 2004 2005 2007 2008 2009 2010 2006 0.41 % 2.54 % Algeria 0.11 % 2.34 % 0.51 % 0.85 % 1.41 % Bahrain 2.23 % 2.96 % 4.76 % 7.35 % 8.87 % 12.27 % 1.50 % Djibouti 0.00 % 0.01 % 0.02 % 0.13 % 0.29 % 0.61 % 0.91 % Egypt 0.11 % 0.19 % 0.34 % 0.62 % 0.98 % 1.35 % 1.79 % 70 In the case penetration is lower than 1 percent, changes in growth rate of penetration are not significant given the low variance in the number of total connections. 40

53 Impact of broadband on the economy 2006 2007 2008 2009 2010 2004 2005 0.00 % Iraq 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.44 % 1.57 % 2.49 % 3.95 % 4.73 % 0.20 % Jordan 0.88 % 1.28 % 1.43 % 1.57 % Kuwait 0.91 % 1.68 % 1.10 % 1.70 % 2.00 % 4.64 % 4.64 % 4.68 % 4.69 % 4.73 % Lebanon 3.21 % 0.00 % 0.16 % 0.16 % 0.75 % 1.01 % 1.15 % Libya 0.00 % 0.21 % 0.82 % 1.28 % 1.54 % Morocco 1.54 % 1.62 % 1.57 % Oman 0.54 % 0.81 % 0.79 % 1.23 % 1.52 % 1.63 % 0.03 % 1.53 % 3.12 % 7.44 % 7.61 % 9.05 % 8.38 % Qatar 4.78 % 0.30 % 0.88 % 2.44 % 4.01 % 5.36 % 6.23 % S. Arabia 0.28 % 0.00 % 0.01 % 0.03 % 0.04 % Syria 0.17 % 0.33 % 0.06 % Tunisia 0.18 % 0.44 % 0.95 % 2.22 % 3.60 % 4.60 % 0.03 % 1.53 % 3.18 % 5.16 % 7.03 % 8.98 % 9.95 % 10.51 % UAE Yemen 0.35 % 0.00 % 0.01 % 0.01 % 0.05 % 0.11 % 0.23 % Source: ITU World Telecommunications/ICT Indicators Database, www.itu.int/icteye above), the regression coefficients were calculated Based on the 60 observations (highlighted in Table 31 as follows (see Table 32): the Arab States using a Linear Multivariate Table 32 – Assessing the Broadband Economic Impact in Regression Model Standard T-Stat Coefficient Growth in per capita GDP Confidence Interval at P>[t] 71 95% error 0.0207588 0.0109596 1.89 0.064 –0.0012435 0.0427612 Broadband penetration growth Control for population growth – 0.1529595 –3.75 0.000 –0.8812873 –0.2671294 0.5742083 Control for human capital 0.0301354 0.97 0.338 –0.0313306 0.0896679 0.0291687 Control for level of development 0.0098631 –1.15 0.257 –0.0311088 0.0084933 – 0.0113077 Control for changes in oil prices 0.1917185 0.0475012 4.04 0.000 0.096356 0.2870811 Control for Gross Capital 0.0982635 2.86 0.006 0.0837325 0.4782767 0.2810046 Formation – 0.1700399 –0.08 0.935 –0.3554065 0.3273322 Control for level of globalization 0.0140371 0.0227318 Control for bank credit 1.68 0.100 –0.0044809 0.0499445 0.0135549 Constant –7.846972 12.02959 –0.65 0.517 –31.99738 16.30343 60 Number of observations F (8,51) 8.73 Prob>F 0.0000 2 0.5209 R Root MSE 3.7653 Source: Analysis by the author. 71 The two values in the confidence inte rval represent one for the low interval a nd the other for the upper interval at the 95% confidence interval. 41

54 Impact of broadband on the economy According to the model, an increase in broadband penetration of 10 per cent results in an average increase of 0.208 per cent in per capita GDP (measured in real terms). All variables in the model have the correct signs in the sense that they are directly or inversely related to the penetration of broadband: • Population growth is indirectly related to GDP per capita, indicating that the growth in population nts; this result is consistent with that of the reduces the product to be shared by all the inhabita Latin American broadband impact model (coefficient: -0.4469177); • There is a positive relation between human capital (as indicated by educational level) and GDP; however, the coefficient is not statistically significant, probably due to the unreliability of available data; • The more developed countries exhibit lower GDP per capita growth rates (which is consistent with Solow’s hypothesis); • An increase in oil prices has an important and st atistically significant impact on GDP, given that the value of exports tends to increase accordingly; the coefficient for this variable is, as expected, holding the highest significance; • Gross Capital Formation is, consistent with the production function, positively and significantly related to GDP; • Similarly to the results in Katz (2010), the globa lization index appears to have a negligible impact in the countries being studied; • A more developed credit market has a positive effect on GDP growth. In order to conduct a robustness test of the previous linear multivariate regression model, another model using fixed effects by year and country also using robust standard errors was developed. In the new approach, the fixed controls capture the variables that change for all the countries by year (per example the price of the oil), and the specific characteristics of the countries analyzed (education, Level of e only control variable utilized is the growth of the development, etc.). As a result of this approach, th population (with the same dataset that in the previous model). Table 33 – Assessing Broadband Economic Impact in the Arab States using a Fixed Effects by Year and Country Model Growth in per capita Coefficient Standard T-Stat P>[t] Confidence Interval at 95% error GDP Broadband penetration 0.0184833 0.0124923 1.48 0.165 -0.008735 0.0457016 growth -3.35 Control for population -0.1713994 -0.4901746 0. 1463068 0.006 -0.8089498 growth 1.031105 2.60 0.023 0.4295504 4.922719 Constant 2.676135 Number of observations 60 F (2,12) 7.62 Prob>F 0.0073 2 R 0.2279 Source: Analysis by the author. 42

55 Impact of broadband on the economy The result of the regression model with fixed effects shows a coefficient similar to the one yielded in the first approach (a 10% of difference). As a result, an increase in 10 per cent in the penetration of broadband, we should have a 0.185 per cent increase in the per capita GDP. for Arab States, the impact of broadband penetration In summary, according to the results of both models ta GDP is between 0.18 pe r cent and 0.21 per cent. growth of 10 per cent, in the growth of the per capi 4.2.2 General Impact Model Based on the general impact model presented in Table 32, the contribution of broadband to GDP growth for each Arab country was calculated as follows: 72 Count the number of years that broadband penetration is >1% • (column 1) Calculate the total growth between the first year of broadband >1% and 2010 (column 2) • • Calculate CAGR between the first year of broadband >1% and 2010 (column 3) • Multiply CAGR by broadband contribution coefficient to GDP per capita growth per the general model (0.0207588) (column 4) • Multiply results of column 4 by the number of years that broadband penetration level is >1% (column 5) The results are included in Table 34. Table 34 – Arab States: Contribution of broadband penetration growth rate to GDP per capita Years of Growth of Broadband Annual Impact to Total Impact on GDP per CAGR Country Penetration GDP per capita capita for the period penetration >1% 695% 41% 0.86% Bahrain 7 6.00% UAE 7 587% 38% 0.79% 5.50% Qatar 7 448% 33% 0.68% 4.76% 3.69% Jordan 4 201% 44% 0.92% S. Arabia 4 37% 0.76% 3.05% 155% 3 44% 0.91% 2.74% Tunisia 107% 137% 15% 0.32% 2.24% Lebanon 7 Algeria 80% 34% 0.71% 2.13% 3 Egypt 2 33% 33% 0.68% 1.35% Kuwait 6 53% 9% 0.18% 1.10% Oman 3 33% 15% 0.31% 0.94% 5 27% 6% Morocco 0.64% 0.13% Libya 2 14% 14% 0.29% 0.58% Note: It should be noted that some countries, such as Djibouti, Iraq, Syria and Yemen never reached 1 per cent penetration and therefore, the broadband economic impact is 0 per cent. Source: ITU, analysis by the author As Table 34 indicates, broadband has had a substantia l contribution to GDP growth in the Arab region, ranging from 6 per cent of total GDP growth for Bahrain between 2003 and 2010 to 0.58 per cent between 2005 and 2010 for Libya. This finding confirms the rule of “return to scale” where the contribution of broadband to GDP per capita growth increases with penetration. From a policy 72 See comment in footnote 69. 43

56 Impact of broadband on the economy standpoint, governments need to fast-forward the depl oyment of broadband if they want to maximize its economic impact. 4.2.3 Case Studies In the following section, three Arab countries have been analysed in further detail in an attempt to ascertain the specific economic impact of broadba nd on those countries. Three of them, United Arab Emirates, Bahrain and Qatar, exhibit the highest le vel of broadband adoption among Arab countries. The low adoption level until 2003 but has been growing at a other one, Jordan, is a country that exhibits a very fast rate since then (see Figure 10). Figure 10. Arab States: Fixed Broadband Penetration (2002-10), in per cent 14 12 10 8 6 4 2 0 2 0 07 2 0 08 2 0 09 2 0 05 2 0 04 2 0 03 2 0 10 2 0 06 Algeria Jordan Iraq Egypt Djibou ti Ba hrain Qatar Oman Lebanon Morocco Kuwait Libya Saudi Arabia Tunisia Yemen United Arab Emirates Syria Source: ITU; analysis by the author As indicated in Figure 10, Qatar and UAE are in top th ree Arab countries that exhibit the highest rate of penetration growth, while Jordan achieved an adoption of higher than 3 per cent in four years. Table 35 73 presents a comparison of penetration rates and CAGR in two different periods . As Table 34 indicates, both Qatar and the UAE succeeded in maintaining a high CAGR throughout the whole period under study (51% and 42% respectively), while Jordan placed itself in a good relative position when considering fixed wireless connections . The impact of broadband on selected economic descriptive statistics will be analysed in turn. 73 In addition, 2003 represents a key year in the developm ent of broadband in the Arab states as it embodies the combination of multiple trends, ranging from industry liberalization and competition to economic development resulting in added purchasing power to Internet adoption. 44

57 Impact of broadband on the economy Table 35 – Arab States: Comparative Broadband Penetration Growth, CAGR and increase of penetration rate CAGR Increase in Penetration Rate CAGR Country 2003-2010 (in %) 2003-2010 2006-2010 Algeria 2.4800 71% 49% 10.77753 35% Bahrain 27% Egypt 1.7100 51% 56% 4.6335 52% Jordan 74% 16% Kuwait 1.0700 7% Lebanon 2.9500 15% 0% 6% Morocco 1.6136 107% Oman 1.6200 19% 107% 51% 15% Qatar 7.9234 Saudi Arabia 62% 63% 6.0229 Tunisia 4.5973 189% 80% United Arab Emirates 9.6161 42% 19% Source: ITU; analysis by the author 4.2.3.1. United Arab Emirates: The United Arab Emirates exhibits a CAGR of 42 per cent in broadband penetration between 2003 and 2010. However, this rate masks an expected gradual decline in the broadband penetration growth rate (see Figures 11 and 12). Figure 11. U.A.E. Broadband Penetration (2003-10) Source: ITU; analysis by the author Despite the gradual decrease in the broadband growth rate, a comparison with the GDP growth rate indicates a high level of correlation: 84 per cent. The parallel evolution of growth rate in both variables can be clearly seen in Figure 12. This relationship points to a probable dual causation, whereby broadband has an impact on the economy, while economic growth contributes to the adoption of broadband as well. 45

58 Impact of broadband on the economy Figure 12. UAE: Growth rate of br oadband penetration and GDP per capita Source: World Bank; IMF; ITU; analysis by the author While the number of observations for the specific ca se study is not sufficient to ascertain the causality between both variables, we can apply the coefficient of the general model for Arab States to estimate the 74 . By relying on the model specified for all Arab States, economic contribution of broadband in the UAE we can estimate that the average annual contribution of broadband to per capita GDP growth between 75 . 2004 and 2010 is 0.79 per cent 4.2.3.2. Jordan: 10 per cent. However, after growing its broadband Jordan’s broadband penetration in 2003 was only 0. deployment at an annual rate of 100 per cent between 2003 and 2007, the country reached a penetration of 1.57 per cent. While the growth rate declined after that year, the country doubled its penetration by 2009 if we only count fixed connections. When wire less connections are considered, Jordan’s broadband penetration in 2010 is approximately 5 per cent (the fifth higher in the region). 74 The general model remains statistically significant becaus e the analysis of multiple countries generates a sufficient number of observations. 75 This estimate considers the CAGR of broadband penetration between 2004 and 2010, because in 2004 the UAE surpasses the barrier of 1% penetration. The average CAGR for that period is 38%, which is then multiplied by the owth estimated in the mode l of figure D-3 (0.0207588) coefficient of GDP impact of broadband penetration gr 46

59 Impact of broadband on the economy Figure 13. Jordan. Broadba nd Penetration (2003-10) Source: ITU; analysis by the author The growth in broadband penetration and GDP per capita are fairly correlated: 76 per cent. This is clearly depicted in Figure 14. oadband penetration and GDP per capita Figure 14. Jordan: Growth rate of br Source: World Bank; IMF; ITU; analysis by the author As shown in Figure 14, broadband growth rate is related to GDP, in some cases with a two-year lag. This lag can be, in some cases, explained by the complex se t of causal links existing between the two variables. of broadband, the effect emerges only after For example, when assessing the economic impact 47

60 Impact of broadband on the economy broadband has resulted in changes in the adoption of e-business processes in enterprises (a process that requires more than one year to occur after broadband is deployed). In order to estimate the contribution of broadband to economic growth, only the years when penetration exceeded the floor of 1 per cent were consider ed (2007 to 2010). The CAGR for that period was 44 per cent, which, when multiplied by the broadband penetration growth coefficient of the model included in Figure 14, yields an average annual increase of GDP per capita of 0.92 per cent. 4.2.3.3. Qatar: Qatar is the Arab country with third highest broadband penetration growth rate by 2010 (behind Bahrain and the UAE). It was one of the first countries in the Arab region that surpassed the barrier of 1 per cent penetration, having continued its growth trend unt il reaching almost 10 per cent by 2010. Between 2003 and 2007, Qatar has been doubling the number of broadband lines year over year. After 2007, although the growth rate declined significantly, the penetration continued to grow steadily. Thus, between 2003 and 2010, the penetration rate CAGR was 51 per cent and the penetration reached over 8 per cent (see Figure 15). Figure 15. Qatar: Broadba nd Penetration (2003-10) Source: ITU; analysis by the author Contrary to the other two case studies, the GDP per capita and broadband penetration growth rates are not significantly correlated, since oil prices are a significant driver of GDP growth. In the case of Qatar, oil prices are, as expected, the primary and most signific ant driver of GDP growth. This raises the relevance of having introduced oil prices in the broadband impact general model for Arab States. Considering that Qatar surpassed the broadband penetration floor of 1 per cent in 2004, the CAGR of broadband penetration between that year and 2010 was 35 per cent. By multiplying that number by the broadband coefficient in the Arab States general model (0.0185964), it is estimated that broadband contributed an average 0.65 per cent to the annual GDP growth. 4.2.3.4 Saudi Arabia: The impact of broadband on employment oyment database containing information for its To estimate the impact of broadband on Saudi empl 13 provinces was compiled. It was composed of the following variables: 48

61 Impact of broadband on the economy Table 36 – Saudi Arabia: Variables used to es timate the broadband impact on job creation Source Series Observations Variable Forty Fifth Annual Report – Unemployment Dependent variable Change rate in unemployment rate (2007-2008) Saudi Arabia Monetary Agency Broadband penetration per Broadband Independent variable Communications and (2007-8) Information Technology Δ population penetration Commission Variable to control for Facilities authorized to provide Annual Report – Saudi Infrastructure health services (2008) Arabia Monetary Agency differences in level development I infrastructure Tourism Annual Report – Saudi Independent variable Change in the number of domestic tourism trips (2007-2008) Arabia Monetary Agency development Variable to control for Annual Report – Saudi Percentage of households with Infrastructure development II access to potable water differences in level Arabia Monetary Agency infrastructure Government Change in the number of projects Annual Report – Saudi Independent variable spending I funded by the government (2007- Arabia Monetary Agency 2008) Independent variable Government Annual Report – Saudi Change in the value of projects funded by the government (2007- Arabia Monetary Agency spending II 2008) Source: Author The model results are as follows (see Table 37): Table 37 – Saudi Arabia: Broadband impact on job creation Unemployment rate Coefficient Standard error T-statistic P>[t] Change in broadband penetration per –0.2434 0.02935 –8.29 0.000 population 2007-8 0.899473 0.170589 5.27 0.002 Facilities authorized to provide health services (2008) Change in the number of domestic tourism 0.337701 0.093622 3.61 0.011 trips (2007-8) 0.13798 Percentage of households with public 4.73 0.652031 0.003 access to potable water Change in the number of projects funded –2.31629 0.761245 –3.04 0.023 by the government (2007-8) Change in the value of projects funded by –0.92765 0.341837 –2.71 0.035 the government (2007-8) 0.303 Constant –16.1099 14.306 –1.13 13 Number of observations F (6,6) 34.84 Prob>F 0.0002 2 R 0.9400 Root MSE 7.8648 Source: Author 49

62 Impact of broadband on the economy The independent variables (the change in broadband penetration, change in the number of domestic tourism trips, change in the number of publicly-funded projects and change in the value of publicly- funded projects) had a negative impact on the level of unemployment. We found that a 10 per cent increase in broadband penetration decreases the change in the unemployment rate by 2.4 percentage points. This coefficient seems high in comparison to those found in studies for other countries. It should also be noted that because of lack of reliable data , a human capital variable could not be included. In consequence, the broadband variable could be capturing the effect of the missing human capital variable. ized to provide health services and percentage of The control variables (the amount of facilities author households with public provisioning of potable water) had statistically significant coefficients with positive signs. As expected, two of the other variables consider ed (variation of the number of and value of projects funded by the government) had a negative impact on unemployment. However, we found an unexpected sign on the tourism coefficient: the increase in the number of domestic tourism trips increased the gap in unemployment. 4.3 Asia Pacific In the case of Asia-Pacific, the analysis of the economic impact of broadband was only conducted on a country basis, given the difficulty in capturing general effects in a cross-sectional model. 4.3.1 India: the impact of broadband on employment and economic growth In the case of India, a data availability problem si milar to the one encountered for Brazil was identified. Therefore, the analysis had to be conducted at an aggregated level, which impacted the robustness of estimates. The information regarding broadband penetration is reported according to the telecommunications circles in which the Indian territory was divided during the privatization process. Each circle may represent only one province, (e.g., the Assam Telecom circle only includes the Assam province), or may include more than one province, (e.g., the Maharashtra Telecom circle includes the Maharashtra and the Goa 76 provinces) . The definition of circles has changed on several occasions, but in this study the definition of 77 2006 has been used . In order to estimate the impact of broadband on I ndian employment, a database containing information 78 for India's 20 circles was compiled . It contained the following dataset: Table 38 – India: Variables used to estimate the broadband impact on job creation Variable Series Source Observations Indiastat Independent variable Δ Broadband Broadband penetration (2007-8) penetration Employment Employment growth (2008-9) Indiastat Dependent variable Control for Variable to explore the relationship GDP per capita by region (2005) Indiasta t economic of GDP and employment development 76 Bihar includes Jharkhand, Delhi includes Noida, Gugaon, Ghaziabad and Faridabad, Maharashtra includes Goa, Madhya Pradesh includes Chhattisgarh and Uttar Pradesh includes Uttaranchal. 77 Andaman and Nicobar Islands, Andhra Pradesh, Assam, Bi har, Gujarat, Haryana, Himachal Pradesh, Jammu and Kashmir, Karnataka, Kerala, Madhya Pradesh, Mahrashtra, New Delhi, North East, Orissa, Punjab, Rajasthan, Tamil Nadu, UP East, UP West, West Bengal. 78 of information made it necessary to While there were 20 telecom circles de fined by the government for 2006, lack consolidate two circles into one. 50

63 Impact of broadband on the economy Series Observations Source Variable Measure of Number of micro, small and Indiastat Variable to explore the relationship medium enterprises (2006) economic activity of employment and development of micro, small and medium enterprises Financial Number of banking offices per Indiastat Variable to co ntrol for differences in the level of banking infrastructure development I Lakh (100,000 of Population) 2002 Financial Bank credit per capita by region Indiastat Variable to control for differences in development II 2002 financial access Infrastructure ntrol for differences in Total road length per hundred Variable to co Indiastat sq. Km by area (km.) 2001 development level of infrastructure Average population growth by Population Variable to control for differences in Indiastat growth province 2001-11 population growth Source: Author. The model results are as follows (Table 39). Table 39 –India: Broadband impact on job creation Employment rate Standard error T-statistic P>[t] Coefficient Growth in broadband penetration .0282529 .0151963 1.86 0.090 2007-2008 Number of micro, small and medium 0.056 .0000461 .0000216 2.14 enterprises 2006 GDP per capita by region 2005 –.0002114 0.071 .0001056 –2.00 .2778866 1.56 0.147 Number of banking offices per Lakh .4333618 (100,000 of Population) 2002 –.0003247 .0001892 –1.72 0.114 Bank credit per capita by region 2002 .0102269 .0059689 1.71 0.115 Total road length per hundred sq. km by area (km.) 2001 Average population growth by province 1.73853 .7382998 2.35 0.038 2001-11 –5.14944 3.834851 Constant –1.34 0.206 19 Number of observations F (4,14) 1.16 Prob>F 0.3983 F (2,310) 0.4237 2 R 0.4237 2 Adjusted R 0.0570 Root MSE 1.6382 Source: Author. Four variables exhibited significant coefficients : growth of broadband penetration between 2007 and 2008, the number of small and medium enterprises in 2006, GDP per capita 2005, (although the negative pulation growth. According to the model, an increase sign lends itself to different interpretations), and po 51

64 Impact of broadband on the economy in 1 percentage point in broadband penetration growth results in 0.028 percentage points increase in the employment rate. ation level of 0.64 percentage points in 2009. According to ITU, India reached a broadband penetr Considering that India’s labour force has already passed the 500 million people frontier, this model jobs in direct and indirect ways. This result suggests that broadband has generated nearly 9 million becomes more important taking into consideration the latest estimates provided by the Reserve Bank of India forecasting an increment of 220 million to India’s workforce by 2030. In addition to estimating the impact of broadband on job creation in India, its contribution to the se, a dataset for India's 20 circles was built (see country's GDP growth was estimated. For this purpo Table 40): Table 40 – India: Variables used to measur e the impact of broadband on GDP growth Series Source Observations Variable Economic growth Per capita GDP growth by Indiastat Dependent variable region 2007-2008 Variable to control the contribution Total road length per Infrastructure Indiastat hundred Sq. Km. of are (Km.) of physical capital to economic development growth 2001 Broadband penetration Indiastat Independent variable Broadband penetration Δ (2007-8) Participation rate in Variable to control for contribution Indiastat Human capital of human capital to economic secondary schooling 2001 growth Population growth Average population growth Indiastat Variable to control for differences by province 2001-2006 in population growth Variable to control pre-existent Indiastat Point of departure of GDP per capita by region economic growth 2000 level of economic development Source: Author. The results are as follows: Table 41 – India: Contribution of broadband to GDP growth GDP Growth (2007-8) Coefficient Standard error T-statistic P>[t] .00300 .0006727 4.46 0.001 Infrastructure development Growth of broadband penetration .0158414 1.97 0.070 .031284 2007-2008 .0133936 .0803297 0.17 0.870 Human capital –0.85 Population growth by province 2001- 0.413 1.272319 –1.075241 2006 0.40 GDP per capita by region 2000 .0000278 .00007 0.698 0.725 Constant 2.753619 7.668025 0.36 19 Number of observations F (5,7) 8.02 Prob>F 0.0012 2 R 0.3241 2.2265 Root MSE Source: Author. 52

65 Impact of broadband on the economy Broadband penetration was found to be statistically sign ificant and exhibited the expected positive sign. According to the coefficient of this variable, a 10 per cent increase in penetration will result in an increase of 0.3128 percentage points in regional GDP. The GDP 2003 variable was included to control for the starting point of development. Its coefficient sign is positive and not significant, which according to the be interpreted as a sign of divergence among the regions (circles) of conditional convergence theory could India. The results of this model should be carefully interprete d because of the potential bias. It is possible that the casual relationship between dependent vari able, (GDP growth 2007-2008) and the independent 79 variable, (broadband penetration growth 2007-2008) runs in both directions, which would yield biased and inconsistent estimates of the structural para meters. This endogeneity problem seems to cause a strong upwards bias, producing a high coefficient in comparison to those of other studies and prior models. In conclusion, while the econometric models of broadband impact on Indian employment have yielded strong estimations, the one focused on understanding the contribution of the technology to GDP growth has to be discarded until better data sets are available. 4.3.2 Malaysia: The contribution of broadband to economic growth nd to GDP growth in Malaysia. For this purpose, the This case study estimates the contribution of broadba following data was collected (see Table 42): ure the impact of broadband on GDP growth Table 42 – Malaysia: Variables used to meas Variable Series Source Observations Economic growth GDP Growth 2007-2008 Department of Dependent variable Statistics – Malaysia Independent variable Broadband Under Secretariat ICT Broadband household penetration penetration Policy Division Δ 2006-7 Health Hospitals per million population Ninth Malaysia Plan Variable to control for infrastructure I (2006-2010) physical capital 2005 Variable to control for Health Beds in hospitals per million Ninth Malaysia Plan population 2005 (2006-2010) infrastructure II physical capital Infrastructure Construction projects funded by Ninth Malaysia Plan Variable to control for the government 2004 development (2006-2010) physical capital Economic activity Added value of construction Ninth Malaysia Plan Variable to study the (2006-2010) relationship between sector 2004 government funding and economic development Human capital Literacy rate 2000 Ministry of Education Variable to control for human capital Control for Road development index 2005 Ninth Malaysia Plan Variable to control for (2006-2010) physical capital infrastructure Source: Author. 79 Though we are aware of this problem, th e lack of information in this case made estimation by Instrumental Variables or Simultaneous Equations an impossible task. 53

66 Impact of broadband on the economy The results are as follows: Table 43 – Malaysia: Contributio n of broadband to GDP growth GDP Growth (2007-8) Standard error T-statistic P>[t] Coefficient Growth of broadband household 0.077024 0.013247 5.81 0.001 penetration 2006-2007 Hospitals per million population –0.54338 0.146033 –3.72 0.007 2005 0.007824 0.002203 Beds in hospitals per million 3.55 0.009 population 2005 0.0089983 –2.64 0.033 Construction projects funded by –0.02375 the government 2004 Added value of construction sector 0.005661 0.001693 3.34 0.012 2004 Literacy rate 2000 0.126412 6.24 0.000 0.789125 Road development index 2005 –10.6963 2.126227 –5.03 0.002 0.001 –5.66 Constant –52.836 9.339643 Number of observations 15 107.27 F (7,7) 0.0000 Prob>F 2 R 0.9010 1.1898 Root MSE Source: Author. The independent variable, growth in broadband penetrat ion per household, has a statistically significant coefficient with the expected positive sign. It indi cates that an increase of 10 per cent in broadband penetration will contribute to 0.7 percentage points to regional GDP growth. This result has to be put in a context of an economy that has a service sector contri buting more than 55 per cent of the GDP. It should be noted that this estimation, based on penetration per household, is lower than the 0.4 per cent impact inhabitant estimated for Malaysia by McKinsey and on GDP per 10 per cent of broadband penetration per 80 . Co Inc (2009) On the other hand, the control variables (hospitals per million populations, construction projects funded by the government, and road development index) ha d statistically significant coefficients and with negative signs. These indicate convergence between highly developed and lesser developed regions of the country. 4.3.3 China: The relation between broadband and economic growth To estimate the impact of broadband on China's econom ic growth a database containing information for its thirty provinces was compiled. It was composed of the following variables: 80 The same study forecasts that an increase of 10% in br oadband penetration per inhabitant would have an impact of 0.6% to 0.7% on the GDP of Asia. 54

67 Impact of broadband on the economy Table 44 – China: Variables used to estimate the broadband impact on economic growth Source Series Observations Variable Dependent variable All China Marketing Economic growth Regional economic growth Research (ACMR) between 2008-2009 GDP tertiary sector / Regional GDP ACMR Importance of a less Importance of tertiary sector (2009) agricultural economy Deposit in financial institutions ACMR Deposits Importance of relationship of financial services and divided by regional GDP (2009) growth Loans ACMR Variable to control for Loans in financial institutions divided by regional GDP (2009) impact of financial access on the economy Secondary institutions (10,000 pop) Education ACMR Relationship between growth and education 2000 Investment Local and foreign investment ACMR Relationship between divided by GDP (2009) Investment and growth Relationship between Gross capital Gross capital/GDP (2009) ACMR acquisition of assets and growth Fixed Capital/GDP (2009) ACMR Variable to control for Fixed capital impact of investment on the economy Euromonitor Relationship between Broadband Difference between broadband penetration in 2009 and 2008 Growth and Broadband penetration growth Source: Author. Unfortunately, the regressions run against this data se t did not yield statistically significant results. The author suspects that part of the reason could lie in some limitations encountered in measuring broadband penetration. Figure 16 clearly indicates that a correlation exists between broadband and economic growth for China's 30 provinces. Figure 16: China: Correlation between ec onomic growth and broadband deployment 0 y = 0.0000141x + 0.0480179 2 R = 0.7831 0 0 (2009) 0 0 Broadband Household Penetration 0 14000 0 2000 4000 6000 8000 10000 12000 GDP per capita in USD (2009) Source: Author 55

68 Impact of broadband on the economy Without pre-supposing a causal link, data in Figure 16 indicates some sort of relationship. The only further analysis that could be conducted allowed us to esti mate the importance that broadband penetration in driving the variance of the level of economic development of a Chinese province. The correlation model results are as follows (see Table 45). Table 45 – China: Correlation between br oadband deployment and economic growth Standard Dependent variable: GDPPP09 T-statistic P>[t] [95% Conf. Interval] Coefficient error Broadband penetration 2008-09 1.867034 1.081138 1.73 0.099 –0.38132 4.115384 0.644611 0.091923 7.01 0 0.453447 Importance of tertiary sector 0.835776 Deposits/GDP –8.71734 –6.21 0 –11.6376 –5.79712 1.404212 3.361515 0.209873 2.22 0.038 Loans/GDP 6.513158 1.515494 Secondary institutions 0.149747 0.055293 2.71 0.013 0.03476 0.264734 (10,000 pop) 2000 0.355613 0.157363 Investment/GDP 0.256488 0.047665 5.38 0 Gross capital/GDP 0.18138 2.07 0.051 –0.00105 0.753351 0.376151 Fixed capital/GDP –0.42741 0.182476 –2.34 0.029 –0.80688 –0.04793 Constant –30.3521 5.867167 –5.17 0 –42.5535 –18.1506 Number of Observations 30 F (8,21)= 10.45 Prob>= 0.0000 R2= 0.7180 Root MSE= 2.2161 Source: Author. According to the results of Table 37, broadband penetration appears to be related to economic development in a significant way. Without assuming a causal relationship, it is reasonable to conclude that broadband has some level of contribution to economic growth in China. It should be noted that this analysis only yields a correlation result between the broadband and the economic variable. Further analytical work is necessary to deepen our understanding of the relationship between the two variables in the Chinese context. 4.3.4 Indonesia: The contribution of broadband to employment growth To study the contribution of broadband to employment in Indonesia, a data panel analysis was implemented. The model was constructed using the unemployment growth rate as the dependent variable. The independent variables used in the model included the unemployment rate lagged one household, the years of education of the labour period, the average of the broadband penetration per force and the difference between the years of education between male and female economic active population (see Table 46). 56

69 Impact of broadband on the economy Table 46 – Indonesia: Variables used to estimate the broadband impact on employment growth Data Set Rationale Type of Variable Source Unemployment Annual change on the PerkembanganBeberapaIndikator Dependent variable growth rate UtamaSosialEkonomi Indonesia – open unemployment rate (2005, 2006, 2007, 2008, 2009, – per province 2010) (2006,2007,2008,2009) Unemployment Annual open Importance of relationship PerkembanganBeberapaIndikator rate UtamaSosialEkonomi Indonesia – unemployment rate per of unemployment rate of previous year and current (2005, 2006, 2007, 2008, 2009, province (2005, 2006, 2007, 2008) change in unemployment 2010) Importance of relationship Euromonitor Broadband Average broadband penetration of broadband and penetration per unemployment household (2005, 2006, 2007, 2008, 2009) Years of education of the Importance of relationship Human capital I PerkembanganBeberapaIndikator of education and UtamaSosialEkonomi Indonesia – male economic active (2005, 2006, 2007, 2008, 2009) unemployment population (2006, 2007, 2008, 2009) The impact of the difference PerkembanganBeberapaIndikator Human capital II Difference in years of in education between male UtamaSosialEkonomi Indonesia – education between male (2005,2006,2007, 2008,2009) and female economic active and female economic population on active population (2006, unemployment 2007, 2008, 2009) The model results are as follows (see Table 47). Table 47 – Indonesia: Broadband impact on job creation Dependent variable: Coefficient Standard 95% Confidence T-Statistic P>[t] Error unemployment growth Interval –23.8118 3.148876 –7.56 0.0000 –30.1063 –17.5173 Lag unemployment growth rate –8.61628 –2.58 Broadband penetration 0.0120 –15.2848 –1.94772 3.335997 Years education male 1.659112 –2.28 0.0260 –7.09309 –0.46006 –3.77658 Differences years education 7.233474 3.895974 1.669609 2.33 0.0230 0.558474 male/female Constant 7.28 0.0000 147.4395 259.0097 203.2246 27.90692 Number of observations 99 33 Number of groups F (4,62) 15.35 Prob > F 0.0000 Source: Author. between the lagged employment rate and the growth As expected, the results show a strong relationship in unemployment. Another variable important in the model is the number of years of education between the male and female population. Finally, the contribution of the broadband variable appears to be an extremely contributor to the reduction of unemployment, with a negative effect of -8.61 per cent. This means that for each 1 per cent increase in the penetration rate of the service among the Indonesian households, the unemployment growth would be reduced it by 8.61 percentage points. It is important to note that the entire effect of broadband on unemployment is a combination of ne w jobs and existing jobs ributed to the unemployment rate. saved that otherwise would have cont 57

70 Impact of broadband on the economy Analysis of Case Study Results 5 The results of the above analyses validate the positive contribution of broadband to GDP growth for developing countries and regions. While limited in the number of countries analysed, one can safely confirm that broadband has a directionally positive economic impact. The following chart compiles the results of the existing literature as well as this study's (see Table 48). While it is not optimal to compare model results acro ss geographic units, the fo llowing conclusions can be drawn. The case studies of Latin America and the Caribbean, the Arab States, Chile, India and Malaysia due to data limitations, some of the country cases yielded statistically significant coefficients. However, ich may jeopardize the consistency of estimators are handicapped by the low number of observations, wh (Chile and Malaysia). However, Koutroumpis (2009), Katz et al. (2010a), the cross-sectional Latin American ls exhibit a higher level of reliability. model , the cross-sectional Arab states mode Table 48 – Comparative estimate of broadband impact on GDP growth Study Region/Country Impact on GDP growth Observations ∆ for each 1% in of broadband penetration Statistically significant coefficient at • 0.023 5 OECD countries with 1% level penetration higher 81 than 30% • 132 Observations 0.014 Statistically significant coefficient at 8 OECD countries with Koutroumpis 1% level penetration between (2009) 20% and 30% • 132 Observations • Statistically significant coefficient at 0.008 8 OECD countries with 5% level penetration under 17% • 132 Observations 0.0256 Statistically significant coefficient at • High Developed Counties in Germany 1% level 214 observations • et al. Katz (2010a) Less Developed 0.0238 • Statistically significant coefficient at Counties in Germany 1% level 210 observations • Statistically significant coefficient at 0.138 • Countries of medium Qiang et al. the 10% level and low economic (2009) development • Human capital carries insignificant explanatory power, which indicates that most variation in her dataset is explained by the dummies and the constant Latin America and the 0.0158 • Statistically significant coefficient (t-statistic = 1.98) Caribbean 49 observations • 82 Present study Statistically significant coefficient (t- 0.02076 • Arab States statistic = 1.62) • 60 observations 81 Denmark, Norway, Netherlands, Sweden, Switzerland. 82 See Appendix D 58

71 Impact of broadband on the economy Study Region/Country Impact on GDP growth Observations ∆ for each 1% in of broadband penetration • Statistically not significant coefficient Brazil 0.008 (t-statistic = 0.16) • 27 observations Chile 0.009 • Statistically significant coefficient (t-statistic = 2.04) 13 observations • 0.031 India Statistically significant coefficient • (t-statistic = 1.97) • 19 observations • Potential endogeneity 0.077 • Statistically significant Malaysia • coefficient (t-statistic = 5.81) • 15 observations China • Statistically not significant coefficients 30 observations • Indonesia • Statistically not significant coefficients 93 observations • Source: Author. As pertains to the results, Qiang et al. (2009) and the Malaysia case study appear to over-estimate the amount of impact of broadband on GDP growth. In the case of Qiang, the overestimation could be the that there is a two sided relationship between the result of a number of factors. First, it is likely dependent and independent variables. In this study the dependent variable is the average growth rate for 1980 to 2006 while the independent variable is the average broadband penetration for 1998 to 2006. The authors rely on the initial level of penetration as in struments, but as the literature on the diffusion of broadband shows both variables, broadband penetratio n and initial level of penetration, tend to be 83 . Second, the authors presented several specifications each of them including one determined by income type of communications technology (e.g. One model only included fixed telephony, another one only included mobile telephony). In all of these models the variables, fixed telephony, mobile telephony and Internet penetration were significant and positive. Thus, it would be probable that the variable broadband penetration 1998-2006 would be functioning as a techno logical index capturing the effect of technologies that contributed to the growth between 1980 and 2006 but were not included in the model that measures the impact of broadband. In the Malaysian case study, part of the overestimation could result from the fact that broadband penetration data is calculated on a per household basi s, rather than per population. Given that the ratio of population to household is roughly 1:3, the approp riate coefficient for Malaysia could be 0.026, which is closer to the range of case study results. All in all, it is fair to conclude and reaffirm the positive contribution of broadband to economic growth with a clear "return to scale" effect. Thus, in count ries with higher broadband penetration, 1 per cent growth in broadband penetration results between 0.023 per cent (Koutroumpis, 2009) and 0.026 % (Katz et al., 2010a) to GDP growth. On the other hand, in the countries with low broadband penetration, the 83 (2009) and Garcia-Murillo (2005). See Sangwon L and Brown J. 59

72 Impact of broadband on the economy es between 0.008 per cent (Koutroumpis) and 0.0158 per cent to contribution to GDP growth rang 0.02076 of the current study. The results of the analyses tend to be more conclusi ve when it comes to the positive contribution of broadband to employment creation for less develo ped countries and regions. The following chart compiles the results of the existing literature and this study's (see Table 49). In this case, all prior research, as well as the results of this study, coincide and indicate that broadband has a positive impact on job creation. In particular, the German study and the Indonesian and Chilean cases, which are based on extensive datasets, yield stat istically significant positive coefficients. The other cases (India, Brazil and Saudi Arabia) have also yiel ded statistically significant coefficients for the explanatory variable (broadband penetration) with sensible signs, i.e. positive when the independent variable is employment and negative when it is unemployment. However, the reduced number of observations (particularly for India and Saudi Arab ia) significantly reduces the models' robustness. Table 49 – Comparative estimate of broadband impact on employment growth •Observations Study Region/Country Impact on job creation for each 1% increase in rate of growth of broadband penetration Employment Unemployment Statistically significant level at Katz et al. • High developed 0.0061 counties in Germany 1% (2010a) • 214 observations • Statistically significant level at et Kentucky, USA 0.14-5.32 Shideler al. (2007) 1% (total employment) Significance level varies • depending on the industry analysed- from not significant to significant • 114 observations (total). Range varies according the industry analysed (from 20 to 120) Brazil -0.0449 • Statistically significant coefficient (t-statistic=1.73) • 27 observations Chile 0.181 • Statistically significant coefficient (t-statistic = -8.29) 13 observations • • Statistically significant India 0.02825 coefficient (t-statistic = 1.86) • 19 observations Present study Saudi Arabia -0.2434 • Statistically significant • coefficient (t-statistic = -8.29) 13 observations • Indonesia -8.6163 Statistically significant • coefficient (t-statistic = -2.58) • 99 observations Dominican Republic -0.2952 • Statistically significant coefficient (t-statistic = -2.22) 32 observations • Source: Author. 60

73 Impact of broadband on the economy Estimation of Broadband Gaps and Investment Requirement 6 The chapter above provided an assessment of the economic returns, both in terms of growth and employment creation, of broadband deployment. It is ob vious that the construction of the infrastructure comes at a cost, which needs to be calculated. Unders tanding the cost of broadband deployment is critical to develop public policies that provide for realistic targets based on economic considerations. Policy makers have taken three approaches to the issu e of investment costs calculation. The first is the conventional engineering one, which is based on estimating the coverage requirements, and then using those estimates to project the necessary investment to fulfil them. This is the methodology followed for the investment estimation of Australia's National Broadband Plan. The second approach, labelled “top- down”, begins by first determining the amount of fina ncial resources to be invested and then sizing the amount of coverage that will be achieved given those resources. To some degree, this is the approach that has been followed in the United States with the Broadband Technology Opportunities Program. Since this was part of the stimulus package passed by th e US Congress, the funds would be assigned through grants with the specific construction plans defined as they are given out. The third approach does not estimate an investment amount. Labelled the “public policy” framework, it defines targets, (such as coverage and speeds), but leaves the amount of investment required unaddressed. This is the approach followed in Germany's and Brazil's national broadband pl ans, and, to some degree, in the United States. The following section presents a methodology that de termines investment requirements. The objective is not to provide extremely precise estimates but to ga uge the investment in broadband required in order to have a sense of the resultant social and economic returns. 6.1 Methodology The methodology used to estimate broadband investment is based on the assumption that policy makers have defined targets in their national broadband plan . As mentioned above, these targets generally refer to metrics such as coverage, adoption, and speed. In other words, the plan defines the percentage of the population that needs to be served by broadband, and the specific service level at which it needs to reached, (typically in terms of download speeds.) ainst the current situation of broadband deployment. Once defined, the targets have to be compared ag level (municipalities, counties, administrative Data on current coverage at a disaggregated enough departments) is in many cases not available either because the government does not keep accurate records of coverage and service level or because th e private sector considers this information to be proprietary for competitive reasons. Some countries, however, have public records with excellent quality 84 n and the targets allows estimation of the . The comparison between the current situatio of information deployment objectives. Using these figures it is po ssible to calculate the number of broadband access lines to: 1) cover the “white” spots (unserved areas), 2) upgrade the “grey” spots (areas with inferior service measured by low access speeds), and 3) deploy high speed lines in priority areas. Once the number of lines per service target is estimated, they are multiplied by the costs per broadband line. This last step should take into account the type of platform. In order to determine the costs per line, the estimates from deployment experience in Europe and the 85 . These costs need to be adjusted by factors such as urban density, United States can be utilized 86 economies of scale, and experience curve . This calculation yields the total investment required for wireless and wireline technologies. The total investment is then split according to three cost categories: 1) 84 For example, Germany’s Federal Mini stry of Economics and Technology publis hes the Broadband Atlas (BMWi 2009b). 85 It is important to realize that these figures may not tran slate exactly in cross country co mparisons due to difference in labor markets, in particular construction wages. 86 See, in particular, Elixman et al. (2008). 61

74 Impact of broadband on the economy 3) telecommunications labour. These splits are based construction labour, 2) electronic equipment, and data for NGAN (furnished by a European operator), on cost allocations based on “real life” deployment and for 3G and WiMAX for a US operator. The next two sections will review examples of investment requirement calculation. The case studies to be 87 analysed include Germany's National Broadband Plan and Brazil's Broadband Plan. 6.2 The National Broadband Plan in Germany According to estimates based on publicly available da ta published in the national broadband strategy, of 88 all German households , 39 million have access to some type of broadband technology. Of these, 36.7 million have DSL capability, 22.0 million are passed by cable TV networks (and, therefore potentially connected via a cable modem), and 730,000 can access the Internet via fixed wireless or satellite technology. Furthermore, it is estimated that 10.9 m illion households are able to connect via VDSL, while only 240,000 could be based on fibre to the home (FTTH). These estimates are based upon announcements of fibre optics technology bein g deployed by telecommunications operators and municipal networks. 89 , Germany plans to raise the access speed of residential broadband In addition to improving coverage users. As of 2009, the national broadband strategy reports that 98 per cent of all German households have access to broadband Internet with transmission rates of at a minimum 384 Kbps and 92 per cent of households are served by lines with at least 1 Mbps. About 2.8 million households are in “grey spots”, meaning that they have broadband access between 384 Kbps and 1 Mbps. The remaining “white spots”, which comprise 2 per cent of households (or 730,000), are located in areas with low population densities or near the outer boundaries of already connected areas. Furthermore, Germany is undergoing three levels of fibre deployment: 1) Fibre to the Main Distribution Frame for ADSL 2+ services for selected cities, 2) Fibre to the street cabinet for VDSL services, and 3) Fibre to the home (DTAG, NetCologne, M’net, and Stadtwerke Schwerte). In the case of VDSL, Deutsche Telekom has announced that 90 per cent of households located in top 50 cities (which are estimated to be 10.9 million) can have access to VDSL. Simultaneously, the municipal networks have launched or are planning to deploy FTTB networks. Deployment is focused on densely populated areas of their home markets. Total coverage, based on these targ ets under construction, is of 240,000 homes. The major cable players have all upgraded networks to two way Hybrid Fibre Coax, generally relying on DTAG ducts. Kabel Deutschland’s acquisition of Orion in the first quarter of 2008 resulted in 7.6 million upgraded households passed (71 per cent of ho mes passed). Unitymedia has 6.3 million upgraded households passed (72 per cent of homes passed). Kabel BW has upgraded 91 per cent of its network, which 3.3 million homes. All players are offering DOCSIS 2.0 and plan to roll out DOCSIS 3.0 by the end of 2010. Turning to the demand side, as of the end of 2008, there were 22.6 million broadband lines in Germany (Source: Bundesnetzagentur Jahresbericht 2008, 2009) . Based on this statistic, Germany has reached 58 per cent household penetration, or 27 per cent of its population. In light of the supply perspective we see that 58 per cent of households actually serv ed by any combination of broadband technologies purchase service (see Table 50): 87 This calculation is based on prior rese arch contained in Katz, R. L., Vaterlaus, S., Zenhäusern, P., Suter, S. (2010a). The impact of broadband on jobs and the German economy. Intereconomics , January-February, Volume 45, Number 1 , 26-34. The results were originally presented at the Confeder ation of German Industries in Berlin on June 17, 2009. 88 Total households: 39.7 million, Fede ral Statistical Office, Germany, 2006. 89 e a given population. It is different from penetration, which Coverage is defined as the ability of the technology to serv means actual adoption. 62

75 Impact of broadband on the economy Table 50 – Households passed by broadband Technology Subscribers Coverage Connected/ passed in % (million) (million) 20.9 57 DSL 36.7 1.6 7 Cable modem 22.0 0.092 13 Fixed wireless, satellite 0.730 FTTB 0.240 0.043 18 39.0 22.6 58 Total (assuming overbuilds) Sources: DT; Kabel Deutschland; Unitymedia; M-net; Bundesnetzagentur. 90 The Federal Government of Germany has agreed : on the following two broadband strategy targets 1. Nationwide capable broadband access (1 Mbps) no later than the end of 2010. 2. Provide 75 per cent of German households wi th access to a broadband connection of at least 50Mbps by 2014, specifically with the goal that such access lines should be available as soon as possible throughout the country. The deployment actions required to meet these targets are three-fold. First, the 730,000 unserved households (white spots) will be covered by a mix of wireless and wireline technology. It is assumed, following Deutsche Telekom's recent announcement th at 250,000 unserved households will be covered by DSL technology, while the remainder will be covered by wireless technologies. The second action will be to upgrade the 2.8 million “grey spot ” households to 1 Mbps connection. The third target of the National Broadband Strategy is that 75 per cent of households will have to be served by at least 50 Mbps by 201 4 with higher bandwidths to follow. This was structured in two stages: • oyed in dense cities limited to 50 Mbps, it is Upgrade to FTTH: Given that VDSL technology depl assumed that 9.92 million households (representin g 25 per cent of the German households) will be upgraded to FTTH. Since the current number of households served by VDSL is 10.9 million they are located in the major 50 German cities, it is assumed that the majority of them will be migrating from VDSL to FTTH • Upgrade to VDSL: Here it is assumed that the remaining 50 per cent of households will be be upgraded is calculated by subtracting from upgraded from DSL to VDSL. The number of lines to the target, the coverage that is already provided: 9,920,000 10,900,000 29,791,000 9,920,000 Lines to be households households households households upgraded to - = - - ) ( (moved to (2014 coverage (current VDSL (FTTH covered) VDSL FTTH) coverage) target) dband lines that will be connected as a result of this According to this calculation, it is estimated the broa effort are approximately 18.9 million. following investments to be completed by 2014: To sum up, the broadband strategy will require the • Unserved households (730,000) covered by a mix of wireless (480,000 lines) and wireline technologies (250,000 lines). • “Grey zone” households (2.8 million) upgraded to 1 Mbps. these will come from the 240,000 where FTTH • 9,930,500 households (or 25%) upgraded to FTTH: upgrading the rest from the existing VDSL has already been deployed (municipalities) and by lines. 90 Federal Government of Germany (2009) “Breitbandstrategie der Bundesregierung”, 3. 63

76 Impact of broadband on the economy 18.9 million households (or 50%) upgraded to VDSL: this will be reached by the remaining existing • VDSL lines (980,000) and by upgrading 17.9 million currently DSL lines. foresee the ), other government reports (BMWi, 2009c, Longer-term aspirations, as mentioned in 38 completion of a national ultra-broadband infrastructure by 2020. While this aim has not been underlined Strategy, one can assume a set of “aspirational” by clear policy objectives in the National Broadband targets for 2020: Deploy FTTH to 50 per cent of households. • • Deploy VDSL to the next 30 per cent of households. • Offer broadband services under 50 Mbps to the remaining population (20%). The action implied from these targets is to upgrade an additional 25 per cent of households to FTTH, (which when added to the 25 per cent upgraded by 2014 would reach a total 50 per cent by 2020).The calculation of total investment required has been co nducted for each provision by relying on costs per line. The combined wireline and wireless costs required to cover the unserved households will total EUR 924 million, which are broken down in Table 51: Table 51 – Investment required to cover unserved households Number of households Cost per line Technology Total Investment (EUR) (EUR millions) DSL 250,000 1,200 300 Wireless 480,000 1,300 624 Total 730,000 924 Source: Katz et al. (2010a). The calculation of VDSL and FTTH deployment relies on cost per line gathered by a number of sources (see Table 52): Table 52 – Cost per line estimates for VDSL and FTTH upgrade Source Cases VDSL (EUR) FTTH (EUR) Analysis (2006) 300-400 1,000 Analysis (2008) Netherlands 1,566 ADL (2009) Generic 300-450 1,150-1,700 AT Kearney (2008) Greece 1,206-1,525 WIK (2008) 475 JP Morgan (2006) 1,000-1,500 Iliad (France) Net Cologne (Germany) 1,000 Fastweb (Italy) 1,200 Hillegon (NL) 1,200 Source: Katz et al. (2010a). The cost calculation relies on the ADL figures with in creasing cost per household as deployment of the technology network increases. For example, in the ca se of FTTH the initial 10 per cent of households (3,972,000) will cost EUR 1,150 per household to deploy , the next 10 per cent will require EUR 1,287, and the next 10 per cent, EUR 1,425. In the case of VDSL the first 10 per cent will cost EUR 300 per line, while deploying beyond 50 per cent will require EUR 450. Based on these figures and the number of lines to be to meet the FTTH target is EUR 12,236 million, and deployed, calculated in 5.2, the investment required 64

77 Impact of broadband on the economy EUR 6,747 million. To sum up, the total investment the investment required to meet the VDSL target is required to fulfil the 2014 National Broadband Strategy will be EUR 20,243 million (see Table 53). Table 53 – Total investment required to achieve objectives for 2014 Target Amount (EUR Millions) Address the unserved “white spots” 924 Upgrade the “grey spots” 336 12,236 Deploy FTTH to 25% of households Deploy VDSL to 50% of households 6,747 Total 20,243 Source: Katz et al. (2010a). As a result, the incremental investment required to meet the FTTH target of 50 per cent households 91 served by 2020 will be EUR 15,690 million. 6.3 The National Broadband Plan in Brazil Broadband infrastructure in Brazil is still underdeveloped. In 2008, there were 11,489,000 fixed broadband lines in the country, which represent 7.1 per cent penetration per person (or 18.88% per g from states like the Federal District with household). Broadband deployment is uneven, rangin 52 per cent of households, to Roraima with 0. 30 per cent of households (see Figure 17). Figure 17: Brazil: Broadband household penetration 60% 50% 40% 30% 20% 10% 0% í l l l s s e o e o á a á s a o a a s o o a e s á e o u i i u t r i l r u t a rá r n a c ã g s á n b a n p p i r i r c a h n i n í i u a i m S a i t S s n i u h a a o ra r e e o a ô a o a a g r r e A n P b o P n r g r n a o d o d e ro a N ma t ra B S a P G z a C a a e a d d e m l v n Ge o r c a P A a G J o o o P S o F a o e t A A o s R C d i ã m o n e s d r T R a o t r Ma í S a t e s A d n i n a t e p r d o a n t s o r P M i n Mi s a i E a G Gr R S D o o i Gr t a R o i M R Source: IBGE Pesquisa Nacional de Amostra de Domicilios 2008; IPEA; analysis by the author. 91 2014 (EUR 12,236 million) and the second 25 per cent tranche The difference between the first 25 per cent achieved in 1) the first tranche benefits from the 240,000 households achieved in 2020 (EUR 15,690 million) is due to two factors: already served by municipal networks roll-out and, more importantly, 2) the cost per line in the second phase rises from EUR 1,150-1,425 to EUR 1,500-1,700. 65

78 Impact of broadband on the economy In addition, download speeds exhibit significant shortfalls (see Table 54). Table 54 – Brazil: Breakdown of download speeds (2009) Speed Per cent of lines 128 kbps-255 kbps 9.5% 256 kbps – 511 kbps 21.2% 512 kbps – 0.99 kbps 26.6% 24.3% 1 Mbps – 1.99 Mbps > 2 Mbps 18.5% Source: Cisco/IDC (2010). While no public information is available regarding broadband coverage (e.g. households being passed by broadband infrastructure), it is assumed that the demand gap amounts to 20 per cent, which means that in addition to the population that currently subscr ibes to broadband, and additional 20 per cent of households can purchase service but do not due to ei ther economic, educational or device accessibility reasons. Under this assumption, and considering the 40 per cent growth in broadband penetration that has taken place since 2008, the last year that household penetration was collected, the author estimates ely 49 per cent of households (see Table 49). that broadband coverage amounts to approximat Brazil's national broadband plan does not define coverage targets. For the purposes of this estimate, it is stipulated that the country will achieve universal access of 1 Mbps. With this in mind, it is necessary to 1 Mbps and how many lines need to be deployed to estimate how many existing lines to be upgraded to achieve universal coverage. Assuming that current coverage is 49 per cent of ho useholds and that 57 per cent of total lines are under 1 Mbps, it is estimated that 16,2 60,000 lines would need to be upgraded. In addition, assuming that supply gap would amount to 26,469,000 households. To coverage reaches 49 per cent of households, the will require upgrading 16,260,000 lines and deploying sum up, the fulfilment of the stipulated target 29,470,000 additional lines. The calculation of total investment required has b een conducted for each programme (upgrades and new deployment) by relying on costs per line. It is esti mated that the cost per upgraded line is USD300, while 92 As such, the combined wireline and the cost to deploy a new broadband line would average USD450 wireless costs required to cover the unserved households are depicted in Table 55: 92 The deployment cost is an approximate estimate that av erages the deployment of ADSL in relatively dense areas (In the US the estimate (USD300) and the construction of broadband wireless infr astructure in rural areas. approximates USD 1,800/line). The USD 450 per line value was reported by Giusti, M. (2010). Ier Foro Iberoamericano para impulse de la banda ancha. June 21, Sao Paulo, Brazil. 66

79 Impact of broadband on the economy Table 55 – Broadband penetration and estimated coverage (2009) UF Total Federative Penetrati Broadband % of Broadband % of Total Households Unit Households Households on + Households broadband Broadband (2009) households (2008) (2008) demand (2009) Households (2009) gap (2008) 28,600 40,057 187,175 186,748 21.40% 41.40% 12 Acre 15.31% 884,165 11.17% 7.99% 886,184 98,947 Alagoas 31.17% 27 70,646 1,044 0.63% 165,474 1,462 0.88% 20.88% 16 Amapá 165,097 Amazonas 797,134 13 8.01% 798,955 89,395 11.19% 31.19% 63,826 29 Bahia 595,190 14.06% 4,242,106 833,622 19.65% 39.65% 4,232,440 34.57% Ceará 10.43% 2,383,294 347,234 14.57% 247,918 2,377,863 23 755,130 386,779 51.22% 756,855 541,722 71.58% 91.58% 53 Distrito Federal Espírito 38.61% 1,064,608 294,160 27.63% 1,067,039 412,000 32 58.61% Santo 52 Goiás 342,817 18.43% 1,864,486 480,149 25.75% 45.75% 1,860,237 31.13% Maranhão 129,109 21 1,625,014 180,830 11.13% 1,621,311 7.96% 51 Mato Grosso 928,017 213,169 22.97% 930,136 298,564 32.10% 52.10% 45.34% 50 Mato Grosso 25.34% 734,761 133,255 18.14% 736,439 186,637 do Sul 31 Minas 6,126,839 1,210,850 19.76% 6,140, 832 1,695,915 27.62% 47.62% Gerais 15 1,941,274 150,302 7.74% 1,945,708 210,513 10.82% 30.82% Pará 1,086,860 114,679 10.55% 25 Paraíba 160,619 14.74% 34.74% 1,089,342 41 Paraná 3,392,013 937,449 27.64% 3,399,760 1,312,990 38.62% 58.62% 26 Pernambuco 2,481,854 256,223 10.32% 2,487,522 358,866 14.43% 34.43% 9.02% 22 56,531 6.45% 877,940 79,177 875,939 29.02% Piauí 33 Rio de 5,258,154 1,317,226 25.05% 5,270, 163 1,844,905 35.01% 55.01% Janeiro 24 Rio Grande 35.64% 888,764 99,484 11.19% 890,794 139,337 15.64% do Norte 43 Rio Grande 3,662,250 798,518 21.80% 3,670,614 1,118,403 30.47% 50.47% do Sul 11 42.51% Rondônia 452,541 72,886 16.11% 453,575 102,084 22.51% 14 Roraima 117,126 0.30% 117,394 486 0.41% 20.41% 347 40.89% Santa 1,960,334 42 29.26% 1,964,811 803,326 60.89% 573,559 Catarina 35 São Paulo 12,916,358 3,800,540 29.42% 12,945,858 5,323,031 41.12% 61.12% 28 Sergipe 567,747 62,859 11.07% 569,044 88,040 15.47% 35.47% 17 Tocantins 41,980 11.08% 379,827 58,797 15.48% 35.48% 378,961 Total 57,714,525 11,999,946 20.79% 57,846,339 16,807,106 29.05% 49.05% Sources: IBGE Pesquisa Nacional de Amostra de Domicilios 2008; IPEA; Ci sco/IDC (2010); analysis by the author. 67

80 Impact of broadband on the economy Table 56 – Investment required to fulfil targets Programme Number of Cost per line Total Investment USD households (USD) Upgrade 16,290,000 USD 300 4,887,000,000 13,261,500,000 New deployment 29,470,000 USD 450 Total 18,148,500,000 Source: Author. As a result, in order to achieve universal coverage and upgrade of under 1 Mbps lines to 1 Mbps, Brazil will require an investment of USD 18 billion. 6.4 Conclusion This section presented a methodology on how to estimate investment requirements to meet deployment targets as stipulated in national broadband programme s. By applying the methodology to the German and Brazilian cases, it was possible to gauge the significant funding requirements of these plans. Assuming that a large portion of that investment will come from the private sector, it is pertinent to explore the policy frameworks to be relied upon to stimulate the funding of future deployment. 7 The Role of Public Po licy and Regulation in Boosting the Development of Broadband We turn now to a consideration of the policy tools necessary to promote broadband deployment and adoption in order to realize broadband’s potential to contribute to economic growth and the creation of jobs. The policy tools required to stimulate the deployment of broadband range from the formulation of ion policies and the identification of cases where national broadband plans to the enactment of competit the government should intervene in order to addre ss specific market failures. In addition to supply- oriented policies, additional measures need to be enacted to promote adoption of broadband by certain social groups and firms that may not be naturally inclined to adopt the technology. This section is based on a review of best practices in countries that have succeeded in reaching a high level of performance in the development of their broadband sector. In particular, best practices in the broadband policy domain will be reviewed for certain European countries (Germa ny, Sweden, Netherlands), Asia (Republic of Korea, Japan, Singapore), and Latin America (Chile and Brazil). 7.1 National broadband planning as a tool In recent years, several countries in the develope d and developing worlds have formulated national broadband plans; these plans outline both coverage a nd service targets, as well as policies, with the purpose of achieving near or complete universal broadband service. National broadband plans touch upon four broad policy areas: • The assignment of government assets necessary to reach universal broadband service coverage : the primary focus in this policy area is spectrum allocation. As indicated in the previous section, wireless broadband is the primary platform for reaching unserved and underserved geographic regions in developed countries (also called “white” a nd “grey” zones). It is also widely recognized that wireless broadband will be the primary technology used to provide ample broadband coverage in developing countries. By definition, wireless requires wide spectrum bands to be able to provide broadband access at adequate quality levels, which is primarily an issue of download speeds. In this sense, national broadband plans tend to focus, in many cases, on policies aimed at ting the “freed” spectrum resulting from the reassigning frequency bands (in particular, alloca 68

81 Impact of broadband on the economy digitalization of broadcasting) or searching for “white spaces” (unutilized bands) and assigning them to the wireless mobile communications sector. primarily oriented at addressing demand Investment in promotion of adoption programmes: • policies; the stimulation of the adoption of gaps, these programmes focus on: universal service broadband through digital literacy; economic subsidies; deployment of public access centres; and the development of eGovernment applications in order to promote adoption of broadband. Adoption of a competition policy: based on the premise that competition among service • suppliers is the right model to stimulate broa dband supply, national broadband plans, either implicitly or explicitly, tend to define the most appropriate way to develop market competition in the supply of this service. This involves an e ndorsement of either facilities-based competition (also referred to as platform-based competition) between vertically-integrated players such as the telecommunications incumbent and the cabl e operator or service-based competition (through unbundling of the telecommunication network of the incumbent operator and the sharing of incumbent facilities). If competition between fixed broadband suppliers does not exist or is not feasible, policies could outline performance rules (e.g., rules related to coverage and quality of service) that must be followed by the monopoly provider. Removal of any potential supply obstacles : related to the aforementioned belief that that • competition among service suppliers is the righ t model to stimulate broadband supply, national broadband plans focus on how to lower economic barriers to entry. Relevant policy initiatives ich can range from stipulating rules for duct, could include infrastructure sharing policies, wh mast, and tower sharing to lowering pole attachme nt costs (in aerial networks) to joint trenching rules. This policy area could even lead to potent ial investment by government to deploy national backbones aimed at lowering backhaul costs for broadband wireless players. In addition to formulating policies in the four areas mentioned above, broadband plans tend to stipulate targets to be achieved in terms of deployment, ad option, and quality of service. It should be noted however, that there still is not complete agreemen t among countries about what the appropriate goals for broadband should be (see Table 57). Table 57 – Coverage and speed targets of selected national broadband plans Speed Targets Country Coverage Targets (as per cent of households) for per cent Households United States 100% (2012) • 4 Mbit/s (100%) (2012) • 50 Mbit/s Germany 100% (2014) • 1 Mbit/s (100%) (2014) • 50 Mbit/s (75%) (2014) Singapore 100% (2012) • 100 Mbit/s (95%) (2012) Australia 100% (2012) • 12 Mbit/s (100%) (2012) • United Kingdom 100% (2012) 2 Mbit/s (100%) (2012) Malaysia (33%) 50-100 Mbit/s • 75% (2010) • (42%) 1.5 Mbit/s 50% of urban and 25% of rural households • 75% (512-784 kbps) Brazil European Union 100% (2013) • 30 Mbit/s (100%) (2020) • 100 Mbit/s (50%) (2020) Sources: Author with reference to nationa l broadband plans of specified countries. As Table 56 indicates, the primary area of consensus around national plans is the need to achieve universal (or near-full) penetration of broadband service, which implies the recognition of broadband as a this context, developing countries that are at a public good requiring full adoption by the population. In 69

82 Impact of broadband on the economy countries tend to define coverage targets that, lower rate of broadband penetration than developed while ambitious, shy away from full coverage. In terms of quality of service (as defined by download speeds), there does not seem to be full agreement across countries. There is no shared conception among the plans about what the minimum acceptable download speed should be or what an appropriate “ultra-broadband” target is. Part of the reason why there is not full such targets is, in many cases, based on political consensus on such targets is because the stipulation of itative analysis of costs and social and economic imperatives rather than substantiated on careful quant returns. For example, national broadband plans rarely discuss the expected economic payoff of achieving universal broadband service or the relative benefits of deploying high speed access in certain regions of a given country. The need to ground the setting of target s on careful technological and economic analysis is particularly important since, as shown in the prev ious chapter, targets have a logical impact on investment requirements. licy tools, national broadband plans represent an Despite the lack of agreement on targets and po initiative that can have a high impact at multiple levels: • They create awareness, both within civil society and government entities, about the economic and social importance of broadband service. • The plans represent a way of building consensus and promoting coordination between all areas of government that may have an impact on deployment and utilization of broadband. • National broadband plans have the potential to become state-level policies that transcend the political electoral cycles. • They help build accountability for plan fulfilment at the highest levels of government, particularly the executive branch. d in sections 7.1.1, 7.1.2, 7.1.3, and 7.1.4. Each of these four areas will be further discusse 7.1.1 Creating awareness at the highest national level y tool for articulating a vision of why universal The national broadband plan represents a valuable polic broadband service represents a critical societal challenge from an economic growth perspective. From a practical standpoint, the development of such a vision comprises four sequential steps. First, governments need to publicly reaffirm the collective imperative of deploying broadband as a means of supporting social and economic development goals. Th is requires defining a consensus around objectives and values that link technology adoption to economic and social development. Once this vision is policy makers and civil society around the criticality of developed, it is critical to build consensus between broadband usage. This should be part of a public de bate among all parties that have an impact on deployment and assimilation of broadband technology . Using this shared vision as a foundation, the targets should be defined based on rigorous analysis of investment and social and economic returns, as well as policy tools. The targets provide the context for the development of specific projects and ching target development goal, proactive, multi-year programmes. Using the integrated vision as an overar government planning represents the next step. Some countries embody what can be considered best practices in the field of national ICT (and broadband) planning. In the Republic of Korea, for example, starting in 1995, the government began preparing and implementing five year plans with objectives ranging from broadband universalization to becoming a global IT leader. A significant featur e of the Korea Government sponsored ICT planning 93 . ICT Master Plans are conceive d as tools for facilitating the process remains its holistic character transition into an advanced information society. This implies that planning vectors include not only the deployment of broadband infrastructure but also address services, applications and demand promotion 93 Kim et al., 2010. 70

83 Impact of broadband on the economy t of departure from the broadband development plans in policies. This last point represents a critical poin other developed countries. Planning efforts in other nations tend to have a heavy focus on broadband deployment and, while recognizing the positive spill-ov ers that networks will have on other sectors, they leave promotional efforts in these related componen ts of the broadband eco-system to market forces; this approach could be labelled as “build it and they will come”. Contrary to this philosophy, Korean policy makers tend to use their planning tools (influenced by industrial policy considerations) to address all the ed fashion, generating incentives for broadband components of the eco-system in an inter-connect adoption in the areas of applications and services to follow through the build-up of broadband networks. Additionally, with support of a government rese arch institute, the Korean Information Society Development Institute (KISDI), policy makers in this country were able to develop and refine a broadband technology strategy based on rigorous economic analysis. A country with a sector-wide planning tradition simila r to the Republic of Korea is Japan. The initial push for strategic planning in the broadband area started in 2001, when the government developed the first e - Japan Strategy. The strategic planning process enab led the formulation of annual priority policy programmes focused on implementing of objectives su ch as the universalization of broadband. Similarly, in Sweden, the Information Society for All bill led to the development of the Broadband Support Programme (2001-2007), which focused funding on deploying broadband in rural and isolated areas and building a national backbone. In 2007, the telecomm unications regulator issued a broadband strategy with the objective of achieving universal service, a nd finally, in November 2009, the government released its national broadband strategy. an information society occurred in 1998, when the In Estonia, the first integrated effort to create parliament adopted the Principles of Estonian Information Policy . This principles and objectives set out in Policy 2004-2006, which was this bill were further refined by the Principles of Estonian Information adopted in 2004, and the Estonian Information Society Strategy 2013 , which was approved in 2006. The Estonian Information Society Strategy 2013 establis hed the objective that by 2013, 75 per cent of Estonian residents will be able to access the Internet, while household broadband penetration will amount to 70 per cent. of Korea, indicate best practices in national As these examples, particularly that of the Republic broadband plan formulation, include the articulation of a vision, the integration of broadband goals with the fulfilment of objectives in other areas of the ICT eco-system (applications and services), and a follow- up and continuity built around the on-going formulation of plans and programmes. 7.1.2 Coordinating policies from different private and government entities As indicated above, national broadband plans need to be complemented with detailed blueprints for their implementation. These roadmaps are helpful in generating the appropriate frameworks for introducing changes in the regulatory arena. In particular, the adoption of a clear blueprint guiding broadband development is critical in terms of defining the resp ective roles of the public and private sectors and the potential construction of public-private partnerships. As mentioned in the introduction of this study, the private sector is expected to assume primary re sponsibility for investing in the development of broadband. Along these lines, it is important to determine what the right policy mechanisms are for stimulating investment from the private sector. Additionally, it is important to determine what the role of the public sector will be in addressing potential market failures in achieving universal broadband deployment. Most broadband plans address the models of private and public investments according to a segmentation of geographic areas. The model, popularized by the regu latory authority in the United Kingdom, Ofcom, in 2006, differentiates between “black” areas (where platform-based competition and good broadband service is expected), “grey” areas (where at least one service provider is expected to offer service although quality might not be consistent) and “white” ar eas (where service is not available). Based on this categorization, broadband national strategies, as adopted in countries such as the United Kingdom, Germany, Spain and Brazil, outline principles of pr ivate competition in “black” and some “grey” areas or outright government investment will address the while stipulating that public-private partnerships and/ 71

84 Impact of broadband on the economy plans have articulated the principle of geographic market failure in “white” zones. Numerous broadband Aid guidelines to the national broadband plans in segmentation, ranging from the European Union State Brazil, Germany and the United States. The role of the government in broadband deployment will be addressed in detail in section 7.1.3. In addition to defining the respective public and private sector roles, national broadband plans can articulate the respective roles of specific government entities that may contribute either on the demand or supply side to meet broadband penetration targets. As such, national broadband plans serve to outline the government responsibilities in areas as diverse as science and technology (for deployment of broadband in support of research programmes), education (for promoting digital literacy), health (to foster adoption of eHealth programmes), and ge neral administration (to promote eGovernment applications). For example, the national broadband plan of the United States spells out that more and better targeted funding for research and development (R&D) can have an impact on broadband supply. As such, the plan considers making R&D tax credits pe rmanent to stimulate development of ultra-broadband technologies. At the same time, it proposes coordina ting the creation of development test-beds with the National Science Foundation and the Department of Defence. 7.1.3 Developing state policies that go beyond electoral cycles In addition to plan formulation and programme development, a related best practice in the area of national broadband planning has to do with continui ty and discipline in follow-up. Broadband plans may be more effective when not subject to political imperatives or the need to address an infrastructure- based counter-cyclical policy at times of economic crisis. If endorsed by policy-makers as a primary component of the vision of the country's future, national broadband plans should become a permanent and on-going fixture of economic development. Several countries have reached this level of performa nce in broadband planning. In Chile, for example, the government is undertaking its third iteration of an updated digital agenda; each version of the digital agenda comprises an evaluation of results of past measures and the formulation of new targets and policy tools. In the Republic of Korea, each plan is assesse d in terms of its results at the end of the planning horizon and the results of the assessment are fed back in to the formulation of the next iteration. In that sense, ICT planning in the Republic of Korea has beco me the embodiment of state policies that capture a strategic vision, which in itself represents a consensus of all societal forces in the country. Similarly in countries with a different political system like Ch ina, institutional centralization of broadband policy making was reinforced with government sponsored planning. Senior leadership performance reviews are tied in tangible ways to achieving detailed annual planning targets that specify network capacity expansion, coverage, penetration, and quality standards. 7.1.4 Building ownership and accoun tability of the executive branch Complementing multi-year planning and disciplined follow-up, leadership at the highest levels of government in the promotion and oversight of broadband policy appears to be another best practice. This places responsibility for steering the development of the broadband sector squarely in the hands of the President. In doing so, the executive branch can act as a coordinator among government entities and foster a single national objective to be followed. Broadband deployment and penetration is the result of the combined intervention of many government entities, which requires a higher coordinating figure to make sure all departments proceed according to the same guidelines. In addition, by placing the fulfilment of national broadband objectives within the realm of the executive branch, the government conveys to the public and private sectors the importance that broadband holds for the future of the country. For example, in Brazil, the national broadband plan was developed by the Secretariat of Strategic Affairs of the President of the Republic and directly approved by the President of the country. In China, strong leadership from the top has been a key feature in China’s broadband plan development. The Ministry of Information Technology reports to the country's State Council and is a member of the State Information Leading Group (SILG). The SILG approves and modifies the regulatory framework and future Colombia, the recently announced digital plan was directions for the telecommunications industry. In presented to the nation by the President, in conjunction with all members of his cabinet. 72

85 Impact of broadband on the economy extend their sector intervention beyond multi-year In some cases, governments at their highest levels ructure. In the Republic of Korea, the government planning by actively shaping the broadband industry st 94 intervened in the market “in a focused and strategic way” at several points in the development of the sector, shaping industry structure either with the purpose of creating national champions, fostering export-led industries, or addressing sector sustaina bility. The Korean government often negotiated with 95 . Similarly, the the giant conglomerates (called chaebols ) over their participation in the broadband sector government fostered the consolidation of alternative broadband service providers at times of financial crisis. This required the modification of the original Telecommunications Law . The Japanese experience represents an example of a fine-tuned combination of to p-down sector planning with the creation of a set of incentives to stimulate facilit ies-based competition. While not explicit, the Japan Government has constantly adapted the regulatory framework to facilitate the moderate consolidation of the broadband sector in order to build a competitively sustainable regime. 7.2 Competition policies to stimul ate infrastructure investment 96 for funding broadband network Recognizing that the private sector has primary responsibility deployment, policy mechanisms need to define what the appropriate incentives are to promote private sector investment. In that regard, this section focu ses on articulating the appropriateness of platform- based competitive models. It proceeds first by review ing the main tenets of platform-based competitive models and then provides examples of countries th at have implemented such regulatory frameworks. While endorsing this model, it is agreed that so me situations might prompt governments to define alternative competition policies, around service-based models. The development of facilities-based competition, also labelled inter-platform or intermodal competition, is one of the major overarching objectives of the de regulation of the telecommunications industry. This ntegrated operators that manage their own network model is based on competition between vertically-i infrastructures and have sufficient stand-alone capacity for investment and innovation. The classic example of inter-platform competition is that of the cable TV operator that supplies services such as audiovisual content distribution, broadband access, and telephone services in direct competition 97 . The benefits of this with the telecommunications operator, which supplies the same range of services itive dynamics (prices, services, and user service model include the possibility of multidimensional compet se its level of investment (and, consequently, quality), while stimulating each operator to increa innovation) in its own network. The arguments against this modal suggest that insofar as it typically involves only a few firms, inter-platform competition, does not generate sufficient static efficiencies. In other words, it can lead to tacit collusion between pl ayers, with the resulting sub-optimization of prices for end consumers. Inter-platform competition stands in contrast to serv ice-based competition, which is defined as the model where industry players without infrastructure deliver se rvices to the market by leasing capacity from an incumbent network operator at a regulated wholesale price. By gaining access to the dominant operator's 94 Kim et al., 2010. 95 For example, in the last tranche of privatization of Korea Telecom, the government agreed with SK Telecom the acquisition of 11.3per cent of shares, while LG acquired 2.3 per cent. 96 One should not forget stimulus plans, universal service funds and public-private partnerships being also initiatives to channel government funding. This is particular ly the case in many developing countries. 97 In this context, it is important to discuss whether a broadband platform-based competitive model can develop in countries that lack a highly developed ca ble TV operator. In this case, wireless carriers (or WiMAX -based new entrants) offering broadband access could provide the necessary comp etitive stimulation for wireline carriers to invest in deploying broadband networks. 73

86 Impact of broadband on the economy infrastructure at a regulated wholesale price or through sharing agreements, new entrants can enter the market and set themselves up as viable competitors. Once this occurs, the new entrant will start investing in its own infrastructure when it reaches a certain crit ical mass of subscribers. Accordingly, infrastructure acquisition from the wholesaler becomes the first step in the 'investment ladder'. Thus, at least conceptually, service-based competition is a tempor ary stage in the transition to inter-platform competition. In contrast to platform-based competition, which tends to meet the objectives of dynamic efficiency (such as product innovation), service-based competition is not as effective in fostering infrastructure investment. Owing to the strategic behaviour of vertically-integrated operators, wholesale access negatively affects these operators’ rate of investme nt and their rhythm of product innovation tends to 98 decline. This is what economic theorists refer to as the 'inverted U' behaviour , which means that, in determining asymmetrical regulation over an incumbent operator (e.g. the obligation of wholesale access), there is an optimal level of competition that encourages innovation and investment. Beyond that point, the level of innovation tends to decline becaus e it does not hold strategic value for the integrated operator to share with its competitors any innovation capable of generating competitive advantage. The implications of this point are fundamental. First, if the regulatory obligations of providing wholesale access reach beyond an optimal level, they can have negative social and macroeconomic consequences in terms of limited broadband infrastructure investment and product innovation. Second, if the new entrants do not “climb up” the investment ladder, th e competition model must be reconsidered. This is why the determination of the appropriate competitive model is critical for the broadband industry's future development. An examination of the industry stru cture (not only the existence of cable, but also of an independent wireless player) is an important step in this determination. a temporary stage of service-based competition is Two caveats need to be made at this point. First, certainly better than no competition at all. Servic e-based competition models could work as temporary boosters of broadband penetration, as the European experience indicates. Second, since in general inter- platform competition is the best model to promot e economically-sustainable broadband sectors, some countries with no cable presence might have to cons ider a mixed model where service-level features are combined with a wireline versus wireless platform. These two competition models assume different policy approaches when it comes to broadband promotion. Infrastructure-based competition is base d on the competition between vertically-integrated operators with access to non-replicable passive infrastructure (e.g. ducts, poles, and in-house wiring). This model recognizes, however, that broadband economics do not allow for full competition in all geographies and therefore defines principles for stat e aid and public backhaul in underserved areas. On the other hand, broadband service-based policies fo ster competition among horizontally-integrated operators which have access to wholesale resources (e .g. bitstream, radio access spectrum sharing). It assumes the existence of a single network shared ac ross players competing on the basis of service and pricing. Because of the enormous economies of scal e resident in broadband access, a service-based policy might lead to the emergence of a single transport player that can achieve national coverage. In other words, whether the chosen broadband competition mode l is infrastructure-based or service-based, the economics of the technology will result in a single op erator (either supported by the state or a nationally deployed transport player) serving the rural and isolated regions of a country. Infrastructure-based competition requires the existe nce of two or more nationally-deployed players. Ideally, such a model should comprise a national telecommunications carrier, a single (or regionally focused) cable TV operator, and a stand-alone “pure play” wireless carrier. Table 58 presents information from five countries around the world that have implemented such a model based on such an industry structure. 98 See Aghion et al., 2005. 74

87 Impact of broadband on the economy Table 58 – Selected countries: broadband market shares (*) (1Q2010) United States Netherlands Republic of Chile Canada Korea • Telco 1 (20.3%) Telco 1 (20.3%) Broadband • Telco 1 (44.3%) • Telco 1 (41.7%) • • Telco 1 (42.7%) • • Telco 2 (23.4%) Cable (16.1%) • Cable (36.6%) • • Cable (38.5%) Telco 2 (11.8%) Telco 2 (1.2%) • • Cable (35.5%) Telco 3 (15.6%) • Telco 2 (11.1%) • Telco 3 (6.4%) • Telco 1 (17.1%) • Cable (39.9%) • • Cable (51.7%) • • Cable (68.7%) Cable (85%) Content distribution • Telco 1 (14%) Telco 1 (7.3%) Cable (20.0%) • Telco 1 (4.49%) • • Telco 1 (17.0%) • Telco 3 (17.5%) • Telco 2 (3.25%) Telco 2 (4.9%) • • Telco 2 (1.7%) • •Telco 1: KPN Companies • Telco 1: ATT • Telco 1: • Telco 1: KT • Telco 1: Bell Telefonica Canada Telco 2: SK/ •Telco 2: • Telco 2: Verizon • Telco 2: Telus Hanaro Vodafone • • Telco 2: ENTEL • Telco 3: T-Mobile •Telco 3: T- • • • Telco 3: LG Telco 3: Cable: Rogers Telmex/Claro Mobile • Telco 4: Sprint • Cable: VTR •Cable: UPC, • Cable: Comcast, Ziggo Cablevision and TWC (*) Number in brackets depicts market share Sources: Author, based on national regulatory agencies and telecom and cable operator reports The industry structure in the countries depicted in Table 50 not only includes a facilities-based telecommunications operator and one (or more) cable operator, but also a second mobile/landline operator competing with the landline operators on telecommunications operator and at least one mobile an intermodal scale. Those who propose service-based competition argue that the countries mentioned above represent the Indeed, the argument is based on the fact that in exception to the rule of building viable competition. these countries, the position of cable TV is quite developed and has created a natural inroad for the creation of inter-platform competition. Yet, it is inte resting to note that while this is true, the regulatory authorities in these countries did not adopt the inter-platform competition model by building on cable r experimenting with service-competition models and TV’s position, but rather adopted such a model afte ructural determinism (based on the axiom “service- identifying its limitations. Accordingly, beyond any st based competition was never given any consideration because there is a strong cable industry”) or an orderly progression in the 'investment ladder' process, the countries that have adopted the platform- competition model have done so on the basis of experimentation and testing, which led them to recognize its advantages. All the regulatory authoritie s of these countries tried initially to implement a service-competition model. intentions included the entry of a large number of The industry's initial response to these regulatory virtual competitors and a reduction in prices but at th e same time, a deceleration of investment (as in the United States and Chile). At the same time, the industr y started a process of cons olidation giving rise to players who competed in every sector of the industry (primarily, telephony, broadband, mobile and 99 content distribution), demonstrating the actual viability of inter-platform competition . The consolidation of the cable and mobile industry in the Netherlands is one example of a phenomenon that can also be seen in the mobile industry in Chile, the telecommunications and cable industry in the United States and the broadband sector in Korea. 99 See Katz, 2008. 75

88 Impact of broadband on the economy In view of this consolidation, the national regulato r in those countries recognized that the process for creating strong competitors with good financial heal th and a capacity for maintaining a certain rate of innovation and investment had to do less with an 'i nvestment ladder' and more with the natural “creative destruction” associated with competition and return s to scale that characterize a capital-intensive industry such as telecommunications. At the presen t time, the platform-competition model has finally been adopted by the regulatory authorities in all of these countries. In short, the adoption of the inter- platform competition model arises not so much from structural determinism as from the combination of the regulator's pragmatism and industry consolidation processes, resulting in an industry model that is ws for achieving universal broadband policy targets. both viable and balanced and that ultimately allo The question that still needs to be asked at the present time is to what extent have these countries sacrificed static efficiencies in favour of consolid ation? In other words, are we witnessing a situation where the joint dominance of a few operators is leading to market failures in terms of broadband development? Again, the study of the above cases shows that this is not the case. Table 59 indicates that in the transition towards the inter-platform competit ion models, the end user interest in innovation and low prices has been preserved. ased competition countries (December 2009) Table 59 – Performance metrics of platform-b Metrics United States Netherlands Republic of Chile(*) Canada Korea Broadband Population 26.4% 37.1% 33.5% 10.4% 29.59% penetration Households 63.5% 77% 95.9% 31.5% 82% Relative position OECD:15 OECD: 1 OECD: 5 LATAM: 1 OECD: 11 USD 19.99 USD 22.49 Broadband USD 27.48 Minimal USD 20.83 USD 30.47 subscription pricing (USD PPP) Relative position OECD: 14 OECD: 16 OECD: 29 LATAM: 2 OECD: 22 Average 14.3 32.8 100.0 1.5 19.6 Advertised download download speed speed (Mbit/s) Relative position OECD: 23 OECD: 5 OECD: 4 LATAM: 1 OECD: 15 Fibre as a percentage of 4.9% 11.3% 48.8% 0% 0% broadband accesses is author considers it more pertinent to evaluate it (*) Note: While Chile is an OECD country since May 2010, th against its Latin American neighbours. Sources: Author, based on data from OECD and national regulatory authorities Table 59 illustrates that in countries in which the inter-platform competition model was adopted, there were no noticeable market failures with regard to the development of broadband. First, the four developed countries tend to have a high level of penetr ation. Canada, while being in 11th position within OECD countries, has a broadband penetration of 82 per cent of households. The United States, which holds the 15th position in terms of OECD broadba nd ranking position among OECD countries, has a penetration rate in excess of 80 per cent in 35 states comprising 85 per cent of the country's population. Similarly, Chile exhibits the highest broadband penetr ation among its Latin American peers. Second, with respect to pricing, the US, Canada and Netherlands ar e within the mid-range of OECD countries; while the Republic of Korea has high minimum broadband prices, these prices do not appear to have materially impacted the level of adoption relative to other OECD countries. Finally, with the exception of the United States and Canada, access speeds in the other countri es are among the highest of their peer groups. This anomaly can be explained by the peculiarities of the two North American countries. While as advertised download speed, the US lags other OECD countries, both telecommunications incumbents and 76

89 Impact of broadband on the economy while deployment of fibre and DOCSIS 3.0 is well cable operators offer services in excess of 20 Mbit/s, ahead than most service based competition countries. Canada's delay in introducing higher speed offerings and deploying fibre is explained by the specific failed privatization of Bell Canada which delayed 100 . any infrastructure upgrade plans of this incumbent The successful experience of infrastructure-based comp etition models needs to be considered as a model in the context of country specific circumstances. For example, it is very likely that in countries where the industry structure comprises a single full service-ba sed provider and a stand-alone wireless player, an infrastructure-based model could likely result in market failures (e.g. large broadband unserved areas). In this situation, regulators need to pragmatically recognize that wholesale obligations of access on the telecommunications operator could be the more approp riate approach to stimulate entry of new players and boost a competitive regime. Alternatively, if the pr esence of a cable TV operator is limited to certain regions of the country (as it is so often the case), a hybrid model that combines infrastructure-based competition in selected geographies with service- based competition in others might be the most appropriate approach. In fact, the latter approach, labelled “geographic segmentation” has been implemented in many developed countries around the world. 7.3 Role of government intervention in promoting broadband deployment Should the government actively intervene in the development of broadband? The role of government in promoting the deployment of broadband can be inferred from the paradigms governing the application of universal service policies to wireline communications. Nevertheless, the broadband challenge might require new types of government intervention. While agreeing that private sector investment is the primary funding of broadband development, one rkets are not sufficiently developed to offer sound should recognize that, in some cases, broadband ma financial investments for carriers. If one assumes th at the government needs to address this potential market failure, the question remains as to what is the best way for the state to intervene. It is initially assumed that private sector investment tends to gravitate to areas where demand and demographic density guarantee an appropriate rate of return (see Table 60). According to the geographic segmentation principle, determining where government involvement is necessary is the first policy decision. Communities in a specific country need to be divided into one of three groups: those that are, or can be, served by ma rket forces; those that can become self-sustaining if they are given assistance with initial investment; and those that cannot become self-sustaining and require on-going funding. There are two potential routes that governments can take in order to address this particular market failure. The first is to directly enter the unserved regions as a service provider. The second approach is to generate the necessary stimuli in order to render the market more attractive to private sector investment. In general, state-owned facilities are the less desirable option. They tend to be less innovative, lack checks and balances, require more regulation (especially to enforce open access), and may have unintended consequences for utility behaviour (e.g., pricing distortions, “erosion of public 101 , etc.). Some of these risks will be addressed below. good” syndrome 100 Prompted by competition from the cabl e TV operators, Bell Canada plans to ha ve VDSL2 cards available to 100% of its network by YE2010, while beginning in 2H10, all new neighbourhoods in Quebec City will get FTTH with 100 Mbps capability. 101 We define this as the gradual transformation of the mission and operating paradigm of a municipal broadband network from a public service not for-profit utility to that of a private commercial enterprise. 77

90 Impact of broadband on the economy Table 60 – Market structure and demand characteristics Market structure More than 3 One operator 2-3 operators No operator operators Dense urban areas with high business High and residential density Urban areas/towns with primarily Medium residential density Rural areas with sparse residential Low density Density and size of demand Rural areas with Very low very low density Source: Author. If the generation of appropriate incentives to ease the burden of private investment is the more suitable approach, how can those areas of stimuli be determ ined? A sustainable broadband business is predicated on the number of potential subscribers, the aver age revenue to be derived per user, the capital curring operating expenses. In this context, a investment required to enter the market, and its re challenges (or “choke points”), and two strategic broadband business case faces two structural market and/or operational constraints. To begin, the broadband market in a small community may be too small. If primary demand is not sufficient, even if a company is a monopoly and completely controls the market, network deployment may not be profitable. The demand challenge has to do with building critical mass and, consequently, leveraging the industry economie s of scale. While larger company size does not necessarily lead to lower costs, companies that have higher market share in the communities tend to have lower unit costs. Second, investment in equipment may be too high rela tive to operating profits. This leads to lengthy horizons for a positive return on the investment, or, in other words, puts a heavy burden on businesses in the short run. Strategically, businesses can also face challenges if a low market share in a structurally small market negatively affects revenue streams. T hus, because of competition, a large investment may not have a profitable return unless a business is able to capture a sufficient share of demand. The capital expenditure structural challenge differs according to the type of network: in fixed broadband networks, construction costs are the largest cost item. In wire less broadband networks, the primary cost category is 102 backhaul infrastructure. Government intervention can render a private sector business case sustainable by taking several initiatives that positively impact the investment model. First, governments can put in place mechanisms in order to reach the level of critical mass that makes entering the market a worthwhile venture for providers. It can do this directly by adding its own demand to the natural market, or indirectly by subsidizing subscribers to make prices more affordable . If demand is low because there is little interest or too few potential adopters, the former is probably the best the course of action. If the area is too poor to afford broadband at prices sustainable to providers, the latter strategy may be the best course of action. 102 Backhaul service, which includes the lines required to inte rconnect the base station to the network, also represent a high recurring expense. 78

91 Impact of broadband on the economy s in the targeted area by providing low-cost real Government can also help lower the capital expenditure ovide grants to fund capital investments or reducing the estate for central facilities. Alternatively, it can pr costs of obtaining rights of way and/or spectrum a ccess. Some of these policy tools are reviewed below. Bundled demand Generally, the best way to induce private investment in broadband infrastructure is to “bundle demand”. The government does not need to artificially interven e in the market; it can act as an anchor user to guarantee revenues during the ramp-up phase of br oadband installation. Local governments can pro- actively coordinate demand for broadband access from public administration, public safety, local schools, and health care facilities in order to create an “anchor tenant”. Once the demand “consortium” is structured, the government negotiates a wholesale ra te and long-term contract with a broadband service provider in order to create a flow of revenues th at eases the initial economic pressure and reduces investment risk. Additionally, the government can st imulate demand from the private sector by working at the grass-roots level. Centralized efforts such as the establishment of Broadband Expertise Centres and library access and demonstrations are also helpful. Such efforts are particularly useful to spreading broadband knowledge, conducting training, and develo ping interest among organizations that do not utilize broadband. This will be discussed in more detail in the sections below. The Netherlands has developed a number of good practices in this regard. Subscriber subsidies Subscriber subsidies should be used sparingly because of their distortion potential. However, in certain cases they can be very beneficial. As mentioned abov e, a subsidy targeted at economically-disadvantaged subscribers is one of the appropriate uses of this a pproach. Such a subsidy addresses the social inclusion problem that faces governments that are seeking to ensure universal service. Fiscal incentives are also a useful form of subsidy: a reduction in taxes to sm all and medium enterprises has been found to stimulate broadband adoption in industries that have a strong impact on economic output. In Sweden, for example, tax incentives are given to businesses and residential tax- payers who signed up for broadband services: 103 ble up to a maximum of SEK 5000 (or roughly USD740). 50 per cent of the subscription costs are deducti Infrastructure sharing In order to reduce backhaul costs, infrastructure sh aring (e.g., backbone and towers) should be allowed and encouraged. Infrastructure sharing alleviates cost pressures on competing providers. If multiple broadband providers are not sustainable, sharing or consolidation may produce a broadband access “utility”. It allows operators to capture economies of scale and reduce investor risk, which is tantamount to lowering costs. There are several other ways to reduce costs to network providers in a given area. National regulators may reduce right of way or access costs (e.g. spectrum costs or pole attachment fees). They may also attempt to regulate backhaul costs, although in general, states do not have the capacity to do so. A way to address this last issue is to provide grants for capi tal investment, particularly backhaul capital costs or recurring expenses. These grants could take several forms: a subsidy for purchasing backhaul services (e.g. T-1 lines) from an operator or direct underwri ting of government-owned backhaul facilities that could offer services at a lower-than-market pricing to remote operators. Government as a risk taker In the last resort, if private investment does not fl ow after suitable incentives are provided, governments can act as a risk taker without resorting to public ownership. One possibility is to subsidize the incumbent telecommunications carrier and upgrade broadband to the “utility” status. In greenfield situations, 103 e average taxpayer has a marginal tax rate 20 per cent. This represented a key stimulus in a country where th 79

92 Impact of broadband on the economy a universal access network. This may induce strong governments can contract for the construction of competition for government contracts and lower the initial costs of the operation. Afterward, the government can auction the right to operate the broadband infrastructure to highest qualified operator. This process gives the government the option of cr eating a monopoly for wholesale-only or an open access “utility” operator. In effect, any loss that is su stained upfront is a one-time infrastructure subsidy much like building a highway system. In most countries, local governments already play a ro le in broadband deployment. In the US, there is a legal framework allowing municipalities to operate a broadband operator in response to a failure of the private sector to deliver service. By 2009, there were 66 municipalities that are already operating fibre networks and over 40 more that are planning to do so. In Sweden, there were 136 municipalities with fibre-based networks in 2009. In Germany, there were 25 city networks in 2009, some of which controlled 50 per cent of the local market (e.g., Cologne and Hamburg). Finally, in the Netherlands, there are currently 16 municipal fibre projects covering most major cities (e.g., Amsterdam, Rotterdam, and at there is no single business model driving municipal Almere). However, it is important to mention here th broadband networks. Municipal networks can follow on e of the following schemes: 1) closed networks, where the municipality directly provides retail services; 2) the municipality is a wholesaler to a single retail service provider; 3) the municipality offers open access or wholesale transport to multiple retail service providers; and 4) the municipality offers dark fibre. It should be noted that public investment in broadband could come at the expense of three risks: • First, municipal networks have the potential to create access bottlenecks. While funding is typically provided to municipalities in order to deploy infrastructure in areas where infrastructure competition is not feasible, these broadband service providers can start behaving as commercial entities, either by deploying infrastructure in areas where competition was feasible or regularly refusing to provide dark fibre or access to their infrastructure to competitors. In Sweden, for example, where funding was provided to municipalities in order to deploy infrastructure in areas where infrastructure competition was not feasible, these municipal broadband service providers begun to behave as commercial entities. As the regulator noted, providers deployed infrastructure in areas where competition was feas ible or regularly refused to provide dark fibre 104 . or access to their infrastructure to competitors • Second, municipal networks may shift their mission from public to commercial service. In a display of non-competitive behaviour, municipa lities can post broadband investment in their wholly-owned electric utility’s balance sheet to benefit from lower borrowing costs. That has been the case with municipal networks in Germany and Switzerland. t violated, some municipal networks experience • Finally, even if the public service mission is no the form of operating inefficiencies (which difficulties serving their customers. This may take prevent them from showing a positive financial pr ofile), cumbersome customer provisioning in a multi-provider system, and/or difficulty in managing the network and resolving service problems. That has been the experience in the United States and Canada with some of the municipal fibre 105 . networks In a risk profile similar to the one linked to direct en try into local access networks, direct state investment in backbone networks may also be problematic. To be gin, the government entry in the provisioning of long-haul fibre optics can be jeopardized by the lack of coordination between the time of policy formulation and the launch of the company, result ing in lengthier delays. Furthermore, the public broadband provider may lack operating and business in dependence and the ability to assume institutional 104 See PTS (2008). Dark Fiber: market and state of competition. Stoc kholm, Report PTS-ER-2008:9, p. 23. 105 See Mitchell, Christopher (2008) “Municipal Broadba nd: Demystifying wireless and fiber optic options”, Broadband , February, p. 42, and Curri, Michael. The South-Du properties ndas “Municipally owned network: from fiber optic pioneer to cautionary tale”. Strategic Networks Group. 80

93 Impact of broadband on the economy responsibility. The terms and conditions of network access might not be clearly specified, and the regulatory authority could be overstretched. In sum, a lack of coordination and coherence may cause deregulation and liberalization of telecommunications to conflict with government re-entry into the 106 . broadband market International experience is helpful in outlining the ar eas of opportunity for and the risks associated with government intervention. The evidence above indicates that the only case where government viate the constraints of businesses and stimulate private intervention is sustainable is when it aims to alle investment. Government intervention that aims to pre-empt private investment is not likely to prove sustainable (see Table 61). Table 61 – Options for government intervention in broadband provisioning Is project sustainable and profitable? Yes No Alleviate the constraints of the business • Pre-emption (“crowding out”) of • Yes private investment (Germany, case to stimulate private investment Switzerland, Netherlands) • Re-creation of access bottlenecks (US) • Erosion of the public utility model (US, Sweden) intervening? Is government No • Market addresses the need of Supplier of last resort • public good Source: Author. tuations where a broadband enterprise is sustainable According to Table 55, government intervention in si and profitable could result in the “crowding out” of private investment; this has been the case with some municipal networks in European countries, such as Germany and Switzerland. On the other hand, if the broadband business is not profitable or sustainabl e, the more appropriate form of government or structural constraints of the business case. Finally, intervention is the alleviation of the “choke points” government intervention could be conceived as appropriate in cases where, even after incentives are put in place, the private sector would not invest. In sum, government should intervene in broadband and wireless deployment, but only by facilitating market forces, not by pre-empting them. This proces s requires strong cooperation between governments, communities, businesses, and operators to identify supply and demand conditions and tailor services to unmet needs. Specifically, in the areas where broadband is deployed, it is important to identify the barriers to consumer adoption; in the areas where th ere is no broadband service, it is important to identify the barriers to sustainable market entry. The next step would be to develop a plan that incentivizes private investment, for example, by promoting public-private partnerships, encouraging market competition, refining supportive regulati on, and modifying spectrum management policies. 7.4 Stimulating innovation in applications and services Countries that have succeeded in building a highly developed broadband sector have transitioned from developing policies on a sector-by-sector basis (tel ecommunications, software, science and technology, and computing) to an integrated and comprehensive mode of policy-making. An integrated approach to ICT policy development recognizes the interconnected domains of ICT (infrastructure, demand, 106 See the experience of Infacom in South Africa, as describe d in Gilliwald, Allyson (2009). Presentation to Alternatives for velopment, Brasilia, 16 November 2009. Broadband Infrastructure and Access De 81

94 Impact of broadband on the economy cations, broadcasting, and IT applications and production, and adoption) and sectors (telecommuni devices). This integrated policy approach has two primary dimensions. Outline overarching objectives for the ICT sector An integrated policy approach translates initially into the formulation of a vision of the future of ICT for anning effort. In the Republic of Korea, for example, the country; this vision then guides the multi-year pl each of the multiple plans formulated by the government has been guided by an overarching visionary objective such as “reach world class ICT performance levels by 2010” ( 1996-2000 First National Informatization Promotion Plan ), “build a knowledge-based society” ( Cyber Korea 21 ), “development of broadband leadership” ( Broadband IT Korea Vision 2007 ), and “broadband convergence and ubiquitous networks” ( ). u-Korea Master Plan In Japan, the government developed an overarching strategic policy in 2006 labelled u-Japan , which was guided by three targets: 1. Elimination of non-broadband served areas, establishing that by the end of 2010, broadband service should be available to 100 per cent of the population, while high speed broadband should be available to 90 per cent of the population. 2. By the same year (2010), 80 per cent of the popul ation should value ICT as a tool to address social needs; this should be measured by the level of adoption and assimilation of applications and services, particularly in the eGovernment domain. 3. Finally, in the same year (2010), 80 per cent of the population would be IC T literate in order to feel at ease accessing the Internet and computer technologies. mprise infrastructure, digital literacy and social Interestingly, the goals in Japan's 2006 strategy co objectives. Link broadband deployment to industrial policy objectives In addition to formulating an overarching vision for the ICT sector, moving from a sector-specific regulatory policy to a comprehensive industrial policy recognizes that the development of a telecommunication sector and the creation of export-o riented IT services and software industries have to be linked. In the Republic of Korea, policy makers determined that meeting demand domestically and leveraging the industrial power of big conglomerates could allow the country to build an export base in electronics, IT, and communications. Initially, however, objectives were articulated in terms of meeting internal demand for an upgraded telecommunication in frastructure and entering the electronics arena. According to this approach to ICT sector development, incubation of an export-oriented industry is linked to funding development and adoption of its products in the domestic market. A key policy objective of all Korea master plans has been the articulation of i ndustrial policies such as research and development promotion (R&D), the development and diffusion of i ndustry standards, training of ICT resources, the promotion of e-Government applications, and the provision of seed capital for infrastructure deployment. This development fund benefits from private sector contributions through spectrum licensing fees, a percentage of revenues from operators, and interest earning loans. As such, one of the fund’s primary objectives is to reinvest the profits of the ICT sector in the sector itself. Over time, the guiding principle for the formulation of policies evolved toward “buildin g the information society”. Based on the overarching goal of developing an advanced information society, Korea formulated several successive master plans, which featured both supply and demand-side polic ies. Finally, Korea's policies regarding broadband development were always focused on the developmen t of an applications and services sector both benefiting and acting as a stimulus for infrastructure usage. As a result, the development of broadband acted as a stimulus for the creation of a content industry. Among the newly-created industries, Korea 82

95 Impact of broadband on the economy 3.4 billion domestic content industry, as a well as a counts a USD 8.3 billion online gaming industry, a USD 107 home-grown Internet search sector. With a similar objective of promoting the development of an equipment manufacturing industry, the ICT International Competitiveness Ministry of Information and Communication in Japan set up the in 2007 aimed at promoting Japanese products and developing world markets Enhancement Program through a collaboration of industry, academia, a nd government. This programme has been actively endorsed by the ICT manufacturing sector. In addi tion, the development of ICT strategies has been constantly supported by large domestic high technolo gy companies such as Canon, Mitsubishi, Nintendo, Panasonic, Sony, and Toshiba. Similarly, several Latin American countries, such as Panama and Brazil, link broadband development goals with objectives to expand their presence in international software markets and/or creating a local call centre industry. 7.5 Stimulating broadband demand Thus far, the debate surrounding the digital divide in the use of Internet and broadband has been primarily focused on the statistics regarding computer ownership and broadband penetration. The major issue in the eyes of public policy and public opinion has been the need to increase the adoption of the service through the expansion of the technological coverage. The underlying assumption is that by reducing the obstacles for infrastructure investment, the digital divide challenge would disappear. Yet, while without a doubt supply-side issues such as the gap in investment contribute heavily to the digital divide, demand for broadband services also plays a key role in explaining service penetration. The following section identifies the causes of the demand gap and outlines a set of policies aimed at tackling these issues. 7.5.1. Inhibitors of broadband adoption: the broadband demand gap gap is defined as the number of households that For the purposes of this study, the digital demand choose not to purchase a subscription to broadband serv ices even though they are served by a network. lculate because broadband coverage (in other words, the Historically, this statistic has not been easy to ca access broadband services) is usually not measured by amount of households that have the capability to public or regulatory agencies. However, the last year has seen the development of numerous national 108 which have necessitated a thorough analysis of the broadband coverage gap. For broadband strategies 92 per cent of all US households were capable of example, in the United States, according to the FCC, broadband access via cable modem and 82 per cent co uld purchase service via DSL. In the National 109 , the number of unserved or underserved housing units amounted to Broadband Plan released in 2010 7,000,000. However, penetration statistics indicate th at only 62 per cent of US households subscribed to the service. Thus, 30 per cent of households that had the capability of acquiring broadband access chose not to subscribe. This gap in broadband service demand was also identified in Germany. According to the National Broadband Strategy, published in 2009, 98 per cent of all households (39.7 million) in Germany are capable of purchasing broadband access. Coverage is divided as follows: 36.7 million households have access to DSL platforms; 22 million are served by cable television (and therefore could access the Internet through cable broadband); and 0.73 million are capabl e of subscribing to wireless broadband through satellite or other fixed wireless services. Table 56 estimates the demand gap for a number of countries. 107 See Kim et al. (2010). 108 See Katz et al, 2009. 109 www.broadband.gov/download-plan/ 83

96 Impact of broadband on the economy Table 62 – Broadband demand gap Country Households Demand Gap Households passed(*) connected Australia 89% 69% 20% 96% Denmark 20% 76% France 100% 77% 23% Germany 98% 58% 40% Israel 100% 83% 17% Italy 55% 40% 95% Republic of Korea 100% 93% 7% Spain 93% 61% 32% Sweden 100% 11% 89% United Kingdom 100% 68% 32% United States 92% 62% 31% (*) Note: Household passed is defined as a resi dence where the broadband network is deployed; e residence is linked to the network for this differs from connected, which means th provisioning the service. Sources: Analysis by the author, based on data from EU; FCC; BMWi; OECD; PTS – Sweden; and Israel Minister of Communication. As Table 62 indicates, the broadband demand gap in the developed world ranges from 7 per cent in the Republic of Korea to 40 per cent in Germany and Italy. While statistics for developing countries are not available, the broadband demand gap is expected to be higher. In Argentina, for example, the two telecommunications carriers could provide broadband to 91 per cent of households, while cable TV operators could offer service to 82 per cent. In this country, broadband household penetration is 29 per 110 cent, which means that the demand gap is 62 per cent It is evident that in all countries, a significant portion of the population does not subscribe to broadband Internet for reasons beyond service availability, such as affordability, lack of di gital literacy, or limited interest. An understanding of this problem and its causes is critical in order to put in place an appropriate set of policy tools for promoting broadband adoption. 111 it is argued that individuals respond to the introduction In a study on ICT adoption inhibitors in Spain, and spread of ICT either by adopting a technology after being exposed or by not adopting it. Within those 112 We believe that these two that adopt the technology, another s ubgroup exists – the “sporadic users”. groups, the exposed non-users and the “sporadic user s”, are at the heart of the demand-side digital divide. Though the percentage of people who do not use the Internet and who consequently do not subscribe to broadband service varies by country, the figure is generally significant. The study on Spain referenced above estimates that in July 2007, non-users (outsiders and sporadic users) accounted for 52.9 per cent of the adult population. Research conducted by Pew Internet & American Life indicates that in June 2009, non-users accounted for 21 per cent of the US population. In the UK, the figure was 26 per cent in March 110 Investment prospectus of Mult icanal and Telecom Argentina (2010). 111 See Red.es, 2007. 112 These individuals make use of the technology, but only on a periodic basis (such as bi weekly, monthly or quarterly). 84

97 Impact of broadband on the economy 2009. However, this last figure reflects the state of adoption within the population; in Spain, the equivalent is 40.2 per cent. ng the Internet? Studies in the United States and What are the reasons cited by non-users for not adopti the United Kingdom place the reasons into four categories (see Table 63). reasons for not accessing to the Internet (2009) Table 63 – United States and the United Kingdom: Percentage of answers Reasons United States United Kingdom Relevant (lack of interest, busy doing other tasks, other reasons) 45% 60% Price (the cost of broadband is too high, does not have a computer) 15% 28% Service availability 14% 16% Easy to use (difficulty – senior citizen – person with disability) 22% 16% Source: Horrigan, J. (2009); Ofcom (2008). As shown above, both the order of importance and the percentage of responses in each category are fairly consistent: limited relevance is the most import ant factor, followed by service price, availability of service, and comfort or knowledge required for use of the service. The United States and the United Kingdom cases illustrate the reasons for non-adoption in developed economies. In developing countries, however, availability and affordability of services are likely to be more important than, for instance, the lack of interest. While there are no studies to this author's knowledge, absence of relevant content could also be a major constraint. cussed above in section 2.7.3), it is important to With the exception of the availability of service (dis understand the reasons that inhibit broadband adoption. Is there a socio-demographic profile that can help explain the limited relevance (or lack of interest), the cost barrier, and the difficulty in usage? Studies from the United States, the United Kingdom, and Spain have found that the reasons for not adopting the Internet or broadband are remarkably consistent. Inhi bitors to broadband adoption are clustered around: 1) level of education (completion not above secondary school); 2) age above 65 (related, in some cases to absence of children in the household); 3) location in rural areas; 4) disadvantaged socio-demographic ducation and employment status (unskilled workers, groups, which is also correlated with level of e retirees and homemakers); and 5) income (less than net EUR 900 per month). The effects of socio-economic status on the process of technological adoption have already been studied 113 . However, education merits further analysis because it can be influenced by public policy. elsewhere According to OECD statistics, at 93 per cent, the Republic of Korea is the country with the highest level of broadband penetration. While telecom regulatory fact ors (referred to above) explain the absence of a supply-side digital divide in the Republic of Korea, th e education level in this country is a key explanatory variable that accounts for the minimal demand-side di vide. The Korean population is comparatively more educated than the Japan or United States population. The average number of years of education completed in the Republic of Korea is 15, one more th an the Japan average. Moreover, this figure is 50 114 The spill-over effects of education on broadband per cent higher than the average for US households. adoption are a factor that should influence public policy-making both in education and in ICT. In that 113 See, for example, Rogers, Everett. And Shoemaker, F. (1971). Communication of Innovati on. New York: Free Press and, Stoneman, P. (1976) “Technological Diffusion and the Computer Revolution, the UK experien ce”, department of Applied Economics Monograph 25, Cambridge University. 114 See Kalba (2006). 85

98 Impact of broadband on the economy netration is not only a result of technology-based regard, it is important to emphasize that broadband pe policy, but should also be addressed in educational policy objectives. Finally, the affordability issue, which is so important in developing countries, must be emphasized. This ic crisis, which indicates how significant the effect has been also highlighted in the context of the econom 115 of income could be on the decision to adopt broadband at the household level. 7.5.2 Relevant policies aimed at addressing the broadband demand gap In addition to deploying policies to stimulate infrastructure development aimed at achieving wide service coverage of key technologies, leading information societies implement several demand-side policies aimed at promoting broadband adoption. Introduction of tax incentives Governments of countries with high performing ICT se ctors tend to introduce tax incentives designed to encourage the purchase of equipment. In addition to the Swedish example described above, in Japan, firms investing in ICT solely for their own use have th e option of either taking a 10 per cent credit from corporate tax or a special depreciation equivalent to 50 per cent of the acquisition cost. Developing e-government services By actively developing e-government services, go vernments can generate additional incentives for consumers and small businesses to join the information society. Such e-government services could include, for example, the electronic submission of tax returns, an e-procurement service for small and medium enterprises selling goods and services to the government, platforms for tele-commuting, and the development of portals that allow the interaction be tween the government and enterprises for e-business transactions. This initiative is generally complemented by the implementation of digital literacy programmes that include subsidies for acquiring PCs and online educatio n programmes targeted at the elderly and disabled, such as the programmes implemented in the Republic of Korea. In the case of small businesses, the Japan Government encourages small and medium enterprises to voluntarily install new IT platforms to reform business management and improve productivity by providing training, collecting and disseminating best practices, and supporting collabo ration with local communities. Enabling environment for SMEs In the small and medium enterprise (SME) sector of developing countries, the implementation of business processes and applications enabled by broadband is st ill limited. Productivity and global competitiveness of the SME sector are affected negatively by limited broadband adoption. Conversely, broadband adoption by large enterprises is high due to afford ability, abundance of training, and spill-over factors. 116 , the primary reasons for the low level of broadband uptake As discussed by the author in prior studies among SMEs are: limited access to investment capital; comparatively high technology costs; and lack of training. Regarding capital investment and monthly servic e costs, it is important to note that a significant proportion of SMEs, particularly in the developing world, do not receive fixed monthly income because they operate outside of the formal economy. Their income is generally daily or weekly and is dependent upon the type of labour performed; thus they canno t borrow long-term or purchase products that require a fixed monthly payment such as PCs, servers or Internet access. These enterprises are generally forced to use prepaid wireless, Internet booths or cybercafes, and rented PCs. 115 For example, the Pew Research study indicates that during 2009, 9 per cent of Internet users in the United States have canceled or downgraded their Internet service due to economic pressures. 116 See Katz, 2009b. 86

99 Impact of broadband on the economy In addition, many of the entrepreneurs that run SMEs (which are primarily microenterprises) have a very limited level of technological training. In developing countries, a large number of SME owners are subject to a generational gap because they were not exposed to Internet technology as they grew up. Therefore, they lack the necessary training to operate a computer or to use broadband to improve business efficiency. This lack of training translates into a fear of using technology and ignorance of its capability to create economic value. Both the public and private sector must address this obstacle. The private sector must redefine product development processes so that the services market ed toward SMEs are not simply “impoverished” versions of those offered to larger companies. Produc ts developed for the SME sector must be tailored to its needs for processing and transmitting information, its economic capacity for acquisition and operation, and the level of training it will require to operate the technology. The specific needs of the sector will only be met if these dimensions are properly understood . On the other hand, governments must actively contribute to the extension of technology training a nd education. Continuing education courses that focus on SME owners and employees can vastly improve productivity. ICT adoption by SMEs is also limited by cultural-educ ational factors. In emerging countries, SMEs tend to restrict the use of ICT to accounting and finance, while neglecting its application to production processes. A survey by the Chilean Ministry of Economy found that only 2.6 per cent of Chilean companies used ICT to increase the efficiency of business processes other than accounting and finance. Yet the most worrisome observation in the survey was the following : 80 per cent of companies reported that they did not implement ICT in areas other than finance and accounting because they lacked of the technological expertise necessary to understand how it would be useful. Hence, in order to successfully stimulate SME adoption of broadband, it is necessary to take steps beyond offering incentives to the sector and creating the right conditions. SMEs must also be informed a nd educated about the strategic capabilities of ICT. ss to and retention of skilled ICT workers. Because of Another obstacle facing SMEs is the difficulty in acce a systemic shortage of technical personnel, large companies offer wages to graduates of higher education that SMEs cannot match. Even where SMEs manage to hire graduates, retention rates are very low. geographical asymmetry. As a result of uneven Lastly, ICT adoption by SMEs is obstructed by telecommunications network deployment, SMEs that operate in urban centres tend to have better access to broadband infrastructure and technological capital, whereas those that operate outside of such areas are marginalized. nd networks, but also by a lack of incentives. For ICT adoption is not only marred by undeveloped broadba example, many governments of emerging countries have failed to enact laws that promote ICT adoption unications, and so forth. The lack of incentives such as tax incentives, subsidies for remote telecomm impacts SMEs in two ways: it affects both supply and demand. On the demand side, high adoption costs restrict the adoption of platforms th at allow SMEs to enhance efficiency. On the supply side, small retail providers of ICTs (m ainly in products and services) cannot compete with suppliers of goods or services that control a la rge share of the market. Consequently, small firms supplying ICT products and services tend to mimic the behaviour of market leaders. This behaviour restricts the adoption of ICT among SMEs because it increases the costs of promotion for suppliers. The second problem on the supply side concerns the prov ision of telecommunications services. Competition must be stimulated in the telecommunications services industry in order to reduce access prices. The agenda promoting ICT adoption by SMEs should ha ve the following focus: economic issues (such as cost reduction and incentivization); education; and th e development of products specific to SME needs. As the economic factor represents a fundamental obstacle for ICT promotion, governments must prioritize the search of solutions in this field. In th is vein, the adoption of tax benefit programmes, special financing, and subsidies are recommended, following models of universal telephony service that have been adapted for the SME segment. Second, since training is such an important factor in ICT adoption, investment in continuing education programmes is recommended. These programmes can teach SMEs how to take advantage of new market products worldwide. Another measure to be technologies, for example, web sites can be used to 87

100 Impact of broadband on the economy ations and social networks (such as Facebook, but considered is the development of participatory applic inesses to share experiences and form alliances to targeted to SMEs). This medium would allow small bus improve market access. This mechanism should promote the best practices of companies or government administrations. tion and use of ICT plays a crucial role in creating The support of consultancy services regarding the installa awareness about the potential of these technologies for SMEs. Finally, in the new products segment, the development of ICT packages for SMEs that include computers, with maintenance support and usable software, voice communication and broadband services is recommended. These packages would be financed in monthly payments, including the fee for the use of the services. Promoting deployment of basic infrastructure While the deployment of basic broadband infrastructu re is necessary for ICT adoption, it is equally necessary to provide stimuli and incentives for the purchase and installation of products and services for the residential and corporate markets. In order to ac hieve these goals, a vast number of measures that will contribute to ICT adoption can be implemented. First, in the context of promoting the adoption of wireless handsets, this author has established the importance of lowering taxes on purchases of hardwa re. (See the discussion below.) This tariff reduction on equipment should be extended to fiscal programmes that tax the usage of telecommunication services. Such tariffs and taxes negatively affect th e rate of ICT adoption. Therefore, the income these taxes and tariffs generate for national treasuries sh ould be evaluated in the context of the negative impact that they have on the adoption of ICT and thus on enhancements to productivity. In the same vein, to speed the rates of acquisition and modernization of equipment, the use of accelerated depreciation accounting schedules should be considered. Finally , the establishment of discounts or rewards for enterprises that use ICT for their transactions with th e government could become an additional incentive. This will have a positive impact on the use of e-gove rnment and on the use of ICTs by SMEs. In order to equipment, SMEs should be provided with financial stimulate the use of credit for the acquisition of ICT forecast tools, certified by financia l institutions, which will help enterprises in the process of requesting a loan from a bank. Another important point is the increase in the number of telecentres. For example, even countries such as Brazil, which boasts 5,000 centres, should increase their number. Regarding specific recommendations, the extension of opening hours of telecentres in order to serve schools, communities, and businesses throughout the day and evening should be considered. What can we expect if a region is not successfu l in promoting ICT adoption by SMEs? Given the importance of SMEs in the economies of developing co untries, a policy failure in this domain could have a significant negative impact. The capacity of the SME sector to enter international production networks and to export to other markets will be greatly reduc ed. Thus, overall economic growth will also be reduced. This task is extremely urgent: a failure to promote ICT adoption will jeopardize competitiveness in international markets. As studies by the World Economic Forum have indicated, there is a strong correlation between the development of ICT infrastructure (as measured by the Network Readiness Index) of a country and its level of competitiveness. ate the adoption of ICT by SMEs. The government It is the responsibility of governments to stimul apparatus is capable of generating spillover effects similar to those generated by the Republic of Korea chaebols. Chile Compras is a good example. This is a portal designed to promote SMEs’ sales of primary inputs to the government. To participate in the eco- system of Chile Compras, the SMEs must adopt Internet platforms. Thus, the programme benefits SMEs, as well as the government. Companies must adopt technology, which leads to improvements in productivity and sales, and the state benefits from access a greater number of suppliers of inputs. The resolution of the digital divide entails solving coverage problems in neglected economic corridors and meeting the needs of SMEs. These challenges can be met directly by municipal and provincial governments. These levels of government are capabl e of more efficient allocation of resources than national governments, and tend to be directly responsible for the accumulation of intangible capital, (e.g., barriers to technology adoption by SMEs. education), which is one of the foremost 88

101 Impact of broadband on the economy It is important to analyse the experience of Asian countries where SMEs represent the centre of gravity of the economy. In the Republic of Korea, SMEs account for 99.8 per cent of enterprises and are responsible 96 per cent of enterprises, and in India, 90 per cent for 87 per cent of jobs. In Malaysia, they account for of enterprises are SMEs and 86 per cent of jo bs in the formal economy are found in SMEs. been implemented in order to increase ICT adoption In Asia, three important sets of public policies have by SMEs. First, improve SME awareness of the critical role that ICT plays in improving performance. The primary focus of this policy, which is promoted by chambers of commerce and provincial government agencies, is training SME entrepreneurs. Second, provide training that not only focuses on ICT, but also focuses on changes in structure and processes that will help SMEs absorb the value of ICT, i.e. what we called the accumulation of intangible capital in section 2.1. Thes e training projects must be tailored so they address the application of ICT specifically to the SME sector. Third, create an environment that incentivizes ICT adoption by SMEs. This refers to the provision of ta x incentives and financial tools that allow SMEs to access technology. Some incentives directly stimulat e ICT adoption by SMEs (through subsidies, etc.) However, an ICT friendly environment can also be achi eved through indirect incentives. For example, tax deductions can be offered to large enterprises that, in the course of their purchasing operations, help small companies acquire ICT. Other indirect incentiv es include subsidies for broadband installation in industrial parks. This practice is extremely common an d successful in countries such as India and Malaysia. 7.6 Addressing taxation as a barrier to broadband adoption it comes to broadband penetration (see Table 64). The developing world lags significantly when Table 64 – Comparative broadband penetration (per population) (2010) Population Penetration Continent/Country Western Europe 19.7% North America 27.7% Asia 4.9% Latin America 6.5% 1.6% Africa and Middle East Sources: ITU; Euromonitor; World Bank; analysis by the author. Cognizant of this wide disparity, many government s in the developing world are implementing public policies aimed at stimulating broadband deployment and adoption. For example, in Malaysia, the 117 . Wireless government objective is to reach a household broadband penetration rate of 50 per cent broadband is the technology of choice to achieve this target. For this purpose, the government has issued new spectrum licences to four companies that will roll-out new wireless broadband services based on WiMAX platforms. Furthermore, to rationalize capital investment, th e government has imposed sharing requirements for towers among high-speed downlink packet access (HSDPA ) and WiMAX operators. Finally, as an incentive for operators to roll-out their broadband networks, th e government also approved tax allowances on expenditures on last-mile broadband equipment. In general terms, most developing countries' policy makers now envision mobile broadband as a key lever to address the digital inclusion gap. With the exception of countries where 3G licences have not yet been 117 See Malaysia Communications and Multimedia Commission. The National Broadband Plan, 2006. 89

102 Impact of broadband on the economy se in wireless broadband services combined with the auctioned, all countries register a continuous increa deployment of 3G-enabled handsets and devices (see Figures 18 and 19). Figure 18: Mobile data as a percen tage of service revenues (2003-10) 35% 30% 25% 20% 15% 10% 5% 0% 10 08 08 07 06 06 05 05 04 03 Q Q 2Q 4Q09 2Q09 4Q 2Q 4Q 2Q07 4 2Q 4Q 2Q 4Q04 2 4Q China S. Africa Brazil Malaysia Mexico Source: Adapted from Katz et al. (2010c) from Merrill Lynch (2010). Figure 19: 3G Phone subscribers as a percen tage of all mobile subscribers (2007-2010) 25% 20% 15% 10% 5% 0% 7 10 2009 200 20 2008 Mexico Brazil South Africa Malaysia Source: Adapted from Katz et al. (2010c) In this context, high taxation on mobile devices and services could have a detrimental effect on the public policy strategy aimed at deploying broadband. With the few exceptions of countries like Malaysia, which has implemented a benign taxation system based on extremely low value-added tax, many developing countries have introduced taxes that could nega tively affect service diffusion (see Table 65). 90

103 Impact of broadband on the economy Table 65 – Mobile taxation approaches in selected economies Handset Services Country Fixed Other Other Fixed Taxes VAT Customs Duty Taxes VAT Taxes Taxes (USD) (USD) 21% 0-20% - – - - – - Argentina 18% 0.04-0.10 18% Burkina Faso 1% - – - 13.30% Bangladesh 35% 11.76 15% 20% - – - 11.63 15% 18% 16% - – - 18% Brazil 9.30% 13.35 3.70% Ghana 12.50% 2.50% 12.50 9.50% 5.50 6% 4.33 60% Iran - – - Malaysia 5% - – - - – - 10% - – - - – - Mexico 16% - – - 16 0.10% - – - - – - 3% 14% 7.60% 14% South Africa - – - - – - Sri Lanka 33% 15% Tunisia 18% 5% 10% 8% Venezuela 14% 1.56-6.25 14% 14% Source: Adapted from Katz et al. (2010c). the total cost of ownership of mobile service varies The impact of these different taxation approaches on widely. For example, in Mexico, the impact of taxes on total cost of ownership is 18.4 per cent; in South Africa, it is 15.2 per cent; in Brazil, it reaches 29.8 pe r cent, while in Bangladesh it is 54.8 per cent. On the ile cost of ownership amounts to only 6.1 per cent. other hand, in Malaysia, the effect of taxes on mob Taxation of mobile services appears to have an impact on the deployment of mobile broadband. For example, all things being equal, there may be some association between the very high level of taxes in Brazil and its very low penetration level of 3G handset s. On the other hand, Malaysia has a low level of taxes and a high 3G penetration rate. Similarly, an inverse relationship appears to exist between tax burden and adoption of data services when measured by wireless data as per cent of service revenues (see Figure 20). If taxes limit adoption of wireless broadband, it is re levant to ask what the ultimate impact of reduced penetration might have on economic growth. According to a study by this author, the wealth creation generated by the lowering of taxes was higher th an the accumulated loss in tax collection given the positive spill overeffects of broadband diffusion (Katz et al. , 2010). To conclude, it is safe to assume that a reduction in adoption as a result of incremental taxation could yield a negative impact on GDP growth. 91

104 Impact of broadband on the economy Figure 20. Taxation vs. adoption of data services 60% 50% 40% Malaysia Mexico 30% Brazil revenues 20% 10% South Africa Wireless data as % of service 0% 30 25 0 35 40 45 50 15 5 10 20 Overall Taxes Source: Adapted from Katz et al. (2010c). 8 Conclusion The objective of this study was first to provide ev idence on the economic impact of broadband, while effects is still in its infancy, primarily due to the recognizing that the discipline aimed at measuring these limitations in data, but also resulting from the still evolvi ng analytical tools. In spite of this situation, the research, and this study, has already generated a considerable amount of proof that broadband has y, both in terms of fostering growth and creating considerable positive spill-over effects on the econom ee areas. First, the evidence of both the existing employment. The study conclusions are structured in thr research literature and this study will be summarized. Second, based on the experience gathered in the course of this study, recommendations will be made in terms of data that needs to be collected by public fine the study of economic of broadband. Finally, a administrations going forward in order to further re number of policy implications will be highlighted. 8.1 The nature of the evidence of broadband economic impact The evidence generated for this study as well as the results of prior research validate the positive contribution of broadband to GDP growth both for developed and developing countries and regions. While limited in the number of countries studied , these analyses confirm that broadband has a directionally positive economic impact. While it is not optimal to compare model results across geographic units, the following conclusions can be drawn. First, the quantitative case studies of Germany, Latin America and the Caribbean, Chile, Malaysia and cating a positive effect on GDP growth. These results India yielded statistically significant coefficients, indi add to the evidence generated by the studies conducted in the developed world by other researchers. ia and India cases are based on a low number of However, due to data limitations, the Chile, Malays observations, which may jeopardize the consistency of estimators. The ability to generate a larger set of observations for these last studies would allow us to safely conclude on the positive effect that broadband has on economic growth of developing nations. Nevertheless, Koutroumpis (2009), Katz et al. ates models exhibit higher levels of reliability. (2010a), and the cross-sectional Latin America and Arab St Second, an aggregated interpretation of the evidence would appear to point out to the validation of a critical mass hypothesis: the higher penetration of broadband, the more important is its contribution to economic growth (see Figure 21). 92

105 Impact of broadband on the economy Figure 21: Relationship between broadband penetratio n and importance of contribution to GDP growth Source: Author’s analysis. As Figure 21 demonstrates, it would appear to be a direct relationship between broadband penetration and strength of economic effect. The results of the analyses also consistently valid ate the positive contribution of broadband to job particular, the Germany, Indonesia, Brazil and Chile creation in developed and developing countries. In ts. The other cases (e.g. India, and Saudi Arabia) have cases yield statistically significant positive coefficien also yielded statistically significant coefficients for the explanatory variable (broadband penetration) with sensible signs – positive when the independent va riable is employment and negative when it is unemployment. However, the reduced number of observ ations limits the robustness of estimates of the models. 8.2 ering to refine impact measurement The need to emphasize data gath The assessment of economic impact of broadband techno logy should be considered an emerging area of research, which is critical in providing evidence in defining policies. The broadband policy arena has been advancing very rapidly since the inception of the techno logy prompted by the shift of information flows to data communications from voice, and the consequent growing amount of investment from the private sector. Unfortunately, it would be fair to say that ec onomic impact studies have not developed at a step commensurate to that of the policy development domain. Yet, triggered by the growing availability of data series, new studies have been completed recently, shedding some light on under which conditions, and to what extent, can broadband contribute to econ omic growth and job creation. At the same time, the studies have allowed to identify the gaps in da ta availability. Governments, especially regulatory authorities, should emphasize the gathering of data in order to facilitate the analysis of economic impact, a key cornerstone of policy making. As a primary take-away of this study, a set of indica tors for measuring the economic impact of broadband is proposed and included in Appendix B. Data requirements range from the aggregate macro-indicators to of consumption patterns and user needs. the micro-data that provide an indication 93

106 Impact of broadband on the economy 8.2.1 Disaggregated data for ICT, broadband and economic indicators The foremost need is for data on all indicators to be disaggregated at the regional level within a country (county, canton, department, even postal code). The lower the level of aggregation is, the more data points there are and, with more data points, the effe cts of broadband can be estimated more precisely. In points are needed to obtain an adequate level of addition, the more variables that are used, the more precision. Since growth models rely on at least five or six variables, it is necessary to have a large number of observations (at least 30-40) to estimate the effects of broadband. The second reason for disaggregated data needs is that it allows research to deal with local fixed effects. Even with a rich set of controls, it is hard to measure the effects of broadband using large levels of aggregation. All sorts of questions ar ise as to how comparable the data are. For example, it is difficult to compare penetration rates or education (which are two variables used in growth models), across countries, or even between cities and rural areas. Quality may differ tremendously between countries, therefore two people with the number of years of education or both with broadband Internet may be getting very different services if one is in a developed and another is in a developing country. Therefore, the lower the level of aggregation is, the more accurate results will be. 8.2.2 Quarterly data Quarterly data is another way to increase the nu mber of observations and thus the accuracy of estimations of broadband effects. In many countries employment is reported on a quarterly basis. Therefore, if broadband subscriptions are also reported quarterly, it will be possible to estimate the relationship between broadband and employment more accurately. For one, the amount of time required for broadband to impact employment will be known with greater specificity. At the moment, research indicates that there is a lag period between network construction and the moment where indirect employment effects kick in. However, whether this lag is one quarter or two years long is currently unknown. Quarterly data is the only way to observe this phenomenon. More accurate estimations of the effect of broadband on employment are also needed because the relationship is complex and also subtler than that of broadband on GDP. While employment is clearly raised by network construction, the long-term effects on empl oyment are less clear. For example, broadband may enable outsourcing or online shopping could hur t employment at local commercial businesses. From an economic point of view, in the short-run, increases in labour productivity may lead businesses to substitute capital for labour. While there are also many positive effects of broadband on employment, the relationship is very subtle and more precise estimates are needed to quantify it; as of now the results of research vary, some papers find no relationship, (the negative and positive effects are balanced), while others find a positive effect. Quarterly data will allow for more precision, which will go a long way towards solving this problem. Finally, growth models necessitate data from at least two points in time per country, state or other unit of observation. Therefore, if data is collected quarterly, research may be conducted as soon as one year after regulators begin releasing new time series. Ho wever, with annual data, two years is the absolute minimum needed for researchers to use new data. 8.2.3 Range of broadband download speed Policy makers and researchers agree that the speed of Internet access matters. So far, research has proven that the move from dial-up to broadband (be it DSL or cable modem) has a positive impact on productivity. However, it is yet unclear whether there is a linear relationship between speeds and economic impact. The question is: at what speed sh ould broadband be offered in order to maximize economic impact? In order to answer this question, regulators need to keep data on the number of subscriptions by speed. For one, this would greatly facilitate cross-country co mparisons. It is unrealistic to assume that the s the same service quality as his counterpart in a average broadband customer in a developed country ha developing nation. Yet, because these data are not available economic studies are forced to make this 94

107 Impact of broadband on the economy assumption. This may have created a large range of problems. For example, studies disagree whether broadband is more useful to urban areas and developed countries. However, more developed countries often offer faster broadband services, so it is difficu lt to tell whether the effects increase because the impact of increasing access speeds is very larg e or developed countries can more successfully use broadband services. Data compiled on the number of subscriptions by range of speed will help researchers quantify the marginal returns on speed. In turn, this will help countries optimize their broadband plans. For example, it will become clear whether policy should focus on the development of new technologies (such as fibre to the home deployme nt) or increasing coverage of basic technologies. There is no point in investing in 100 Mbit/s services if they don’t offer any socio-economic gains relative to 10 Mbit/s services. 8.2.4 Data on wirele ss broadband Internet Wireless Internet represents the platform of choice for meeting the demand for broadband in developing ped countries in terms of countries. It is also increasingly important for develo the shift to wireless of a coming increasingly sophisticated and as a result great deal broadband usage. Wireless data plans are be their speeds and capabilities are comparable to fixed-line broadband plans. In light of the surge in wireless data plans, (which is expected to continue thro ugh the next few years), developed countries will experience a substantial increase in broadband subscriptions, even though the fixed line market is in some cases saturated. Without data on wireless data plans, socio-economic benefits will be wrongly accounted for. This problem is only exacerbated in developing countries, where, due to lack of infrastructure, the majority of Internet connections are through wireless networks. Fixed-line broadband and wireless data subscriptions simultaneously in creasing and therefore without data on mobile broadband subscriptions the economic benefits will be attributed solely to fixed line broadband. This will severely influence estimations on the importance of broadband, because mobile data plans are just as if not more important than fixed lines. 8.2.5 Data measuring: the demand gap There is much debate over the short-term and long-term effects of broadband networks. From an differentiate between direct effects of roll out and econometric point of view homes passed allows us to other effects. At the moment econometric models have a great amount of difficulty measuring direct effects because if subscribers rise during one year it is usually as a result of construction in the previous year. However, we tend to measure the effects of subscriber gains on the next year. Because homes uction, the confounding effect would be solved if passed are a more accurate measure of network constr ort-term and long-term effects would be especially this data were collected. Differentiating between sh useful for understanding the relationship between broadband and employment. As mentioned above, the long term effects of employment are the most controversial. However, when the both short and long term effects are measured at the same time, we risk overlooking the subtleties of the long-term effects. In addition, homes passed would allow researchers to gauge the demand gap, defined as the population that could subscribe to broadband but do not. This is critical since in many cases, broadband policies are primarily focused on supply stimulation (e.g. how to stimulate further investment in unserved and underserved regions) when the first and easiest problem to be tackled to increase penetration would be the demand side (e.g. what type of education prog rammes, digital literacy campaigns, and potential subsidies could be implemented to stimulated adoption). In that sense, it is critical to gather data in terms of coverage, quality and speed by region of a given country. In addition, survey data on household and enterprise broadband utilization should be extremely useful to determine policies tackling the demand gap. 8.2.6 Variables for income endogeneity There is good reason to believe that, like most other goods, broadband is income elastic. That is to say broadband not only drives income growth, but inco me also drives broadband demand. This two-way relationship is termed endogeneity and is very difficult for researchers to deal with. It may severely 95

108 Impact of broadband on the economy P. The most convincing econometric tool to deal impact estimations of the impact of broadband on GD with this is to rely on a system of equations. However, this process is not feasible for most countries because of data availability. The following variable s would greatly enhance researcher’s capabilities for using systems of equations. They allow the estimati on of broadband supply and demand, which solves the problem of income endogeneity. In the first place, both supply and demand are functi ons of price. Therefore, in order to make use of simultaneous equations, it is necessary to have a meas ure of price (such as average or median price paid), for each geographical unit of observation. Though it is far less desirable, if price is unavailable, competition can be used as a proxy. This can be me asured by the number of providers serving a certain area (however this is obviously only useful at small levels of aggregation such as the zip code or county). Competition is also useful because it can be used as a proxy for subscription rates. For example FCC only collects data on the number of firms that offer broadband in a given zip code, not the penetration level. rted by government. This share is so large that Secondly, a large amount of broadband ventures are suppo is problematic to broadband supply without government incentives. Government surplus is not a good enough proxy – in fact it is misleading. One would think that governments with greater surpluses would be able to offer more incentives, but research such as Koutrompis (2009) has observed the opposite effect in OECD countries. It would seem that large deficits are indicative of big spenders. However, it is very likely that this would not hold up if we analysed developing countries where broadband plans might be the first to be cut in a situation of financial need. Clearly, in order to adequately describe supply, researchers need direct data on government incentives and investment. 8.3 The policy kit for stimulating br oadband deployment and adoption This study also introduced a methodology aimed at calculating the investment required to achieve full broadband penetration in section 6.1. That methodol ogy was demonstrated in the cases of a developed ile the economic effects are substantial, the level and a developing country. The results indicate that, wh lined in national broadband plans is substantial. of investment required to achieve the targets out of social and economic “business cases” of such an Accordingly, the next step should be the development be the return of a national broadband plan and its investment. In other words, what is it going to associated price tag? Finally, the study focused on the policies that have pr oven to be most successful in stimulating not only investment in broadband but also adoption by population and businesses. The tool kit is wide ranging comprising not only national broadband plans but al so the definition of competition models, and the development of demand stimulation tools, which should also extend to fiscal policies. The purpose of this study has been to outline challenges and opportunities of what we believe to be a critical task for the years to come. Bibliography Aghion, P., Bloom, N., Blundell, R., Griffith, and Howitt, P. (2005). Competition and innovation: an Quarterly Journal of Economics inverted-U relationship. , vol. 120(2): pp. 701-728. Atkinson, R., Castro, D. & Ezell, S.J. (2009). The digital road to recovery: a stimulus plan to create jobs, boost productivity and revitalize America . The Information Technology and Innovation Foundation, Washington, DC. Atkinson, R. and Schultz, I. (2010). Broadband in America. New York , NY: Columbia Institute for Tele- Information. Atrostic, B.K. & Nguyen, S.V. (2006). How businesses use information te chnology: insights for measuring technology and productivity. U.S. Bureau of Census, Washington, DC. Battese, G. & Coelli T. (1995). A Model for Technical Inefficiency Effects in a Stochastic Frontier Production 20 (2), 325-332. Function for Panel Data. Empirical Economics , 96

109 Impact of broadband on the economy The Quarterly Journal of Economics, Barro, R. (1991). Economic Growth in a Cross Section of Countries. 106 (425), 407-443. Basu, S. & Fernald J. (2006). Information and communications technology as a general-purpose technology: evidence from U.S industry data (Working Paper Series 2006-29). Federal Reserve Bank of San http://www.frbsf.org/publications/e conomics/papers/2006/wp06-29bk.pdf Francisco. Retrieved from Bundesnetzagentur (2009). Jahresbericht 2008. Retrieved from www.bundesnetzagentur.de/media/archive/15901.pdf. BMWi (2009a). Breitbandstrategie der Bundesregierung, Stand (2009, February). Retrieved from www.bmwi.de/Dateien/BBA/PDF/breitbandstrategie-der- bundesregierung,property=pdf,bereich=bmwi,sprache=de,rwb=true.pdf . Der Breitbandatlas des BMWi BMWi (2009b). . Retrieved from www.zukunft-breitband.de/BBA/Navigation/breitbandatlas.html . BMWi (2009c). Konjunkturgerechte Wachstumspolit ik Jahreswirtschaftsbericht 2009 , www.bmwi.de/BMWi/Redaktion/PDF/Publik ationen/jahreswirtschaftsbericht- 2009,property=pdf,bereich=bmwi,sprache=de,rwb=true.pdf . Cisco/IDC (2009). Barometro de la banda ancha: Brasil . Clarke, G. (2008). Has the Internet Increased Expo rts for Firms from Low and Middle-Income Countries? Information Economics and Policy 20 Crandall, R., Jackson, C., & Singer, H. (2003). The Effect of Ubiquitous Broadband Adoption on Investment, . Washington DC: Criterion Economics. Jobs, and the U.S. Economy Crandall, R., Lehr, W., & Litan, R. (2007). The Effects of Broadband Deployment on Output and , 6 Employment: A Cross-sectional Analysis of U.S. Data. Issues in Economic Policy . Czernich, N., Falck, O., Kretschmer T., & Woessman, L. (2009, December). Broadband infrastructure and economic growth (CESifo Working Paper No. 2861). Retrieved from www.ifo.de/DocCIDL/cesifo1_wp2861.pdf The Internet Ecosystem: Employment Impacts of National Darby, L. F., Fuhr, J. P. Jr., & Pociask S. B. (2010). Washington, DC: American Consumer Institute for Citizen Research Broadband Policy. de Vries, G. & Koetter, M. (2008). How does ICT enhance productivity? Evidence from latent retail technologies in Chile. de Vries, G., Mulder, N., Dal Borgo, M. & Hofman, A. (2007) ICT Investment in Latin America: Does it Matter for Economic Growth? CEPAL. Dreher, A, Noel G., & Pim M. (2008). Measuring Globalization – Gauging its Consequences . (New York: Springer). The Impact of Broadband on Growth and Productivity . Fornefeld, M., Delaunay, G. & Elixmann, D. (2008). A study on behalf of the European Commission (DG Information Society and Media), MICUS. Garcia-Murillo, M. (2005). International broadband deployment: The impact of unbundling . Communications & Strategies, 57. 83-108. Galperin, H. and Jordan, V. (2010). Speeding up the digital revolution: broadband in Latin America and the Caribbean . Santiago: Chile, Economic Commission for Latin America Gillett, S., Lehr, W., and Osorio, C., & Sirbu, M. A. (2006). Measuring Broadband's Economic Impact. Technical Report 99-07-13829, National Technical Assi stance, Training, Research, and Evaluation Project. Gordon, R. (1998). Monetary Policy in the Age of Information Technology: Computers and the Solow Paradox. Paper prepared for the conference Moneta ry Policy in a World of Knowledge-Based Growth, rsion, June 3, 1998. Quality Change, and Uncertain Measurement. Bank of Japan, Corrected ve 97

110 Impact of broadband on the economy Greenstein, S. & McDevitt, R. (2009). The Broadband Bonus: Accounting for Broadband Internet's Impact www.nber.org/papers/w14758 on U.S. GDP . (NBER Working Paper 14758). Retrieved from Greenstein, S. & McDevitt, R. (2010). Evidence of a Modest Price Decline in US Broadband Services (NBER Working Papers 16166). Retrieved from www.nber.org/papers/w16166 . Islam, N. (1995). Growth Empirics: A Panel Data Approach. The Quarterly Journal of Economics, MIT Press, 110 (4), 1127-70. Horrigan, J. (2009). Home broadband adoption 2009 . Pew Internet and American Life Project. IPEA (2010). Analise e recomendacoes para as politicas publicas de massificacao de acceso a internet em . Brasilia, Brasil, Comunicados de IPEA. abril 26, 2010. banda larga Jipp, A. (1963), “Wealth of nations and telephone density”, (july), Telecommunications Journal pp. 199-201. Jonscher, C. (1982). Productivity change and the growth of information processing requirements in the economy. Doctoral dissertation, Harvard University, 1982. Jorgenson, D., Ho, M., Stiroh, K. (2006a). The industry origins of the American productivity resurgence . oh, K. (2006b). Productivity growth in the new October 2006.Jorgenson, D., Ho, M., Samuels, J., Stir d at the 2006 Intermediate Input-Output Meeting millennium and its industry origins. Paper presente July 26-28, Sendai, Japan. Jorgenson, D., Ho, M, Samuels, J., Stiroh, K. (2007). Productivity growth in the new millennium and its industry origins. Paper presented at Sloan Industry Studies Conference, Boston. Kalba, K. (2006). Korea, Japan and the U.S. Broadband Gap . Unpublished manuscript. The economic and societal benefits of ICT Kaplan, (2007). use: an assessment and policy roadmap for Latin America and the Caribbean . Arlington, VA: CompTIA. Katz, R.L. (2008a). La competencia entre plataformas: teoría y resultados . Editorial Enter, Madrid. Katz, R. L., Zenhäusern, P. & Suter, S. (2008b). An evaluation of socio-economic impact of a fiber network in Switzerland. Polynomics and Telecom Advisory Services, LLC. Katz, R., L. & Suter, S. (2009a). Estimating the economic impact of the broadband stimulus plan (Columbia Institute for Tele-Information Working Paper). Retrieved from www.elinoam.com/raulkatz/Dr_Raul_Katz_-_BB_Stimulus_Working_Paper.pdf Katz, R. L. (2009b). La Contribución de las tecnologías de la in formación y las comunicaciones al desarrollo económico: propuestas de América Latina a los retos económicos actuales . Madrid, España: Ariel. Katz, R. L. (2009c). The Economic and Social Impact of Telecommunications Output: A Theoretical Intereconomics Framework and Empirical Evidence for Spain, , 44 (1), 41-48. Katz, R. L. (2009d). Paper Estimating broadband demand and its economic impact in Latin America. submitted to the ACORN REDECOM Conference 2009, Mexico City, September 5, 2009. www.acorn-redecom.org/program.html Katz, R. L., Vaterlaus, S., Zenhäusern, P. & Suter, S. (2010a). The Impact of Broadband on Jobs and the German Economy. Intereconomics , 45 (1), 26-34. Katz, R. L. (2010b). La banda ancha: un objetivo irrenunciable para Brasil . Paper presented at 54o Panel Telebrasil. Guaruja, Agosto. Katz, R.L., Flores-Roux, E. & Mariscal, J. (2010c). The impact of taxation on the development of the mobile broadband sector . GSM Association, London. Kim, Yongsoo; Kelly, Tim; Raja, Siddhartha (2010). Building broadband: Strategies and policies for the . Global Information and Communication Technologies Department, World Bank, developing World Washington, DC 98

111 Impact of broadband on the economy Kolko, J. (2010). Does Broadband Boost Local Economic Development? (Public Policy Institute of California . Working paper). Retrieved from www.ppic.org/content/pubs/report/R_110JKR.pdf Small and medium Enterprises and ICT . Bangkok: Asia-Pacific Development Kotelnikov, V. (2007). Information Programme. Koutroumpis, P. (2009). The Economic Impact of Broadband on Growth: A Simultaneous Approach. Telecommunications Policy , 33, 471-485. ctations, 1993, Cambridge, MA: The MIT Press. Krugman, P. (1993). The age of diminished expe Liebenau, J., Atkinson, R. D., Kärrberg, P., Castro, D. & Ez ell, S. J. (2009, April 29). The UK's Digital Road to Recovery. Retrieved from: http://ssrn.com/abstract=1396687 Measuring the Brazilian regulatory comp liance costs to start a business. Lima Chagas, L. (2009). The George Washington University School of Business Center for Latin American Issues. Working paper. Revenue recovery continues, stock surge Merrill Lynch (2010). . September 24, 2010. Ministry of Economy and Planning of Saudi Arabia, (2010). The Eighth Development Plan. Ofcom (2008). Access and Inclusion Summary of Ofcom research on internet access, use and attitudes. London, England: OFCOM. Information and Qiang, C. Z., & Rossotto, C. M. (2009). Economic Impacts of Broadband. In Communications for Development 2009: Extending Reach and Increasing Impact , 35–50.Washington, DC: World Bank. Red.es. (2007). Inhibidores de uso de las TIC en la sociedad española . Madrid: Enter. ons Infrastructure and Economic Development: A Roeller, L-E, & Waverman, L. (2001). Telecommunicati Simultaneous Approach. The American Economic Review , 91 (4), 909-923. Rincón-Aznar, A., Robinson, C., & Vecchi, M. (2006). The productivity impact of e-commerce in the UK, 2001: Evidence from microdata . The National Institute of Economic and Social Research, United Kingdom. Sangwon, L. & Brown J. (2009). broadband deployment: Examination of A cross-country analysis of fixed adoption factors and network effect . Warrington College of Business Administration working paper The Latest Economic Development . Saudi Arabian Monetary Agency. Annual Reports (2005-2009). Shideler, D., Badasyan, N., & Taylor, L. (2007, September 28-30). The economic impact of broadband deployment in Kentucky. Telecommunication Policy Research Conference , Washington D.C. Shiu, A. & Lam, P-L. (2008). Causal relationship between telecommunication and economic growth: a study of 105 countries. Paper presented at the 17 th Biennial Conference of the International Telecommunications Society (ITS). Montreal, June 24-27. New York Times Book Review , 36. Solow, R. (1987, July 12). We'd better watch out. Thompson, H., & Garbacz, C. (2008). Broadband Impacts on State GDP: Direct and Indirect Impacts. Paper presented at the International Telecommunications Society 17th Biennial Conference, Canada. van der Ark, B., Inklaar, R. & McGuckin, R. (2002). “Changing Gear” Productivity, ICT and Services: Europe and the United States, GGDC Research Memorandum Nr. 60. van der Ark, B., Inklaar, R. & McGuckin, R. (2003). ICT and Productivity inEurope and the United States, Where Do the Differences Come From? CESifo Economic Studies , 49 (3), 295-318. Varian, H., Litan, R., Elder, A. & Shutter, J. (2002). The net impact study: the projected economic benefits of the Internet in the United States, United Kingdom, France and Germany , Available from: www.cisco.com , also available at www.itu.int/wsis/stocktaking/docs/activities /1288617396/NetImpact_Study_Report_Brookings.pdf 99

112 Impact of broadband on the economy Waverman, L., Meschi, M. & Fuss, M. (2005). The Impact of Telecoms on Economic Growth in Developing Countries. , 2, London, United Kingdom. Vodafone Policy Paper Series Waverman, L. (2009, February 29). Economic Impact of Broadband: An Empirical Study. London: LECG. 100

113 Impact of broadband on the economy Appendix A Methodologies and Data Utilized in Measuring Broadband Economic Impact Input/output analysis to measure multipliers of broadband deployment A.1 This approach focuses on determining how much value added and employment is generated through the national rollout of high speed broadband services in the access network. The importance of a sector can be measured by direct and indirect effects on the wh ole economy. The direct effects can be expressed in terms of indicators like the sector’s contribution to total value added, growth or its importance as an employer. However, these single indicators do not represent the full importance of a stimulus programme. Complex relationships develop between i ndustries because each sector sources goods and services from other sectors. In consequence, investme nts in one sector trigger demand indirectly in other sectors as well. These networked relationships mean that the effect of investment of a high speed broadband rollout programme is greater than the direct effects would suggest. The indirect effect can be measured using parameters known as multipliers; they estimate the factor by which the direct effect must be multiplied to determine the total impact of investment in a single sector on the national economy. It is important to mention, however, that, in some cases, investment does not fully materialize in job creation. First, a portion of the investment can be “leaked” to other economies due to the fact that some intermediate inputs, such as equipment, ar e manufactured in other countries. Second, while the eation in a fairly deterministic way, a number of models might predict that investment leads to job cr institutional factors could stand in the way of this effect to happen. For example, the public funds could nts act as obstacles for networks to be deployed. be available for investment but bureaucratic impedime Input-output tables enable the calculation of the impa ct of additional inputs in specific sectors on the economy as a whole. The relationships between the sectors at the inputs stage trigger additional demand and thus increase production in other sectors. The sum of all these effects is the multiplier for the total veral ways and also for several economic dimensions. volume of goods. Multipliers can be calculated in se the total volume of goods in an economy, for the There are, for example, goods-related multipliers for value of total production or for the value added. Ther e are also multipliers for labour market parameters such as the size of the workforce or the number of hours worked. A.1.1 Methodology The estimation of countercyclical effects comprises tw o steps: the estimation of investment required to fulfil the targets of the broadband stimulus plan and th e calculation of resulting economic effects through input-output analysis (see Figure A.1). The coverage and service targets established by the broadband stimulus plan are used to estimate the investment required to deploy the broadband infrastructure. These targets were compared against the current situation of broadband deployment. The comparison between the current situation and the targets allows estimating the deployment objectives in terms of number of lines to: 1) cover the “white” spots (unserved areas), 2) upgrade the “grey” spots (a reas with inferior service measured by low access type of platforms (wireless, DSL, VDSL and FTTH). speeds), and 3) deploy additional lines for different 101

114 Impact of broadband on the economy Figure A.1: Structure of input-output table Output side (use side) Sector End Input-Output table demand (each column of the input-output matrix Input side reports the monetary value of an Employment and Broadband industry's inputs and each row + - = = represents the value of an stimulus production effects industry's output). plan State Exports Investments Households Imports Volume of Goods Inputs Gross domestic product Goods/Sector Inputs + Value added = Gross production + Imports = Volume of goods Source: Based on the Bureau of Economic Analysis (BEA) make- and use-tables Once the number of lines by service target is estimated, they are multiplied by the costs per broadband line by type of platform. In order to determine the costs per line, the costs from deployment experience in Europe and the United States were relied upon, adjust ed for factors such as urban density, economies of scale, and experience curve. This calculation yields the total investment required for wireless and wireline ing to three cost categories: 1) construction labour, technologies. The total investment is then split accord 2) electronic equipment, and 3) telecommunications labour. These splits are based on cost allocations based on “real life” deployment data furnished by op erators. The resulting process yielded the amount of total investment by cost category (see Figure A.2): Figure A.2: Methodology for estimating broadband deployment costs Investment required Total investment Broadband strategy Future broadband by input targets required infrastructure • Construction • Source: Broadband •Wireline ” white • Cover spots and “ •Electronics Plan •Wireless ” spots grey upgrade “ • Telecommunications • Deploy high speed infrastructure Cost per broadband Cost structure per Current broadband broadband line line infrastructure ” spots •“ White •“ ” costs Real life • Construction labor, •“ Grey ” spots electronic equipment, (EU/US) • DSL/VDSL/cable/ telecommunications • Adjusted for density, satellite labor, civil and electronic economies of scale, engineering, etc. experience curves Legend: Input Calculated Source: Author Once the investment input is calculated, the estimati on of employment and output effects can be done. Input-output tables help calculating the direct, indirect, and induced effects of broadband network interrelationship of these three effects can be construction on employment and production. The measured through multipliers, which estimate how one unit change on the input side effects total employment change throughout the economy. 102

115 Impact of broadband on the economy To calculate employment effects resulting broadband deployment, one relies on input-output matrices published by government census and statistics depart ments (see below). However, in order to be utilized in this analysis, the input-output matrices need to be formatted to calculate the employment multipliers. Once the table is reformatted, one calculates the mult ipliers. From the I/O-table it is possible to obtain multipliers for total industry supply and additional variables as value added and employment. The calculation of the multipliers for the total industry suppl y uses the direct requirement table, which is also e (DR) is calculated by the following formula: called Leontief-Inverse. The direct requirement tabl DR = (I – A)^-1 with A = I/O-table / total industry supply (division of each cell of intermediate domestic supply by total industry supply) I = Identity matrix ease of the total industry supply by one additional The sum of the columns per industry reflects the incr unit of demand in this specific sector. A correction for the share of imports on total industry supply results in the total domestic production of the industries. The multiplying of the share of value added of total domestic industry production results in the value added multiplier. Using labour productivities it is possible to calculate the job effects now. input/output analysis A.1.2 Data utilized for The following data sets are needed to conduct the input-output analysis (see Table 6). Table A.1 – Data utilized in input-output studies Data Remarks Availability Rationale This is the investment To be calculated Breakdown of investment by sectors Investment in input that will trigger based on broadband (i.e. manufacturing of electronic growth in output and/or benchmarks programme equipment, construction, jobs telecommunication) Input-output-table To be supplied by Required to understand government inter-sectoral statistical units relationships and/or central banks Number of employed Required to calculate In the same classification as the To be supplied by input-output-table government employment effects persons statistical units To calculate investment inputs, data on current broa dband coverage and service quality is required. Data public sources, although it is available through the on current coverage is typically difficult to find from 118 service providers in each country . In other cases, the regulatory authority might have information regarding broadband coverage. Once the investment input is calculated based on the methodology described in section A.1.1., one needs to move to measuring the broadband construction effe ct. The starting point is a make-use table from input-output statistics, from which imports are excluded to reflect domestic production. In addition, as mentioned above, the matrix might need to be cons olidated in a more reduced number of industries to 118 Some countries, however, have up to date information on broadband coverage. For example, Germany possesses the Broadband Atlas produced by the Federal Ministry of Econom ics and Technology (BMWi 2009). As part of its National Broadband Plan, the United States is conducting an effort at developing its broadband map. 103

116 Impact of broadband on the economy r the US study, Katz and Suter (2008) relied on three reflect sector codes of the country being analysed. Fo original matrices and data sources: 1. Bureau of Economic Analysis: Matrix fr om the 2002 Benchmark Input-Output Accounts ndustry (“Employment and Earnings Online,” January 2. Bureau of Labor Statistics: Employment by I 2008 issue) 3. Oxford Economics: Sector share of employed persons by sector in the USA The I/O-table (Figure A.1) was built based on the Bure au of Economic Analysis (BEA) make- and use-tables using a methodology from Chamberlain Economics LLC. To obtain an I/O-table that can be used to calculate multipliers that reflect domestic production it is necessary to exclude imports from the make- table. The resulting I/O-table from BEA data has the dimension of 133*133 industries. Due to the fact that the employment data used for further calculations is in a NACE code with 28 industries the I/O-matrix is transformed to a 28*28 industries matrix. In their study of Germany, the input-output matrix supplied by Eurostat (Eurostat 2009), and originally developed by the Federal Statistical Office (Destatis 2009) was utilized. Beyond developed countries, input-output matrices might not be available and therefore, the possibility of conducting this kind of anal ysis might be greatly limited. A.1.3 Advantages and disadvantages of input/output analysis Input-output analysis to assess broadband deploy ment impact has certain advantages. Input-output tables are easy to communicate and are based on proven interlinks between sectors. The results of the short-term direct and indirect effects of broadband analysis are particularly robust in terms of estimating investment on employment and value added. for predicting investment impact, two words of However, while input-output tables are a reliable tool caution need to be given. First, input-output tables are static models reflecting the interrelationship between economic sectors at a certain point in time . Therefore, they cannot measure dynamic processes of sectoral adjustment in response to changing demand. Since those interactions may change, the matrices may lead us to overestimate or underestimate the impact of network construction. For example, if the electronic equipment industry is outsourcing jobs overseas at a fast pace, the employment impact of broadband deployment will diminish over time and part of the counter-cyclical investment will “leak” overseas. Second, it is critical to break down employment effects at the three levels estimated by the input-out table in order to gauge the true direct im pact of broadband deployment. Having said that, all these effects have been codified and therefore, with the caveat of the static nature of input-output tables, we believe that the results are quite reliabl e. Third, induced effects are calculated based on numerous assumptions, which need to be carefully formulated. For example, under close to full employment conditions, induced effects do not materialize. A.2 Econometric analysis to meas ure externalities of broadband This approach for measuring the impact of broadband on economic growth and job creation entails specifying regression models where GDP growth, employment and other output metrics are a function of broadband deployment and broadband penetration. A number of econometric studies have been generated to measure broadband's impact on GDP gr owth and employment. Because of limited data availability, most studies tend to focus on the developed countries (see Table A.2). 104

117 Impact of broadband on the economy Table A.2 – Econometric studies of broadband externalities Title Period Aggregation Author Area Year 2003-05, State 2007 The Effects of Broadband Deployment Crandall 48 US States et on Output and Employment 04-05 al. 2009 et 20 OECD 1996-07 National Czernich Broadband Infrastructure and al. Economic Growth 1999-04 County 2005 Broadband and Economic Lake County, Ford & Development FL Koutsky et al. 2010 The Impact of Broadband on Jobs and Germany 2003-06 County Katz the German Economy 2010 Does Broadband Boost Local Economic Zip Code, County, US 1992-06 Kolko Development? Individual 2009 22 OECD 2002-07 National Koutrompis The economic impact of broadband on growth: A simultaneous approach LECG Economic Impact of Broadband: An National 15 OECD 1980-07 2009 Empirical Study State/Zip Gillett et al 2006 Measuring Broadband's Economic 1999-02 US States Impact Qiang & 2009 Economic Impacts of Broadband 120 Countries 1980-06 Country Rossotto The Economic Impact of Broadband Shiedeler, Zip Code 2007 et 2003-05 Kentucky Zips Deployment in Kentucky al. Thompson & State 2009 Broadband Impacts on State GDP: 2001-06 48 US States Direct and Indirect Impacts. Garbacz Waverman Country 1980-03 2005 The Impact of Telecoms on Economic 102 Countries Growth in Developing Countries et al. Country Latin America 2004-09 Brazil 2006-07 State Chile 2003-08 Region The impact of broadband on the Katz 2010 economy India 2007-08 Telecom Circles Malaysia 2007-08 States Saudi Arabia 2007-08 Provinces Source: Compiled by the author. A.2.1 Methodology Econometric studies rely on regressions and therefore necessitate historical datasets to determine the effect of broadband on GDP, employment and other economic indicators. In order for these studies to investigate causal effects (rather than correlations), models are developed that account for factors, other than just broadband, that may influence economic i ndicators. In slightly more technical language, a regression is informed by the historical levels of an economic indicator and the factors that influence it. at maximize the amount of explained variance: that Then, given a certain model, it assigns coefficients th is to say it assigns the set of relationships between factors and the indicator that will cause the projected value of the indicator to most closely mimic what happe ned historically. For example, level of education in age; thus it is important that this variable is a given country drives GDP growth as well as broadband us present in a regression model, because if it does not, we may attribute the effects of education to broadband. In that vein, economists specify models th at account for each of the factors that influence economic indicators. An optimal regression breaks up th e variation of an indicator, such as GDP. It assigns how important each factor is in determining the indicator, or how a set of coefficients which measure 105

118 Impact of broadband on the economy much a change in each of the factors caused the economic indicator to change. Hence, we are left with the effects of broadband keeping other factors constant. This is why broadband effects are conceptualized in terms of, all other variables (education, fixed capital investment, infrastructure) being constant, and the change in broadband penetration that would explain GDP growth. There are three types of model estimation procedures used to assess the economic impact of broadband: us equations. The cross- cross-sectional regression; panel data and simultaneo sectional procedure relies on one observation per unit (country, county, region, etc. ) and in the case of studying change in variables; at least two points in time are needed. It includes independent variables such as broadband penetration, level of tertiary education, fixed capital investment (see below) and the dependent variables (such as GDP growth and employment and unemployment growth). This methodology is the most commonly used because it is rare the case when more than two years worth of data across variables is available. Given the issue of carefully estimating the direction of causa lity (what drives what), it is advisable to lag the variables collecting data for independent variables in year 1 and regressing them against dependent variables in year 2 or more. Panel data and simultaneous equations are two techni ques that further help econometric analyses study causation rather than correlation. They are among the most successful techniques that have been employed in the papers that analyse broadband economic effects. Panel data is a time series for multiple geographic areas, (i.e., it is both a time series and a cross-sectional dataset). This allows researchers to account for time fixed effects and geographical fixed effects. For example, if a dataset were confined only to 2008, it would be extremely difficult for researchers to separate the effects of the recession from their growth models. The reliance on the panel data approach is to allow unobserved differences in preferences a nd technology across regions or countries, differences that if were not taken into account in a cross-sectio nal regression could cause biased estimators (omitted variable bias).As these differences are not easily measurable, they can be treated as unobserved individual effects in the panel data regression framew ork. From an econometric point of view, the panel data model will correct the omitted variable bias, where the omitted variable captures the differences across countries. The panel data approach requires the compilation of time series for multiple geographic areas, (i.e., it is therefore, both a time series and a cross-sectional dataset). The third methodology -simultaneous equations- is used to deal with endogeneity—or a cycle where factors cause the indicators to change and vice versa. This problem is particularly pronounced in the study of broadband’s effect on GDP, GDP per capita and income. Research unilaterally agrees that broadband increases GDP and income. However, numerous studies on broadband demand have also shown that broadband is income elastic (or that an increase in income substantially increases broadband demand). Therefore, when we model the effect of broadband on income (or GDP, etc.) we must have a way to account this reverse effect (income elasticity), or ou r estimate will be biased. Simultaneous equations do just this: regressions are performed that simultan eously estimate broadband demand, supply and impact on income, solving the problem of endogeneity. The key disadvantage of this approach, particularly for developing countries, is the lack of data availability (particularly, prices and supply side variables). A.2.2 Data utilized for econometric analysis The econometric methods require the gathering of data for both dependent and independent variables in terms of their rate of change in order to determine to what extent changes in broadband penetration affects the economy. The dependent variables that are required for this analysis are GDP per capita and employment or unemployment rate. In some cases, when it comes to regional studies, analysis has been conducted considering Gross sales (see Table A.3). 106

119 Impact of broadband on the economy Table A.3 – Data utilized in econometric studies Dependent Variables Independent Variables Control Variables • GDP at starting time of period • Annual or quarterly Annual or quarterly • rate of change of rate of change of GDP • Level of education: Per cent of population with tertiary broadband degrees; Illiteracy rate; Years of schooling; Annual or quarterly • penetration participation rate in secondary school rate of change of employment Regional Investment as percentage of regional GDP • • Annual or quarterly • Per cent of households with electricity or running rate of change of water unemployment Number of projects and added value of construction • • Annual number of projects financed by the state SME’s • Number of hospitals per inhabitant; number of beds in hospitals per inhabitant • Access to financial services: Number of banking offices and bank credit per capita • Industry concentration: Contribution of financial services, commerce and manufacturing sectors to regional GDP • Importance of tourism in the region (number of domestic tourism trips) • Cost index for interstate trade costs • Cost to create new business • Regional Gini Coefficient Percentage of people living in urban centres • • Total road length per hundred sq. Km by area; Road development index Population growth rate • • Globalization Index; Global ization Index per region A number of observations need to be made regarding this list: • Level of education: in countries where the minimum level of education is secondary school, tertiary education is determining variable • Regional investment as per cent of regional GDP: It is critical to understand if the increase in employment or GDP is driven by investment othe r than broadband; this is why analyses consider e pre-existing quality of infrastructure, but also several control variables that measure not only th the change in infrastructure deployment as a proxy of investment • projects financed by the State: In the case of Number of projects and added value of construction countries where the economy is run by the government , it is imperative to introduce this variable as a way for not over-estimate the effect of broadband • Appendix B presents the variables utilized by the most important econometric studies. A.2.3 Advantages and disadvantages of econometric analysis The primary advantage of the econometric modelling is its capacity to link projections of broadband penetration, growth and productivity by relying on macro-economic causal models that rely on historical time series and cross-sectional analysis. More specifically, the methodology can provide estimates on employment growth and productivity based broadband network externalities, and generates results and identifies productivity and employment effects at the industry sector level. On the disadvantage side, since the impact of investme nt on productivity is generally lagged, time series Furthermore, the analyses require data at a fairly data sets need to be somewhat long for reliability. 107

120 Impact of broadband on the economy difficult to identify effects at the regional level, granular level (e.g. postal code). Finally, it is more although this can be addressed with disaggregated data A.3 Measuring consumer surplus of broadband The measurement of broadband consumer surplus is not as common as the econometric studies discussed above. Shane Greenstein and Ryan McDevitt, from Northwestern University, pioneered this methodology for the United States broadband sector and have recently replicated this analysis for a number of selected countries: Canada, the UK, Spain, Mexico, Brazil and China. The authors' theoretical framework is based on the notion that a new good (broadband) provides benefits that are additional to the old (dial-up access). The objective is, therefore, to calculate those benefits, which they call “the broadband bonus”. The approach the authors used was to rely on price declines to “trace out demand curve for broadband”. In doing so, they assume constant/falling nominal prices (falling real prices), which explains growing use at households. A.3.1 Data utilized in measuring the consumer surplus of broadband The data required for conducting the assessment of consumer surplus of broadband for the United States comprised the following sets: • Number of broadband subscriptions • CPI for Internet access • Household use of dial-up and broadband • Price of broadband service Estimation of revenues of household broadband services • In extending the analysis to other countries, the auth ors needed to gain access to data on number of broadband and dial-up subscribers broken down between residential and enterprise users extending back to at least 2002. In addition, residential and enterprise data series for the equivalent period is required. A.3.2 Methodology The objective of this methodology is to calculate a metric similar to the estimate of the broadband bonus developed by Greenstein and McDevitt (2009) for the Un ited States: consumer su rplus and net gain in producer revenue (broadband revenue minus lost dial-up revenue), expressed in a single currency for comparability. The first step is to calculate broadband revenues for the whole country. This is done by multiplying the number of subscribers by a price index for the given country, expressed in real terms for the last year of the series. Estimates are calculated for each year of the series. The second step is to estimate the consumer surp lus. The basic methodological premise in measuring consumer surplus is that a decline in real prices, re sulting from the combination of general price inflation with flat or no growth in nominal prices generates consumer surplus. Such declines are common in all these economies for broadband. By analysing the historical trend, one can observe the growth in consumer surplus, whose vector depends on the change in prices levels and change in revenue. The third step adjusts the estimates for the replacement of dial-up by broadband, presuming that dial-up would have been used, had broadband never diffused . Since prices for dial-up service might not be available, Greenstein and McDevitt rely on estimate s of cannibalized dialup revenue using OECD’s figures for dial-up use and an estimate of the price of dial -up service, also from OECD, whose latest published e size of dial-up revenue cannibalized by broadband. number was for the year 2000. This allows seeing th 108

121 Impact of broadband on the economy The calculation of the so-called “broadband bonus” is done by adding broadband revenue to consumer surplus and subtracting cannibalized dial-up revenue. A.3.3 Advantages and disadvantages The model measuring consumer surplus originated from broadband services presumes a stable demand, since core factors shaping demand do not change substa ntially. In that sense, results are quite valid for the short run. ve estimates. These might exclude gains to early On the other hand, the analysis can yield conservati adopters, shifts in demand linked to GDP growth, falling prices of PCs, greater capability of online system, and changing user willingness to pay. Furthermore, the methodology excludes indirect benefits. Having said that, the authors argue that given data availab ility internationally, there does not really exist an alternative approach for comparing countries. 109

122 Impact of broadband on the economy Appendix B Variables Utilized in Econometric Analyses Table B.1 – Dependent variables in econometric studies Author Employment GDPC Crandall Yearly Growth Rate Wages et al. Czernich et al. - GDPC Katz et al. (2010) Period Growth Rate - Kolko Median Income Period Growth Koutrompis (2009) - GDPC LECG - - Lehr et al Period Growth of Employment Rate, Average Salary and Median Income and Number Employed 1994-98 - - Qiang & Rossotto Shiedeler Yearly Growth Rate - et al. Thompson & Garbacz Yearly Growth Rate - Waverman et al. - Income Katz (2010) – Latam Katz (2010) – Brazil Yearly Unemployment Grow th 07/06 Growth Rate Regional GDPC 07/06 Katz (2010) – Chile Quarterly Rate 09/02 Growth Rate Regional GDPC 08/09 Katz (2010) – India Growth Rate Regional GDPC 07/08 Yearly Growth Rate Katz (2010) – Malaysia Katz (2010) – S. Arabia Yearly Unemployment Growth 08/07 Source: Compiled by the author On the independent variables side, in addition to change in broadband penetration, a number of variables are utilized both in the economic and tech nology domain (see tables B.2 and B.3). 110

123 Impact of broadband on the economy Table B.2 – Economic independent variables in econometric studies Author Establishments Investment Labor Force GDP et al. Growth Rate 05/03 or - - - Crandall 05/04 Working Age GDPC - - Czernich et al. Population Ford & Koutsky Gross Sales (Monthly) as - - Population a proxy Katz et al. Period Growth Rate - - Pop. Growth 2000-6 Kolko - - Pop. Density Average Size Koutrompis - - Employed Pop age Millions USD 15-64 LECG Per Hour Growth Rate - - - Lehr et al - Period Growth (99-02) - Number Employed and also 94-98 growth for General, ICT and Small Working Age Qiang & Rossotto Period Growth Ra te - Period Average Population Shiedeler - - - - et al. - - per Capita Thompson & Yearly Growth Rate Garbacz Participation et al. Period Growth Rate Waverman Period Average Working Age Pop - Katz (2010) – Growth Rate 06/04 or Avg Invest. Latam 09/07 03/01 or 06/04 Katz (2010) – Brazil Katz (2010) – Chile Katz (2010) – India Katz (2010) – Growth Rate 08/07 Malaysia Katz (2010) – S. Arabia Source: Compiled by the author 111

124 Impact of broadband on the economy Table B.3 – ICT independent variables in econometric studies Author Telephone Broadband Cable BB Saturation PC Penetration Penetration Penetration Lines Per Capita - - - - Crandall Czernich Yes or No (Dummy) - 1996 1996 - et al. Ford & Koutsky Yes or No Availability - - - - Dummy, 2002 - Katz et al. Penetration Growth 02-03 - - - Kolko - - - - Number of Providers (Proxy for Penetration) 2006, 1999 Penetration - - - - Koutrompis Penetration - 2007-1980 - 2007 LECG 2007 Penetration Lehr et al. Zip: Yes or No Availability - - Square - Penetration Dummy, 1999 also State: (State) Penetration - - Qiang & Rossotto Broadband subscribers per - Period 100 people, Internet users Average per 100 people Penetration, mobile and fixed Shiedeler et al. BB squared - - - Coverage Area/Total Thompson & - Lines Per Thousand People - - - Garbacz - - et al. - - Period Avg. Waverman Penetration, mobile and fixed Katz (2010) – Penetration Growth 03-06 Latam Penetration Growth 05-06 Katz (2010) – Brazil Penetration Growth 03-07 Katz (2010) – Chile Katz (2010) – Penetration Growth 07-08 India Katz (2010) – Penetration Growth 06-07 Malaysia Penetration Growth 07-08 Katz (2010) – S. Arabia Source: Compiled by the author Among control variables, the following are being utilized (see Table B.4). 112

125 Impact of broadband on the economy - - - Other Development, Research and Regulation, Industry shares Competition Investment, (HH Index), Broadband Structure, Age Family Vacat. houses, Terrain slope, - - - Racial Compos - Black Dummy - - Urbanizattion living in areas also Dummy (local), with 500 + % population people/sq km State Percentage Pop - - - - Time Trend - - - - - - - - - Size Country - - - - - - - - - - - - - - - Rent Median 2000 Median - - - - - 2002 Dummys Africa Sub Saharan Regions - - - Census - - - Roads Density - - - - Metropolitan Table B.4 – Control variables in econometric studies - - - - - Climate Temperat. (mean) 1971-01 Index - - - - - - - - - - - Climate - - - - Latam and - - - - Union Membership Union Share of Employment Education Education Primary - cent Degree Per in 1980 Grads/ Pop Years education spent on - College % GDP Grads/ Pop College et et et al. et al. Author Ford & Lehr al. Koutrompis Czernich Katz Koutsky Qiang & Rossotto Kolko Bachelor's al. LECG - Crandall 113

126 Impact of broadband on the economy - - Other Index, Population Growth, sectors, Pop Growth and trade agricultural finance, of Mining, Cost to create Growth new business, Pop Growth Rule of Law offices, Pop banking Gini, Interstate enterprises, Number of trade costs, - - Racial Compos Contribution % Black - - Urbanizattion urban centres by region People living in areas living in urban 2000, % pop - - Yes Time Trend - - - Size Country population population as a per country cent of - - - Rent - - Dummys Rural - - Roads - - - Climate Region Total road length per hundred sq. Km by area Globalization - - - Union Membership Education Schooling Years, population with some level of tertiary education High School education Tertiary in 2002 Participatio Illiteracy Rate in 2002, Years of schooling schooling Primary Grads/ Pop College secondary n rate in in 1980 Education Author & Garbacz et al. Shiedeler, –India et al. Waverman Thompson Katz (2010) – Brazil Katz (2010) Katz (2010) – Latam Katz (2010) – Chile 114

127 Impact of broadband on the economy Other in hospital, Construction projects, Gov., Facilities authorized to provide health services, Tourism, Access to potable water, Number and value of projects funded by Racial Compos Hospital, beds Urbanizattion Time Trend Size Country Rent Dummys Roads Climate Index Union Membership Road development Education Rate Literacy Author Katz (2010) –Malaysia Katz (2010) –S. Arabia 115

128 Impact of broadband on the economy 24 26 Ed.* 10.72 Tertiary 28.2821 18.9653 37.2433 21.7426 4.57912 20.1332 61.1355 1.93629 24.0214 1.57 4.39 2.68 1.99 1.52 2.78 1.97 –0.91 –1.69 –0.80 –1.57 2006–2008 GDP % Growth 7.87798 4.72729 6.719927 4.544442 0.809929 3.973686 3.037594 4.800852 0.909841 1.328786 2.985594 2004–2006 2006 5,125 3,136 5,480 4,713 5,489 8,887 5,688 3,715 4,206 1,224 4.059593 1.36 38.2594 11,893 20,864 2003 6,788 3,880 2,129 1,902 4,353 3,040 4,190 GDP per Capita (USD) 9,283 3,659 3,409 18,722 nd Economic Impact Model 2006 1.4371 5.39179 8.021321 1.940606 6.095567 5.034426 4.049863 2.533507 2.901158 1.892886 Appendix C 2005 0.20506 6.848708 4.111264 0.740275 2.392677 1.737905 1.779003 3.143415 1.825082 5.040323 1.037667 in Latin America and the Caribbean 2004 Broadband Penetration 1.02366 1.895217 0.089943 0.299831 0.595925 1.414778 3.982853 1.717288 1.014174 4.811205 0.656215 2003 0.15437 0.35591 0.064966 0.096824 0.141286 0.188041 915 1.440901 7.021701 12.14815 13.58644 20,708 23,577 2.207713 2.969482 4.347687 6.191195 4,637 8,892 4.291679 1.23 41.0088 0.054143 0.415228 3.913608 0.348107 3.446994 0.532261 0.680761 0.342105 1.018782 1.68686 2.543679 3,113 4,268 1.375247 0.88 19.1449 0.082795 0.092845 0.1931 0.344669 771 958 3.179197 0.37 17.8016 0.162127 0.396652 0.671741 1.122532 2,167 3,686 5.503492 4.25 34.0052 10.84281 11.73019 12.61278 21.76533 10,736 12,568 0.325773 0.485673 0.698415 1.010967 2,173 3,067 2.671857 0.72 17.4876 Dataset utilized for the Broadba Country Barbados Dominica Costa Rica Colombia Chile Nicaragua Bolivia Belize Dom. Rep. El Salvador Jamaica Bahamas Mexico Aruba Argentina Antigua Ecuador Grenada Brazil 116

129 Impact of broadband on the economy Ed.* Tertiary 37.8436 5.66555 8.36397 31.9533 2.17763 42.6653 5.96 35 1.55 0.91 6.64 5.28 2.13 –1.27 53.566 30.5755 18.3828 18.5758 15.8831 29.4352 12.4473 2004-2006 2006–2008 Investment GDP % Growth 7.475425 5.878246 5.898707 10.90417 5.181286 8.678897 5.011022 2004–2006 24.2819 49.6881 14.1111 25.4651 16.7392 30.7857 15.8215 2001-2003 9807 6792 4572 2006 5,212 3,310 13706 61.887 53.4364 54.0161 60.2267 51.4308 53.0914 2004-2006 7613 8575 2003 3529 3238 4,146 2,240 GDP per Capita (USD) 22858 Globalization Index 51.4184 50.7877 48.1822 59.6936 62.1831 2001-2003 2006 16.67537 1.557852 1.720966 5.160188 1.976625 3.429214 2.0579 1.13726 1.20731 0.990163 0.945924 2006-2008 2005 1.266787 0.543597 13.22805 1.333163 3.354612 0.819464 2.704353 Population CAGR 2004 Broadband Penetration 1.32228 2.19363 1.25176 1.16122 0.951002 2004-2006 0.800805 0.527338 1.216242 0.322236 9.070481 0.828051 1.354249 2003 0.453524 0.346132 0.482104 1.057967 4.176324 0.067527 0.044326 0.085065 0.221507 0.525646 2028 4236 4.378311 4.15 12.4278 0.863984 2.322909 4.237365 6.591405 4558 5544 3.959406 –1.30 12.8799 1.401598 2.215718 3.022524 3.822291 Country Bolivia 1.8949 1.76079 54.5436 Antigua Barbados 0.255813 0.260716 50.9108 54.5533 Bahamas Brazil Argentina Belize Aruba 1.7663 1.12019 45.7916 45.721 Country Saint Lucia Virgin Is. Peru Venezuela Paraguay 0.0088 0.05265 0.094848 0.265985 977 1,542 1.824164 1.08 25.9625 Saint Kitts Suriname St. Vincent Trin. & Tob. Puerto Rico Uruguay 0.812278 1.846309 6.38296 3365 5797 Panama 117

130 Impact of broadband on the economy 15.2986 19.9553 18.9363 22.5734 18.2078 44.4276 17.3638 32.6947 26.9405 19.5946 18.2587 17.0393 47.2309 21.7151 2004-2006 Investment 17.3507 21.2559 20.9322 18.3105 13.5683 23.1569 18.8677 37.6115 20.2126 50.7337 24.2193 29.0094 22.3307 16.6489 24.1471 2001-2003 57.122 72.3563 63.3931 58.7649 63.0847 41.6706 45.4881 56.9202 54.0943 58.0445 60.6223 65.6557 45.2084 60.4143 57.8427 46.3896 2004-2006 Globalization Index 70.4928 54.2123 55.5249 42.2787 50.8614 57.3015 44.1024 60.8762 44.2235 59.7317 54.9331 59.3721 63.8743 60.1884 62.6768 46.7015 2001-2003 -0.1044 1.63994 1.05594 1.45501 1.27908 0.09353 1.14469 1.66241 1.39879 1.33853 1.05036 -0.30326 0.999234 0.368135 0.982845 0.389577 0.427532 0.459817 2006-2008 Population CAGR 1.53276 1.29549 1.24472 1.75457 1.09192 0.36223 1.03077 1.74824 1.48418 1.62364 1.00239 0.259903 0.307276 0.156975 0.081186 0.453512 0.367946 2004-2006 Country Venezuela Jamaica 0.613594 0.421592 60.1394 64.6869 29.1088 30.2767 Ecuador Mexico Suriname 1.16198 0.914182 46.0395 48.1803 Trin. & Tob. Saint Lucia Dom. Rep. St. Vincent Dominica Nicaragua 1.27593 1.31174 55.0556 55.3051 26.7796 26.6601 Costa Rica Uruguay 0.097351 0.330656 62.1159 65.3281 11.9323 11.0733 Grenada Panama Saint Kitts Colombia Chile 1.04999 Paraguay 1.90125 1.79783 52.6703 56.0895 17.3689 18.979 El Salvador Virgin Is. Peru Puerto Rico Source: ITU. 118

131 Impact of broadband on the economy 6% 1% 5% 3% -7% 94% 20% 16% 52% 28% 33% 2010 0% -2% 59% 38% 34% 34% 19% 62% 11% 2009 109% 183% 1% 2% 2% 28% 58% 58% 64% 50% 21% 2008 120% 134% 0% 0% 369% 35% 14% 78% 54% 56% 20% 33% 82% 36% 2007 177% 116% 400% 0% 56% 45% 53% 62% 79% 61% 2006 100% 144% 200% 214% 100% 550% 123% 110% 49% Broadband Penetration Annual Growth -7% 33% 61% 73% 2005 108% 104% 290% 120% 500% 43% 72% 49% 38% 12% 2004 100% 232% 2000% nd Economic Impact Model 2010 4.60% 1.62% 4.73% 1.79% 4.73% 6.23% 0.33% 8.38% 0.35% 12.28% 10.51% 2009 5.36% 3.95% 1.35% 0.17% 9.05% 9.95% 0.23% 3.60% 1.54% 4.69% Appendix D 11.93% in Arab States 2008 0.98% 8.88% 2.49% 4.68% 0.06% 7.61% 4.01% 8.98% 2.22% 1.57% 0.11% 2007 0.85% 1.41% 2.34% 2.54% 83% 273% 24% 67% 66% 66% 9% 1.54% 1.57% 2.44% 0.05% 0.04% 0.62% 7.44% 4.64% 0.95% 7.35% 7.03% 0.16% 0.75% 1.01% 1.15% 0.00% 0.00% 0.00% 0.00% 0.13% 0.29% 0.61% 0.91% 1.43% 1.57% 1.70% 1.68% 49% 21% 16% 12% 10% 8% -1% 0.79% 1.23% 1.52% 1.63% 200% 1700% 50% -2% 56% 24% 7% Broadband Penetration 2006 4.64% 0.34% 0.88% 4.78% 0.44% 1.28% 5.16% 0.01% 0.88% 0.03% 4.76% 2005 0.01% 0.44% 0.19% 2.96% 0.28% 0.01% 3.12% 3.21% 0.82% 3.18% 0.18% 2004 1.53% 0.30% 2.23% 0.03% 2.00% 0.20% 0.21% 1.53% 0.00% 0.11% 0.00% Dataset utilized for the Broadba Country Kuwait 0.91% 1.10% 1.28% S. Arabia Qatar Oman 0.03% 0.54% 0.81% UAE Morocco Libya 0.00% 0.00% 0.16% Lebanon Yemen Tunisia Jordan Iraq 0.00% 0.00% 0.00% Egypt Djibouti 0.00% 0.01% 0.02% Bahrain Algeria 0.11% 0.41% 0.51% Syria 119

132 Impact of broadband on the economy 2010 2.22% 2.92% 0.79% 2.45% 0.66% 1.49% 7.93% 1.14% 10.03% 2009 0.81% 2.71% 0.95% 1.14% 1.62% 2.55% -0.67% 14.56% 11.02% 2008 0.69% 2.35% 2.00% 1.41% 1.36% 7.44% 3.60% 18.42% 13.81% 2007 1.05% 1.32% 3.00% 1.25% 2.39% 1.27% 3.34% 14.01% 20.42% Population Annual Growth 2006 2.46% 3.10% 2.07% 2.39% 0.76% 3.23% 1.11% 19.21% 12.08% 2005 5.74% 7.96% 1.00% 3.16% 3.05% 1.25% 2.70% 3.11% 15.06% 2004 5.04% 1.71% 3.32% 4.32% 4.58% 9.72% 3.12% 1.05% -2.66% 2010 0.74% 0.41% 2.55% 3.36% 5.96% 6.21% 2.27% -3.52% -2.24% 2009 3.91% 5.28% 1.96% 0.64% 7.72% -1.81% -7.23% -0.67% -5.07% 2008 0.53% 5.90% 8.44% 5.27% 4.54% 3.44% 4.27% 1.59% -6.49% 2007 5.15% 6.52% 0.20% 5.25% 1.68% 5.19% 5.18% -0.80% -5.04% 2006 4.94% 6.67% 0.00% 0.10% 4.91% 4.55% -4.69% -0.47% -0.50% 2005 2.56% 1.92% 1.93% 2.41% 5.46% 2.95% -0.09% -6.27% -0.36% Per Capita GDP Annual Growth (in constant 2005 prices) 2004 3.63% 3.53% 0.46% 1.44% 0.85% 0.57% 1.50% 2.55% 0.61% 1.23% 1.31% 1.86% 1.63% 1.36% 1.28% 5.80% 6.19% 1.73% 0.82% 3.72% 5.05% 2.17% 2.34% 7.67% 3.67% 3.72% 1.66% 0.26% 2.65% 2.02% 2.07% 4.51% 1.58% 0.66% 1.17% 1.49% 1.99% -2.74% -5.13% -8.48% -8.47% -11.18% -10.54% 7.69% 10.82% 14.67% 15.86% 14.93% 11.76% 8.29% 7.04% 6.96% 1.29% 0.26% 0.83% -8.66% -0.01% 3.13% 3.41% 3.13% 4.43% 4.09% 3.90% 3.44% 1.56% 1.80% 2.91% 3.86% 9.56% -1.68% 1.48% 1.81% 2.15% 1.96% 3.10% 3.29% 2.54% 3.08% 2.80% 3.75% 2.62% 3.48% 2.50% 4.11% 1.43% 2.94% 2.66% -5.90% 0.00% 5.94% 11.08% 0.04% 10.43% 42.63% -3.39% 3.25% -1.37% 6.34% 1.14% -2.17% Country Bahrain Morocco Syria Iraq Tunisia UAE Oman Yemen Egypt Djibouti 1.98% 1.30% 2.85% 3.13% 3.80% 3.02% 1.55% 1.81% 1.84% 1.89% 1.91% 0.68% 2.51% 2.20% Libya Kuwait S. Arabia Lebanon Qatar Algeria Jordan 120

133 Impact of broadband on the economy 8.61 2010 29.80 44.01 11.61 22.33 30.00 26.38 29.78 18.89 35.76 30.67 26.46 25.04 34.88 18.66 2009 29.72 41.18 14.78 35.97 38.93 33.23 24.36 26.13 30.18 26.81 19.26 37.49 20.45 27.90 18.93 2008 33.37 29.72 30.60 22.39 33.23 38.12 23.98 40.25 22.17 26.62 24.36 37.49 20.45 27.90 18.93 2007 30.64 34.22 24.36 21.45 24.39 26.99 37.57 32.48 27.70 26.15 20.85 37.49 25.40 20.45 18.46 13.80 16.31 24.30 21.19 2006 29.50 24.22 18.73 24.36 22.86 29.43 18.73 23.88 29.62 24.39 33.33 29.56 21.00 21.29 16.16 17.52 9.92 2005 31.51 23.15 34.15 18.24 17.98 35.00 24.42 21.77 21.94 28.80 24.36 18.97 20.38 16.43 17.55 Gross Capital Formation (As per cent of GDP) 2004 33.27 25.57 24.87 27.38 29.13 24.16 16.94 19.19 22.16 33.45 24.36 21.51 11.55 22.50 18.19 7.30 8.65 0.03 1.69 9.58 2010 -0.88 -1.41 -2.39 11.26 13.01 11.07 17.04 12.81 14.90 2.72 1.08 1.96 2009 -9.76 -2.27 -0.37 -15.06 -22.79 -11.42 -14.47 -16.48 -12.81 -17.14 -19.92 0.07 9.74 2.01 2008 -2.14 -3.46 -1.38 14.97 10.16 15.03 23.67 18.71 17.67 12.79 20.46 3.79 2.46 5.77 0.14 0.58 3.74 2.81 4.82 4.50 3.22 5.15 2007 -0.30 -0.84 -0.47 6.62 3.71 6.58 5.13 1.08 5.62 7.52 9.00 8.69 2006 -1.59 -0.14 -0.59 -0.92 10.09 Oil Prices (Impact of Price Change) 2.23 2005 -3.50 -1.22 -0.13 -1.95 14.56 11.49 20.80 14.43 10.24 15.26 12.34 19.45 18.92 9.77 1.89 2004 -1.38 -1.03 -2.64 -0.11 13.33 11.55 19.66 13.09 15.20 11.36 17.83 17.61 GDP 2000 532.72 2056.16 1285.74 1475.84 9400.81 1753.41 4612.20 1208.73 7.16 8.48 0.78 0.29 0.95 -0.95 0.71 17.41 30052.76 12489.47 2000 99.25 93.26 92.00 93.09 74.28 99.09 89.53 106.97 115.33 104.25 Human Capital Rank Country Algeria 107.78 1794.41 UAE 88.89 23270.69 Iraq 93.96 1083.82 Jordan Qatar Kuwait 95.50 19434.40 Egypt Oman 91.61 8774.93 Djibouti 32.50 753.12 Tunisia Lebanon Bahrain Syria Morocco S. Arabia Yemen Libya 118.68 6479.71 121

134 Impact of broadband on the economy 0.60 2009 75.66 79.32 19.34 75.38 99.31 75.16 100.48 165.02 2008 -3.99 72.04 11.29 77.69 53.68 98.89 67.27 107.42 169.05 2007 91.38 56.46 17.40 50.07 84.15 10.30 70.34 181.63 110.11 Bank Credit 4.84 2006 41.99 47.03 92.97 70.83 22.51 78.79 190.34 104.97 6.01 2005 30.00 44.14 98.02 70.81 49.63 73.36 176.79 109.74 7.00 2004 10.10 -1.49 -5.30 -17.27 -24.91 -16.31 47.80 68.41 41.71 71.12 90.71 42.73 104.74 177.56 2008 69.67 61.02 46.63 73.20 61.70 62.19 59.92 69.58 65.11 2007 61.49 64.88 73.09 60.31 45.42 60.52 68.50 62.16 69.30 2006 69.17 65.87 59.59 71.67 71.26 61.21 59.48 44.39 59.22 Level of Globalization 2005 71.66 70.95 67.31 42.45 58.77 58.49 62.22 59.70 68.44 2004 53.05 53.68 53.85 49.91 52.70 21.83 7.50 3.93 -3.34 -12.62 -8.89 69.99 58.60 56.84 70.01 55.56 43.24 69.10 60.32 67.46 46.16 45.78 47.95 48.35 49.19 26.65 25.03 24.13 26.49 27.47 32.32 46.13 46.52 46.63 46.99 54.50 -23.43 -43.28 -59.32 -55.63 -49.99 -65.93 74.32 73.98 69.72 70.90 71.00 50.82 57.98 66.61 74.80 88.76 114.52 67.05 67.92 63.91 65.56 64.90 70.16 61.96 62.36 68.83 65.05 65.05 57.28 58.75 58.89 61.38 61.97 34.39 27.92 27.85 32.72 29.22 41.94 42.39 43.24 44.54 45.20 44.59 30.65 35.63 33.51 36.07 36.83 45.10 Country Qatar Algeria S. Arabia Oman Iraq UAE Tunisia Yemen Djibouti Morocco Syria Libya Kuwait Egypt Bahrain Lebanon Jordan 122

135

136 r L egu a tory & market environment International Telecommunication Union L 2012 Telecommunication Development Bureau apri Place des Nations impact of broadband CH-1211 Geneva 20 Switzerland on the economy www.itu.int S b roadband Series roadband Serie b APRIL 2012 Printed in Switzerland Telecommunication Development Sector impact of broadband on the economy Geneva, 2012 04/2012

Related documents

ppacacon

ppacacon

111 TH C ONGRESS " ! LEGISLATIVE COUNSEL P RINT 111–1 2d Session COMPILATION OF PATIENT PROTECTION AND AFFORDABLE CARE ACT [As Amended Through May 1, 2010] INCLUDING P ATIENT P ROTECTION AND A FFORDAB...

More info »
C:\Users\cb47108\AppData\Local\Temp\S1177enr

C:\Users\cb47108\AppData\Local\Temp\S1177enr

S. 1177 One Hundred Fourteenth Congress of the United States of America AT THE FIRST SESSION Begun and held at the City of Washington on Tuesday, the sixth day of January, two thousand and fifteen An ...

More info »
SR288.PS

SR288.PS

113th Congress S. Report ! " SENATE 2d Session 113–288 REPORT of the SENATE SELECT COMMITTEE ON INTELLIGENCE COMMITTEE STUDY of the CENTRAL INTELLIGENCE AGENCY’S DETENTION AND INTERROGATION PROGRAM to...

More info »
G:\COMP\PHSA\PHSA.bel

G:\COMP\PHSA\PHSA.bel

G:\COMP\PHSA\PHSA-MERGED.XML PUBLIC HEALTH SERVICE ACT [As Amended Through P.L. 115–408, Enacted December 31, 2018] References in brackets ¿ ø¿ ø are to title 42, United States Code TITLE I—SHORT TITL...

More info »
aldine.indd

aldine.indd

Aldine Press Books at the Harry Ransom Humanities Research Center The University of Texas at Austin A Descriptive Catalogue by Craig W. Kallendorf ‡ Maria X. Wells Austin Harry Ransom Humanities Resea...

More info »
C:\Users\cbenson\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\h803 enr.xml

C:\Users\cbenson\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\h803 enr.xml

H. R. 803 One Hundred Thirteenth Congress of the United States of America AT THE SECOND SESSION Begun and held at the City of Washington on Friday, the third day of January, two thousand and fourteen ...

More info »
NCDOT Current STIP

NCDOT Current STIP

May 2019 NCDOT Current STIP

More info »
162

162

Black-Box Non-Black-Box Zero Knowledge Vipul Goyal Alessandra Scafuro Rafail Ostrovsky UCLA Microsoft Research UCLA USA INDIA USA [email protected] [email protected] [email protected] Ivan Viscon...

More info »
summer schedule full

summer schedule full

Summer Semester 2019 Class Schedule May 9, 2019 as of Subj Crse CRN Course Type Title Days Time Bldg/Rm Cred Instructor Course Fee Start Date - End Date Campus Part Term BB 321 Financial Accounting I ...

More info »
7B HubCityWormGearDrives

7B HubCityWormGearDrives

Hub City Worm Gear Drives Worm Gear Features B-2 to B-6 Worm Gear Drives B-7 Factory Options Rating Parameters B-8 Pre-Selection Information B-9 How To Select and Order Standard Models B-10 B-11 to B-...

More info »
C:\Users\aholmes4\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\H5515 ~1.XML

C:\Users\aholmes4\AppData\Roaming\SoftQuad\XMetaL\7.0\gen\c\H5515 ~1.XML

H. R. 5515 One Hundred Fifteenth Congress of the United States of America AT THE SECOND SESSION Begun and held at the City of Washington on Wednesday, the third day of January, two thousand and eighte...

More info »
Numerical Recipes

Numerical Recipes

Sample page from NUMERICAL RECIPES IN C: THE ART OF SCIENTIFIC COMPUTING (ISBN 0-521-43108-5) Permission is granted for internet users to make one paper copy for their own personal use. Further reprod...

More info »
oak055628

oak055628

CITY OF OAKLAND CALIFORNIA Fiscal Year 2015–17 Adopted Policy Budget

More info »
G:\COMP\SBA\SBA.bel

G:\COMP\SBA\SBA.bel

G:\COMP\SBA\SMALL BUSINESS ACT.XML SMALL BUSINESS ACT [Public Law 85–536; Approved July 18, 1958] [15 U.S.C. 631 et seq.; 72 Stat. 384 et seq.] [As Amended Through P.L. 115-416, Enacted January 3, 201...

More info »
H.R. 5515 538 TITLE XVII—REVIEW OF FOREIGN  INVESTMENT AND EXPORT CONTROLS

H.R. 5515 538 TITLE XVII—REVIEW OF FOREIGN INVESTMENT AND EXPORT CONTROLS

H. R. 5515—538 TITLE XVII—REVIEW OF FOREIGN INVESTMENT AND EXPORT CONTROLS Subtitle A—Committee on Foreign Investment in the United States Sec. 1701. Short title: Foreign Investment Risk Review Modern...

More info »
Cat11 Rev 0605.p65

Cat11 Rev 0605.p65

& TABLE OF CONTENTS H & H Specialties has manufactured equipment Introduction ... 1 for entertainment and performing arts facilities 2-5 Counterweight Rigging Systems ... ® trade- for over 35 years un...

More info »
base.dvi

base.dvi

Discrete Choice Methods with Simulation Kenneth Train University of California, Berkeley National Economic Research Associates Version dated March 8, 2002 Publisher: Cambridge University Press Schedul...

More info »
Thunder Game Notes 4 30 19

Thunder Game Notes 4 30 19

(( ((Blue Jays)) vs. Trenton Thunder New Hampshire Fisher Cats Yankees ) ) , L HP Zach Logue ( 2 - 0 , 3 . 2 7 ) vs. RHP Albe rt Abreu (0 - 2 7 . 31 ) -------------------------------------------------...

More info »
CityNT2019TentRoll 1

CityNT2019TentRoll 1

STATE OF NEW YORK 2 0 1 9 T E N T A T I V E A S S E S S M E N T R O L L PAGE 1 VALUATION DATE-JUL 01, 2018 COUNTY - Niagara T A X A B L E SECTION OF THE ROLL - 1 CITY - North Tonawanda TAX MAP NUMBER ...

More info »