__ _s _ _ _ Ab POLICY RESEARCH WORKING PAPER 2567 Policy Reform, Economic The digital divide reflects a gap in telecom access, not Growth, and the Digital lower propensity to use the Divide Internet in poor countries. Promoting access for poor households will help, but An Econometric Analysis pro-competitive policy holds the key to rapid progress in narrowing the divide. Susmita Dasgupta Somik Lall David Wheeler The, World Bank Development Research Group Infrastructure and Environment March 2001 I POLICY RESEARCH WORKING PAPER 2567 Summary findings Rapid growth of Internet use in high-income economies persistent gap in the availability of mainline telephone has raised the specter of a "digital divide" that will services. After identifying mobile telephones as a marginalize developing countries because they can promising new platform for Internet access, they use neither afford Internet access nor use it effectively when panel data to study the determinants of mobile telephone it is available. diffusion during the past decade. Their results show that Using a new cross-country data set, Dasgupta, Lall, and income explains part of the diffusion lag for poor Wheeler investigate two proximate determinants of the countries, but they also highlight the critical role of digital divide: Internet intensity (Internet subscriptions policy. Developing countries whose policies promote per telephone mainline) and access to telecom services. economic growth and private sector competition have Surprisingly, they find no gap in Internet intensity. When experienced much more rapid diffusion of mobile differences in urbanization and competition policy are telephone services. controlled for, low-income countries have intensities as Simulations based on the econometric results suggest high as those of industrial countries. While income does that feasible reforms could sharply narrow the digital not seem to matter in this context, competition policy divide during the next decade for many countries in matters a great deal. Low-income countries with high Africa, Asia, and Latin America. The authors' review of World Bank ratings for competition policy have the literature also suggests that direct access promotion significantly higher Internet intensities. would yield substantial benefits for poor households and The authors' finding on Internet intensity implies that that cost-effective intervention strategies are now the digital divide is not really new, but reflects a available. This paper-a product of Infrastructure and Environment, Development Research Group-is part of a larger effort in the group to identify effective policies for narrowing the digital divide. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Yasmin d'Souza, room MC2-635, telephone 202-473- 1449, fax 202-522-3230, email address ydsouza@worldbank.org. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at sdasgupta@worldbank.org, slall@worldbank.org, or dwheelerl@worldbank.org. March 2001. (18 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carlry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center Policy Reform, Economic Growth, and the Digital Divide: An Econometric Analysis Susmita Dasgupta Somik Lall David Wheeler Development Research Group World Bank The authors are, respectively, Senior Economist, Economist and Lead Economist in the Infrastructure/Environment Team of the Development Research Group. Our thanks to William Shaw, Robert Schware and Charles Kenny for their valuable comments on previous drafts of this paper. 1. Introduction Rapid growth of Internet use in high-income economies has raised the specter of a "digital divide" that will marginalize developing countries, because they can neither afford Internet access nor use it effectively when it is available. Dertouzos (1997) anticipates a widening income gap if the global Internet develops under laissez-faire conditions. The UN (1999) and Sachs (2000) concur, calling for substantial aid flows to narrow the technology gap. While Negroponte (1998) counters that the "leapfrog" character of the Internet will enable the poor to catch up quickly, he also advocates major increases in aid for technology development. Current Internet access patterns highlight the potential severity of the problem. Figure 1 uses data on Internet connectivity (subscriptions per capita) to plot a digital Lorenz curve for 55 countries whose total population is 4.5 billion.' Severe inequality is evident, with 90% of the world's Internet subscribers in countries whose population is 15% of the global total. If this imbalance persists, it seems likely to damage the future growth prospects of many low-income countries. Figure 1: International Inequality in Internet Access, 2000 100. 90 70° 60 50 ;nx 40 r o 30 20 10 0 0 10 20 30 40 50 60 70 80 90 100 % of Population In this paper, we investigate the sources and evolution of the digital divide using a new dataset. We focus on two dimensions of the problem: Internet use and telecommunications access. In Section 2, we review the anecdotal evidence on Internet use in developing countries. Drawing on this evidence, we specify and estimate a model of Internet diffusion via mainline telephone systems in Section 3. Sections 4-5 extend the modeling exercise to mobile telephone systems, whose rapid technological evolution may accelerate the spread of Internet access. In Section 6, we use the results to assess the potential role of economic and sectoral policy reform in narrowing the digital divide. Section 7 considers the potential for direct promotion of Internet access as a complement to policy reforms, and Section 8 concludes the paper. 1 Internet subscriber estimates have been provided to the World Bank by Pyramid Research, a subsidiary of the Economist Intelligence, Ltd. 2. Internet Use in Developing Countries: Anecdotal Evidence Although futurists regularly extol the Internet's potential benefits for poor countries, skepticism is common among development researchers. Duncombe (2000), for example, argues that Internet access may mean little to poor Africans because lack of education will prevent them from using the technology effectively. In Duncombe's typology,2 telephone systems require no literacy for independent operation while high literacy and language skills are prerequisites for using e-mail and other Internet services. Presumably, this distinction applies to technology use by poor Asian and Latin American households as well. Limited anecdotal evidence suggests that this typology may be too restrictive. Where Internet services are available, poor households do not seem hesitant or incompetent to use them. Anand (2000) reports active participation by poorly-educated village women in an Internet-based rural information system in Pondicherry, India. Similarly, Bayes (1999) notes the rapid adoption of e-mail as a cost-effective alternative to telephone calls by poor families in Bangladesh. Village women in Lethem, Guyana have used an Internet connection to sell hand-crafted hammocks for as much as $1,000 each (Romero, 2000). Faiola and Buckley (2000) report that Ashaninka Indian villagers in Peru have raised their incomes 10% by using the Internet to market organically grown oranges in Lima. In a recent survey of over 100 developing-country manufacturers, Rajkumar (2000) finds that using the Internet has significantly expanded their customer base, sales, and exports. The new service seems to have been particularly helpful for small enterprises with low er-wage, lower-skill employees. Similar findings have emerged from a survey of north Indian small enterprises by Lal (1996). These cases, while promising, do not fully address Duncomb's concern because specially- trained personnel o'ten maintain such pilot systems. For example, the weavers of Lethem, Guyana have depended on the Internet skills of a village student who was sent to the capital, Georgetown, for special training. The Peruvian Ashaninkas have depended on six tribal leaders who received eight weeks of computer training from the sponsoring agencies. In Uganda, a pilot community telecenter operates through trained personnel who access the Internet for clients (IDRC, 2000). According to Arunachalam (1999), the Pondicherry center cited by Anand (2000) also relies on the skills of specialized staff. One counter-examrple is offered by Mitra and Rana (1999), who report the results of an experiment with a high-speed touch-screen Internet link in a slum area of New Delhi. The link was provided to the community with no instructions and very little guidance, but local children demonstrated remarkable facility in learning how to use the Internet. Mitra and Rana conclude that formal training may be unnecessary for basic Internet access, although more advanced applications will undoubtedly require some instruction. 2 Duncombe (2000), Fig. 5, p. 40. 2 3. Econometric Evidence While they provide useful information, such anecdotes can only offer suggestive insights. In this paper, we use newly-available data for a more systematic assessment of cross- country growth in Internet intensity, or Internet subscriptions per telephone mainline. We focus particularly on the impacts of income, govemment policy and urbanization. Widespread concern about the digital divide reflects the view of Duncombe (2000) and others that human resource constraints may significantly reduce Internet use in low- income countries. If they are correct, then income per capita should be a major determinant of Internet intensity. A second potentially-important factor is government competition policy, which may affect both the supply of Internet services and the intensity of their use by local firms. Our proxy for this variable is the World Bank's rating of competition policy in its Country Policy and Institutional Assessment database. According to the database glossary, this rating "assesses whether the state inhibits a competitive private sector, either through direct regulation or by reserving significant economic activities for state-controlled entities. It does not assess the degree of state ownership p se, but rather the degree to which it may restrict market competition." The index varies from 1 (most inhibition of a competitive private sector) to 6 (least inhibition). As Table 1 shows, it does not automatically consign poor countries to poor ratings. In our econometric exercise, we test the proposition that Internet intensity grows more rapidly in countries with high competition policy scores. Table 1: World Bank Index: Govermnent Inhibition of Competition in the Private Sector (1999) Degree of Inhibition of Private-Sector Com etition Region Low (4-5) Medium (3) High (1-2) Africa Uganda Cameroon Somalia Botswana Ghana Liberia Malawi Nigeria Congo (Rep.) Latin America Guatemala Bolivia Ecuador Brazil Costa Rica Paraguay Peru Venezuela Asia Malaysia China Laos Philippines Indonesia Vietnam _ Bangladesh India We also include urban population in the model, to test the hypothesis that network economies cause Internet intensity to grow more quickly in urbanized societies. Although it would be useful to estimate price and quality elasticities for telephone services, we do not have sufficient data to include these variables. We test the impact of income, policy and urbanization using a Gompertz technology diffusion model first introduced by Chow (1983). In this model, a technology's adoption rate (n) is directly proportional to the log difference between current use and long-run equilibrium use (the latter determined by a set of exogenous variables Xi). 3 (1) n, = O][log n,* log n,] For estimation, we approximate this relation as (2) log n, - log nt, = O[log n,' - log n,-,] where (3) log n: = 80 + Efij log X j, J Substituting and adding a random error term we obtain the following estimating equation: (4) log n, - log n,, = O + E 0j log Xj, - Slog n,-l + , We fit the model to cross-country data on the growth of Internet intensity from 1990 to 1997, using the following equation and variables: (S)logNft -logN>, = - -6lognj -l + logujt-l +OAlogYj,, +OAlogC>1 +jk + ef k where Njt = Internet subscribers/telephone mainlines Ujt = Size of urban population -jt = Income per capita (it = Index of competition policy3 Rj = Vector of regional dummy variables (Sub-Saharan Africa, Middle East/North Africa, Asia, Latin America) Data on Internet subscribers and telephone mainline connections have been supplied by Pyramid Research, a subsidiary of the Economist Intelligence Unit that specializes in telecom data. We have drawn measures of income per capita and urban population from World Bank sources, along with the competition policy index. We summarize the estimation results in Table 2. Although the 44-country sample is not large, the regression statistics are quite robust. The model explains about 96% of total sample variation in the growth rate of Internet intensity. The estimated value of the diffusion parameter is almost exactly one and highly significant, suggesting little adjustment lag beyond the 7-year span of the data. Estimated elasticities for urban population and competition policy are large, positive and highly significant (.83 and .59, respectively, after accounting for the estimated value of 0 (1.04)). The competition policy result is particularly important, because policy changes can occur relatively quickly. Suppose, for example, that an African country adopts measures that improve its 3We have used the 1995 value of this index in the regression because earlier values are not available. 4 competition policy rating from 2 to 3. Our results suggest that this change (a 50% improvement) will increase Internet intensity by approximately 30%. Table 2: Regression Results for Equation (5) Dep. Variable: Log (I/T)19 - Log(IjT)goa9b| Variables Constant -14.76** (7.3)c Log (1IT)1990 -1.04** (31.8) Log (Urban Population)lso 0.86** (12.1) Log (Income Per Capita)lso -0.10 (0.7) Log (Policy Index)199s 0.61** (2.1) Asia 1.68** (4.1) Latin America 1.15** (3.9) Middle East, North Africa -0.04 (0.1) Sub-Saharan Africa 1.26** (2.9) Observations 44 R2 .97 Adjusted R2 .96 F-Statistic 135.1 a I/T = Total Internet subscribers/total mainline connections b Minimum Internet connections set to 1 unit in 1990 c t-statistics in parentheses; White heteroskedasticity-consistent standard error and covariance estimates ** Significant at .05 or greater by the standard criteria The income parameter estimate contrasts strongly with the policy result: Economic development does not have a significant impact on Internet intensity. Indeed, African, Asian and Latin American countries appear (ceteris paribus) to have higher Internet intensities than their counterparts in the OECD. This result is obviously not consistent with the conventional view of the digital divide, which highlights the role of the human resource gap and other development-related variables. A possible counter-argument is 5 that rapid growth of Internet intensity in low-income countries will stall once the urban "elite" (i.e., the small minority of high-income households) have all subscribed. At present, we do not have sufficiently-disaggregated data to test for this kind of discontinuity. However, relatively high-income urban households in developing countries have significantly lower average incomes and education levels than their status counterparts in the OECD economies. For this reason, it seems reasonable to suppose that development would register strongly in the early years of Internet diffusion if it were an important determinant of intensity growth. In any case, the available evidence suggests that differences in competition policies have much greater impact on Internet intensity than differences in income. While income apparently has no impact on Internet intensity (Internet subscriptions per telephone mainline), it obviously has an effect on Internet connectivity (per capita Internet use). The sample countries used to construct Figure 1 fall into three groups whose connectivity differs by orders of magnitude. Group 1 (50 - 400 subscribers/'000) includes the OECD economies, along with two NIE's (Korea, Singapore) and one Middle Eastern oil state (UAE). Group 11 (10 - 50 subscribers/1000) includes the more highly- developed economies of Eastern Europe and Latin America, along with Malaysia and South Africa. Group III (0-10 subscribers/'000) includes all other Sub-Saharan African states and low-income countries in Asia, Latin America, Eastern Europe and the Former Soviet Union. This striking difference between intensity and connectivity suggests that the digital divide is not really new. It reflects a long-standing disparity in telecommunications access, rather than an additional handicap for developing countries. 4. Telecom InnoVAtion and Future Internet Access During the 1990's, most users accessed the Internet by connecting a PC to a telephone line via modem or, in the case of large businesses and universities, to a high-speed internal network. Much of the pessimism about Internet use in low-income countries stems from the belief that it will continue to require skilled use of costly PC systems. However, low-cost Internet access devices are rapidly appearing. For developing countries, the most promising innovation may be the Wireless Application Protocol (WAP), a universal open standard for bringing Internet content to mobile telephones.4 WAP will be embedded in every new digital cellular phone by June, 2001 (IDC, 2000), and these phones will be sold in Africa, Asia and Latin America as well as in the OECD economies. Data transmission rates via WAP-enabled cell phones should converge rapidly with rates that are currently available on high-speed corporate and university networks. Harrow (2000) reports that currently-available technologies can transmit wireless data at ISDN- level speeds (144 Kilobits/sec.); DSL-level speeds (1 Megabit/sec. (Mb)) are imminent; and several leading cell-phone suppliers have projected access speeds of 5.2 Mb by 2003. 4 For detailed information on WAP systems, see www.wapforum.org. 6 WAP-enable cell phones are spreading very rapidly in Japan and Western Europe. According to the Japanese Ministry of Posts and Telecommunications, approximately 10 million Japanese users accessed the Net through mobile communications devices in May, 2000 -- an expansion to 40% of total Internet users since the advent of wireless access (Reuters, 2000). After analyzing recent growth trends, IDC has projected a worldwide user base of over 700 million mobile Internet subscribers by 2002, significantly exceeding its projection of 500 million users who will access the Net through phone lines and wired networks (IDC, 2000). Anticipating rapid expansion of WAP-enabled phones, hardware manufacturers are developing color displays and input devices that will approach PC functionality in very small devices. During the next few years, mobile telephone systems appear poised to join telephone mainlines and cable TV as major platforms for Internet expansion in developing countries. Mobile phone systems have particularly important implications for the digital divide, because they can expand rapidly into peri-urban and rural areas where most poor households are located. 5. Explaining Mobile Telephone Growth in the 1990's During the 1990's, privatization and deregulation of telephone systems accelerated access to telecom services in many developing countries. China's telecom penetration rate rose 25-fold (from 6 to 147 per thousand population); India's 33-fold (from 1 to 33); Latin America's quintupled (from 48 to 248); and Sub-Saharan Africa's rose more than twofold (from 5 to 13). Latin America's current penetration rate is more than half of the OECD rate in 1990; South Africa and East Asia (excluding China) are not far behind. The most significant feature of this growth has been the spread of mobile phone systems. Figure 2 shows that mobile systems now operate in 85% of the International Telecommunication Union's 194 member countries. From a near-zero base in 1990, mobile subscriptions have increased to nearly 45% of fixed subscriptions in East Asia (excluding China); 40% in South Africa; 35% in Latin America; and 30% in China, Sub- Saharan Africa (excluding South Africa), and the Middle East-North Africa region. While rapid growth has been the rule, developing countries have exhibited very different rates of increase and patterns of regional diffusion. In Latin America, annual mobile phone expansion rates have varied from 4% (Guyana) to 171% (Bolivia). In Asia, the growth rate has varied from 28% (Thailand) to 98% (China). Recent data on the spread of mobile telephone systems within developing countries show that geographic coverage is not necessarily limited to a few large cities. In Africa, for example, mobile phone systems are rapidly moving toward coverage of the most populous regions in Senegal, 5 National coverage maps for the GSM (Global System for Mobile Communications) are available from the GSM consortium's Website at www.gsmworld.com. Comparison with population distributions is possible using high-resolution population density maps available from the National Center for Geographic Information and Analysis, University of California at Santa Barbara, at http:l/www.ncgia.ucsb.edu/pubs/gdp/pop.html#GLOBAL. 7 Cote dlvoire, Uganda and South Africa. In other countries (e.g., Guinea, Sierra Leone), coverage is largely confined to the capital cities. Figure 2: International Expansion of Cellular Telephone Systems Cell Phone Diffusion: % of 194 ITU Countries 70 50 40 30 20 10 00 1975 1977 1979 1981 1983 1985 1987 1989 1991 1903 1995 1997 Data Source: rlU To explain these wide variations, we employ another version of the diffusion model in equation (4) for the period 1990 - 1999. We include initial income, policy and urbanization, while adding an income growth term to allow for changes during the 1990's. Recognizing that the growth rates of income and mobile phones may be jointly determined, we have tested the impact of mobile phone growth on income growth in a separate instrumental-variables exercise.6 Finding no significant impact, we have employed a single-equation model (6) for this paper. logM3, -log M l,1 =A -0logMj,i +O/jloguj,-i +/Alogyj,- +OA logcj,t ()+ A4(logy Y_ log Yi, +) + where Mit = Mobile telephone subscriptions U1t = Size of urban population Yjt = Income per capita Cit = Index of competition policy Table 3 reports the regression result, which explains about 77% of the variation in mobile phone growth rates across 99 countries. All estimated parameters have the expected signs and are highly significant by the conventional criteria. The estimated value of 6Our growth equation replicates a recent specification by Easterly (2000), with the addition of a term to capture the possible effect of mobile phone growth. Results are available from the authors on request. 8 Table 3: Regression Results for Equation (6) Dep. Variable: Log (Mobile),s - Log (Mobile)199oa Variables Constant -9.16** (3.2)b Log (Mobile)199o -0.84** (13.0) Log (Urban Population),o 0.78** (6.7) Log (Income Per Capita 0.78** (IPC))irn (3.0) Log (Policy Index)1995 1.06** (2.5) Log (IPC1999) - Log (IPC1990) 2.12** (4.0) Observations 99 R2 .78 Adjusted R2 .77 F-Statistic 49.8 a Minimum mobile subscriptions set to 1 unit in 1990 b t-statistics in parentheses; White heteroskedasticity-consistent standard error and covariance estimates * * Significant at .05 or greater by the standard criteria o (.84) suggests very rapid adjustment toward long-run equilibrium subscription levels in the 1990's. The estimated elasticities for urban population, income per capita and pro- competitive policy are positive, large and highly significant (.93, .93 and 1.26, after accounting for the estimated value of 0). Income growth during the 1990's also has a large estimated elasticity (2.12). Here we see strong evidence of the digital divide: Across countries, long-run equilibrium subscriptions have an approximately unit-elastic relationship with income per capita. This implies a large, persistent gap between high- and low-income countries, at least until the former achieve very high rates of penetration. The urban population result suggests that network economies provide a significant advantage for more urbanized societies. However, the results for the World Bank index and economic growth also suggest that policy matters a great deal. Income growth induces mobile phone expansion with an elasticity greater than two. Although recent studies of economic growth assign increased importance to exogenous shocks (Easterly, et. al., 1993; Easterly, 2000), they continue to find that countries with liberal economic policies enjoy a significant growth advantage. 9 The estimated elasticity for competition policy implies that measures which improve a country's policy rating from 2 to 3 will increase long-run equilibrium mobile phone subscriptions by 63%. In the regression result, this is equivalent to 68% higher income7, or about nineteen years of economic growth at the recent average rate for low-income countries (2.8%). Income remains a prime determinant of the digital divide, but our results suggest that appropriate economic and competition policies can sharply narrow the gap. 6. Policy Reform and Access Expansion To illustrate the implications of policy reform, we use the parameter estimates in Table 3 to predict future mobile telephone growth under different policy regimes. Our baseline forecast for 2000 - 2009 substitutes current values of three righthand variables (urban population, income, the World Bank competition policy index) for the earlier values used in the regression. We use each country's actual log-difference in income for the 1990's as our forecast for that variable. To simulate strongly pro-competitive policy, we set the World Bank index at level 5 for all sample countries. While this would require substantial reform in many countries, it has already been attained by Uganda, Malawi, Botswana, Guatemala, Brazil, Peru, Philippines, Malaysia and Bangladesh, among others Table 1). Assuming general adoption of more liberal economic policies and no adverse shocks in the world economy, we also set annual income growth at 5% (the upper quartile rate for our sample countries in the 1990's). Experiments with separate simulations have shown that each component (policy rating of 5; 5% annual growth) is responsible for about half of our projected change Table 4 summarizes the projections in millions of subscriptions for Asia and Latin America, while Table 5 reports results in thousands for Sub-Saharan Africa to facilitate comparisons among countries in that region. Even the baseline projections suggest very rapid changes ahead. China dominates the Asian estimates, with projected growth from 24 million subscribers in 1998 to 102 million in 2009.8 India's projection is more moderate but still impressive, with growth from 1 million users in 1998 to 21 million in the baseline projection and 42 million in the more optimistic forecast. Large gains are also projected for the other eight Asian countries in the sample. The baseline results for 16 Latin American countries indicate growth from 18 million mobile subscribers to 80 million, while the more optimistic forecast is 145 million. Argentina and Brazil account for over half of the total increase. More moderate gains are projected for the Middle East and North Africa, but the optimistic scenario still projects over 17 million subscribers in the six sample countries by 2009. 7 To produce this estimate, we divide .63 by the mobile-phone-growth elasticity of income (.93, after accounting for 0). 8 In Table 4, China's baseline projection is higher than the "optimistic" projection because its growth rate in the 1990's was significantly higher than 5% per year. 10 Table 4: Mobile Telephone Diffusion in Asia, North Africa, and Latin America Population Mobile Phone Subscribers (hillions) (Million ;) 2009 Pop % With Region/Country 1997 2009 1998 2009 (Growth 5%; Mobile (Baseline) Policy = 5) Phones (5/5 Forecast) EAST ASIA China 1,227 1,399 23.9 102.0 81.9 6 Cambodia 11 15 0.1 0.1 0.2 2 Korea (Rep. of) 46 52 14.0 24.2 36.5 70 Lao P.D.R. 5 7 0.0 0.1 0.2 3 Malaysia 21 28 2.2 5.9 8.2 30 Philippines 73 96 1.6 3.7 12.1 13 Total (Excl. China) 156 197 17.9 33.9 57.3 29 SOUTH ASIA India 961 1152 1.2 20.8 41.5 3 Bangladesh 124 150 0.1 1.9 3.6 2 Pakistan 137 193 0.2 2.4 6.8 4 Sri Lanka 18 21 0.2 0.7 1.2 6 Total (Excl. India) 279 364 0.5 5.0 11.6 3 LATIN AMERICA Argentina 36 42 2.8 23.9 29.0 69 Bolivia 8 11 0.2 0.4 1.1 10 Brazil 164 194 7.8 26.2 61.9 32 Chile 15 18 1.0 11.7 9_X_ .1 _5 Colombia 38 47 1.8 4.7 14.1 30 Costa Rica 4 5 0.1 0.3| 1.0 1 9 Dominican Rep. ___ _ 0_ 3 1_ 2 2.2 22 Ecuador | 2 1 O.E 2.1 14 El Salvador 6 8 0.1 0_7__1 4 Guatemala 11| 1 5 0.1 0.7 1.4 9 Honduras 61 9 Xl 0.5 1.6 19 Nicaragua __r 51 71 0._ 0.3 0.6 8 Panama 3 4 0.1 O.; 0.8 21 Peru 25 32 0.7 4.6 6.4 20 Uruguay 3 3 0.2 1.3 2.3 73 Venezuela 23 30 2.0 2.0 9.4 32 Total 368 451 17.6 79.5 144.8 32 MIDDLE EAST, N. AFRICA Egypt 60 76 0.1 2.9 7.1 9 Jordan 4 7 0.1 0.6 1.1 15 Lebanon 4 5 0.5 2.2 1.5 30 Morocco 28 35 0.1 1.4 4.4 12 Tunisia 9 11 0.0 1.2 2.9 26 Yemen 16 27 0.0 0.1 0.3 1 Total 121 162 0.8 8.5 17.3 11 11 Table 5: Mobile Telephone Diffusion in Sub-Saharan Africa Population Mobile Phone Subscribers (Millions) (Thousands) 2009 Pop % With Region/Country 1997 2009 1998 2009 (Growth 5%; Mobile (Baseline) Policy = 5) Phones (5/5 Forecast) SUB-SAHARAN AFRICA South Africa 38 47 2,500 3,651 14,574 31.3 Angola 11 16 10 58 726 4.6 Benin 6 8 6 77 230 2.7 Botswana 1 1 23 191 484 37.7 Burkina Faso 11 15 3 40 183 1.2 Burundi 7 10 1 3 40 0.4 Cameroon 14 20 7 701 4,518 22.9 Cote d'lvoire 15 21 91 252 1,012 4.8 Ghana 18 25 49 804 2,590 10.5 Guinea 7 10 22 132 412 4.3 Kenya 28 38 11 298 1,783 4.7 Lesotho 2 3 10 49 102 4.0 Malawi 10 14 11 39 139 1.0 Namibia 2 3 20 114 276 10.2 Niger 10 15 1 26 167 1.1 Nigeria 118 166 20 732 3,640 2.2 Senegal 9 13 22 138 554 4.4 Tanzania 31 44 38 90 483 1.1 Togo 4 6 8 31 217 3.8 Uganda 20 29 30 250 321 1.1 Zambia 9 13 5 59 296 2.4 Zimbabwe 11 14 55 200 1,027 7.1 Total Outside South 344 482 441 4,285 19,199. 4.0 Africa I I I I In Sub-Saharan Africa (Table 5), the baseline forecast is 3.7 million South African subscribers and 4.3 million in the other Sub-Saharan countries. Cameroon, Ghana and Nigeria figure most prominently in these projections. In the optimistic case, subscriptions outside of South Africa grow to 19.2 million. While these African projections large in absolute terms, they still imply very low per- capita subscriber rates. At least another decade of rapid expansion after 2009 would be required for convergence with current OECD penetration rates, even under optimistic assumptions. To illustrate the problem, the last column in Table 5 provides estimates of mobile phones per capita in the optimistic scenario. Only a few countries have sizable 12 penetration rates by 2009: Botswana (38%), Cameroon (23%), Ghana (11%) and Naniibia (10%). In Latin America, by contrast, the optimistic scenario projects penetration rates close to current rates in the OECD. Particularly striking examples in Table 4 are provided by Argentina (69%), Uruguay (73%) and Chile (50%). Brazil's forecast is 32%, but this reflects a huge user base (62 million, up from 8 million in 1998). 7. Benefits of Access Promotion Our results suggest that countries with progressive economic and sectoral policies will narrow the digital divide substantially during the coming decade. Large telecom access gaps will persist, however, and particularly for poor households. Targeted intervention could narrow the divide, but social opportunity costs are high in countries where basic educational and health services remain scarce. Are access subsidies warranted? Our own results are cautionary, since we find no causal role for mobile telephone expansion in economic growth during the 1990's (Section 5). However, Easterly (2000), Hardy (1980) and Roller/Waverman (1996) find a significant growth-promoting role for telephone connectivity, and many other studies suggest that telecom expansion has increased the incomes of poor households. These studies have documented three major sources of change: income for telecom service providers, reduced costs for household producers who use these services, and increased incomes from better market information. Income for Telecom Service Providers Provision of telecom services has become a source of income for many households in developing countries. In urban India, the state telecommunications company provides metered telephones to micro-entrepreneurs (Pitroda (1993)). These operators, who are frequently handicapped, charge cash for calls and are billed six times per year by the telephone service. Their bills are discounted 20-25% as a commission for their services. According to Ramadorai (2000), India currently has 2 million urban kiosks, each of which earns from $US 460 to 690 per year. In rural Bangladesh, Richardson, et. al. (2000) report that operators in Grameen's Village Phone program earn a net income of about $US 300 per year. This is 25% of household income for the operators, who are usually women with incomes below the local average (Bayes, 1999). The Village Phone program currently employs operators in 950 villages and plans to expand services to 40,000 villages. Bayes reports an average household size of 6.2 for operators, so a fully- implemented program would increase income by about 33% for 250,000 rural villagers. Fuchs (1998) reports that 1,000 rural telecenters in Senegal created approximately 4,000 new jobs during the period 1992-95 and generated an average net income of $1,600 dollars per year for their owner-operators. According to Braga, et. al. (2000), at least 6,000 telecenters are now operating in Senegal. Income for Telecom Service Users A few systematic studies of telecom's impact on user incomes have been undertaken in South Asia. These studies identify two major sources of increase in incomes: expansion 13 of contact with family expatriate workers, leading to increased repatriation of earnings, and increased revenues for farmers who can get price information without relying on quotes from local middlemen. Pitroda (1993) cites a government study of the impact of a new telephone exchange on a town of 5,000 in Karnataka, India. During the year following installation of a 100-line exchange, the results suggest a rapid expansion of business activity, establishment of frequent communications with relatives in North America and Europe, and an increase of 80% in local bank deposits. In a sample-based statistical comparison of villages with and without Village Phone services in Bangladesh, Bayes (1999) finds price differences of approximately 7% for paddy, 8% for eggs, and 9% for foreign exchange, all attributable to improved knowledge of market conditions. The World Bank (1999) reports a much larger benefit -- price increases of 55% -- for farmers in Sri Lanka who have access to telephones. Consumer Surplus Measures While direct estimates of income and employment impacts are useful, more general welfare measures provide the best summary information about the value of telecom services to poor households. Using a variety of methods, numerous studies have estimated consumer surplus in this context (the difference between households' actual payments for telecom services and the amount they would be willing to pay).9 Saunders, et. al. (1994) report savings between 2.5 and 5.5 times the cost of a telephone call in Andhra Pradesh State, India; 4 times telephone rental cost in Egypt; 10 times the cost of business calls in Kenya; and between 13.5 and 20.1 times the cost of telephone calls in Northern Luzon and Northern Mindanao, Philippines. The same publication reports that measured revenues from 92 public phones in 82 Costa Rican villages represent only 81% of the benefits received by telephone users. The Asian Development Bank (1996) reports savings of 2.6 times the cost of telephone calls in rural Thailand. For the Bangladesh Village Phone program, Richardson, et. al. (2000) estimate consumer surplus at 3-10% of mean monthly household income. Torero (2000) finds that increased access to residential phone services in Lima, Peru since privatization has generated an annual per capita consumer surplus of approximately $US 40 for very low-income consumers, $28 for low- income consumers and $146 for medium-income consumers. Since Peruvian income per capita is approximately $2,500 per year, these are significant benefits. Cost-Effective Intervention While telecom benefits for the poor seem substantial, access promotion programs may be undermined by institutional defects such as corruption, costly administration and benefits capture by more affluent households. Wellenius (1997) argues that the public sector can promote access cost-effectively through appropriate use of economic incentives. In Chile, for example, the government has promoted access for rural communities by auctioning targeted subsidies to private telecom operators. Coupled with careful monitoring of service delivery requirements, this approach has broadened access for low- income Chilean households at very low unit cost to the public sector. India (Ramadorai, 2000), Bangladesh (Richardson, 2000) and Senegal (Fuchs, 1998) have rapidly expanded 9 For a general survey of the evidence, see Bedia (1999). 14 telecom access through micro-credit to small entrepreneurs whose revenues are more than sufficient to cover incremental service costs. Although such documented cases remain limited, the available evidence suggests that poor households value telecom services so highly that rapid expansion is possible at relatively low cost to the public sector. 8. Summary and Conclusions In this paper, we have investigated the determinants of the "digital divide" between high- and low-income countries. Surprisingly, we find that there is no gap in Internet intensity (Internet subscriptions per telephone mainline). Controlling for other factors, developing countries have intensities as high as those of developed countries. While we find no reflection of the human resource gap in this context, our results suggest that policy differences matter a great deal. We conclude that the digital divide is not really new, but reflects a long-standing gap in per-capita availability of mainline telecom services. After identifying mobile telephones as a promising new platform for Internet access, we have studied the determinants of mobile telephone diffusion during the past decade. Our results show that income differentials matter, but they also highlight the critical role of progressive policies. Policy simulations based on our results suggest that feasible reforms could sharply narrow the digital divide during the next decade for many countries in Africa, Asia and Latin America. Our review of the literature also suggests that access promotion would yield substantial benefits for poor households, and that cost- effective intervention strategies are available. 15 References Anand, A., 2000, "ICTs: What Digital Divide?" Women's Feature Service, Pondicherry, India, August 22. Arunachalam, S., 1999, "Information and Knowledge in the Age of Electronic Communication: A Developing Country Perspective," Chennai, India (mimeo.). Asian Development Bank, 1996, "Towards Universal Access: Socioeconomic Impact Study of Rural Telecommunications in Thailand," Midas Agronomics Report, T.A. 2381, November. Bayes, A., J. von Braun and R. Akhter, 1999, "Village Pay Phones and Poverty Reduction: Insights from a Grameen Bank Initiative in Bangladesh," Center for Development Research, University of Bonn, Discussion Paper on Development Policy, No. 8, June. Bedia, A., 1999, "The Role of Information and Communication Technologies in Economic Development -- A Partial Survey," Center for Development Research, University of Bonn, Discussion Papers on Development Policy, No. 7, July. Braga, C., C. Kenny, C. Qiang, D. Crisafulli, D. Di Martino, R. Eskinazi, R. Schware and W. Kerr-Smith, 2000, "The Networking Revolution: Opportunities and Challenges for Developing Countries," InfoDev Working Paper, World Bank, June. Camp, J. and B. Anderson, 1999, "Grameen Phone: Empowering the Poor through Connectivity," Information Impacts Online Magazine, December. Chow, G., 1983, Econometrics (New York: McGraw-Hill). Dertouzos, M., 1997, What Will Be: How the New World of Information Will Change Our Lives (New York: HarperEdge). Dollar, D., and A. Kraay, 2000, "Growth Is Good for the Poor," Development Research Group, World Bank, March (mimeo.). Duncombe, R., 2000, "Information and Communication Technology, Poverty and Development in sub-Saharan Africa," Institute for Development Policy and Management, University of Manchester, UK, June. Easterly, W., 2000, "The Lost Decades: Explaining Developing Countries' Stagnation 1980-1998," World Bank Development Research Group Working Paper. Easterly, W., M. Kremer, L. Pritchett, and L. Summers. "Good Policy or Good Luck? Country Growth Performance and Temporary Shocks," Journal of Monetary Economics, Vol. 32, December 1993, 459-483. 16 Faiola, A. and S. Buckley, 2000, "Poor in Latin America Embrace Net's Promise,' Washington Post, July 9, p. AO1. Fuchs, R., 1998, "Little Engines That Did: Case Histories from the Global Telecentre Movement," report prepared for the Intemational Development Research Center, Canada by Futureworks, Inc., June. Hardy, A., 1980, "The Role of the Telephone in Economic Development," Telecommunications Policy, Vol. 4, pp. 278-286. Harrow, J., 2000, "www.wireless.world," from J. Harrow, "The Rapidly-Changing Face of Computers," http://www.compaq.com/rcfoc, April 24. IDC, 2000, "Mobile Access to Internet Gains Momentum," IDC Forecaster Newsletter, March 21. IDRC Canada, "Nabweru Multipurpose Community Telecentre," summary report at http:Hlwww.acacia.or.ug/html/nabweru.html. Lake, D., 2000, "Moving Beyond PC's," The Standard Online (www.thestandard.com), April 3. Lal, K., 1996, "Information Technology, International Orientation and Performance: A Case Study of Electrical and Electronic Goods Manufacturing Firms in India," Information Economics and Policy, Vol. 8, pp. 269-280. Mitra, S. and V. Rana, 1999, "Children and the Internet: An Experiment with Minimally Invasive Education in India," World Bank, infoDev program, infoDev/iicd Stories online at http://www.iicd.org/base/show_story?id=3778. Negroponte, N., 1998, "The Third Shall be First: The Net Leverages Latecomers in the Developing World," Wired Magazine, January. Pitroda, S., 1993, "Development, Democracy and the Village Telephone," Harvard Business Review, November-December. Rajkumar, A. S., 2000, "The Impact of E-Commerce on Developing-Country Firms: A Survey of Participants in Alibaba.com," World Bank, Development Research Group, July (mimeo.). Ramadorai, S., 2000, "E-Commerce in the Next Millennium: India in Perspective," presentation to the Commonwealth Business Forum, South Africa, January (cited at http://www.tcs.com/news/ceo_notes/htdocs/ fhOO/jan0_0commonwealth.htm). Reuters, 2000, "It's a Wireless World in Japan," June 20. 17 Richardson, D., R. Ramirez and M. Haq, 2000, "Grameen Telecom's Village Phone Programme in Rural Bangladesh: A Multi-Media Case Study," report to the Canadian International Development Agency by TeleCommons Development Group, March. Roller, L. and L. Waverman, 1996, "Telecommunications Infrastructure and Economic Development: A Simultaneous Approach," WZB Discussion Paper, Berlin. Romero, S., 2000, "Weavers go Dot-Com, and Elders Move In," New York Times, March 28. Sachs, J., 2000, "A New Map of the World," The Economist Saunders, R., J. Warford and B. Wellenius, 1994, Telecommunications and Economic Development, 2nd Edition (World Bank: Washington, D.C.) Torero, Maximo, 2000, "The Access and Welfare Impacts of Telecommunications Technology in Peru," Center for Development Research Working Paper, No. 27, Bonn University, June. United Nations, 1999, Human Development Report, Chapter 2, "New Technologies and the Global Race for Knowledge," (United Nations Development Programme: New York) Willenius, B., 1997, "Extending Telecommunications Service to Rural Areas-The Chilean Experience: Awarding Subsidies Through Competitive Bidding," World Bank, Public Policy for the Private Sector, Note No. 105, February. World Bank, 1999, World Development Report: Knowledge for Development, (World Bank: Washington, D.C.) 18 Policy Research Working Paper Series Contact Title Author Date for paper WPS2554 Administrative Costs and the Estelle James February 2001 A. Yaptenco Organization of Individual James Smalhout 31823 Retirement Account Systems: Dimitri Vittas A Comparative Perspective WPS2555 Implicit Pension Debt, Transition Yan Wang February 2001 A. Datoloum Cost, Options, and Impact of Dianqing Xu 36334 China's Pension Reform: Zhi Wang A Computable General Equilibrium Fan Zhai Analysis WPS2556 Household Strategies for Coping with Michael M. Lokshin February 2001 P. Sader Poverty and Social Exclusion in Ruslan Yemtsov 33902 Post-Crisis Russia WPS2557 Decentralization and Accountability: Stuti Khemani February 2001 H. Sladovich Are Voters More Vigilant in Local than 37698 in National Elections? WPS2558 Growth, Inequality, and Poverty: Martin Ravallion February 2001 P. Sader Looking beyond Averages 33902 WPS2559 Deposit Insurance as Private Club: Thorsten Beck February 2001 P. Sintim-Aboagye Is Germany a Model? 38526 WPS2560 Catastrophe Risk Management: John D. Pollner February 2001 J. Pollner Using Alternative Risk Financing 30079 and Insurance Pooling Mechanisms WPS2561 Democracy and Income Inequality: Mark Gradstein March 2001 P. Sader An Empirical Analysis Branko Milanovic 33902 Yvonne Ying WPS2562 Decomposing World Income Branko Milanovic March 2001 P. Sader Distribution: Does the World Have a Shlomo Yitzhaki 33902 Middle Class? WPS2563 A Race to the Top? A Case Study of Tsunehiro Otsuki March 2001 L. Tabada Food Safety Standards and African John S. Wilson 36896 Exports Mirvat Sewadeh WPS2564 Economic Analysis of World Bank Ayesha Yaqub Vawda March 2001 E. James Education Projects and Project Peter Moock 31756 Outcomes J. Price Gittinger Harry Anthony Patrinos WPS2565 Does Fiscal Decentralization Improve David A. Robalino March 2001 0. Picazo Health Outcomes? Evidence from a Oscar F. Picazo 87954 Cross-Country Analysis Albertus Voetberg Policy Research Working Paper Series Contact Title Author Date for paper WPS2566 Convexity and Sheepskin Effects Norbert R. Schady March 2001 T. Gomez in the Human Capital Earnings 32127 Function: Recent Evidence for Filipino Men