102656 Economic Internet Toolkit . for African Policy Makers . An Africa Internet Forum, UNECA and infoDev Project \ I I I I· I I I The Economic Internet Toolkit for African Policy Makers was produced for the Africa Internet Forum under the direction of Robert Schware, Senior Informatics Specialist in the Telecommunications and Informatics Division of the World Bank. The models, data and accompanying text were created by Charles Kenny; Anil Srivastava and Anat Bernstein contributed to the section on examples of Internet use on the continent; and Shampa Banerjee edited and prepared the document for publication. Our special thanks and appreciation for major contributions made to the policy and global Internet sections go to Robert Hawkins. We are very grateful for the help of Lane Smith (USAID), Inta Brikovskis (NAS), Lee McKnight (MIT), Michael Jensen (UNIDO), Michel Menou (CABECA), Nancy Hafkin (UNECA), John Mack (U.S. Department of State) and Raymond Akwule (George Mason University). The project was funded by the World Bank's infoDev program (ID 950920-10): infoDev is a grant-making program designed to encourage re- form and investment for. improving access to information technology in developingcountries. The toolkit will be available on infoDev' s website, part of the World Bank website (www.worldbank.org) .. The findings, interpretations, and conclusions expressed in the Toolkit are those of the authors and should not be attributed in any manner to the member organizations of the Africa Internet Forum or to the World Bank, its affiliated organizations, members of its Board of Executive Directors, or the countries they represent. Foreword Liberalization of the telecommunications sector is progressing across Africa. One of the most important benefits of this trend is that it will make value-added services, particu- larly Internet access, more affordable and reliable for telecommunications users in the continent. ' The Internet need not be a useful tool only for industrial societies. The poor in many African countries, struggling to meet basic needs, often remain poor not only- be- cause they are denied access to physical and human capital, but also because they lack the information necessary to best convert that capital into wealth. By opening wide the door to a huge store of global knowledge, the Internet offers untapped possibilities to address the blight of information poverty. This Toolkit is inspired by the African experi- ence where access to the World Wide Web is helping doctors to save patients, schools to educate children, and communities to create businesses that will lift them out of destitu- tion. Today there is growing exposure to the Internet in Africa. Over the last three years, the number of Internet host sites, excluding the developed market of South Africa, has risen twenty-fold. However, there are still many hurdles to a comprehensive coverage of African nations. Issues that require urgent deliberation include pricing structures, mo- · nopoly controls and licensing charges. Often these are a result of state policies restricted by a short-term view of the economy and its future, or by concerns over the immediate effect of the Internet on telephone company revenues. This Toolkit closely examines these issues. It finds that, in the long term, the Internet cannot be looked upon as a threat to telecommunications companies. It is true that it is one of a range of technological advances that are forcing changes i:ri. the opera- tion of telecommunications systems, but it also presents opportunities for new sources of revenues and new ways to meet the demands of society. The Internet has become a tool for development, with its ability to facilitate the delivery of social services, disaster mitigation, and poverty relief. The Toolkit also finds that the move toward liberalization is likely to have a beneficial effect on Internet roll out, just as it has on basic service pro- vision. The Toolkit is part of a collaborative effort on expanding Internet access in Africa that began in 1995 with the creation of the Africa Internet Forum (AIF). This is a group of donors, users and other interested organizations including the UNDP, UNITAR, USAID, CIDA, NASA, the Carnegie Corporation and the African Networking Initiative - which itself includes groups such as the IDRC, ITU, ECA and UNESCO. The Toolkit is intended to be used in policy dialogues and country assessments, broadly to facilitate the involvement of the private sector in Internet provision, and specifically to help policy makers shape their attitudes toward this exciting and expanding sector of the telecommunications business. James Bond Director, Energy, Mining and Telecommunications Department The World Bank Contents 1 Introdudion and Executive Summary / 1 2 Global Telecommunications, Emergence of the Internet, and Africa / 9 3 The Internet In Africa / 18 4 Policy Issues I 30 BIBLIOGRAPHY I 39 ANNEXES 1 Results of the Model for Costs and Revenues from Communications I 42 2 Calculation of Telecommunications Revenue Lost to Substitution Using FCC Data / 59 3 The Relationship Between Communications Liberalization and Internet Price and Penetration I 60 4 The Importance of Accounting Rates to African Telecommunications Companies / 68 5 Potential Internet Penetration in Africa/ 70 6 The Internet Cost-Benefit Model / 77 I I 1 Introduction and Executive Summary As the old telecommunications regime crumbles around the Definition: email world and a new one emerges, Africa has an unprecedented op- Electronically transmitted messages portunity to vastly improve its information technology and com- sent via the Internet. Email allows a munication infrastructure. African nations, however, must act user to compose messages and trans- mit them in a matter of seconds to one quickly to gain access and contribute to the world's knowledge or more recipients anywhere in the base, communicate with global neighbors, and fully participate world over the Internet. in the development of a global information society. The Internet represents a technology that encapsulates much of the promise Definition: World Wide Web (WWW) of this information revolution. This toolkit aims to assist Afri- A global hypertext system that uses the can policy and decision makers to better understand how the Internet as its transport mechanism. Internet is different, its costs and benefits, and policy issues that Created in 1989 at the CERN research surround this new technology. institute in Switz/er/and, the Web relies Despite the low level of telecommunications development on hypertext transfer protocol, an in the African continent, the Internet has expanded relatively Internet standard that specifies how an application can locate and acquire rapidly over the past few years. Private, nonprofit, and public resources and information stored on sector Internet service providers have sprung up to help exploit another computer on the Internet. the opportunities presented by this new technology. At the time this report was written, 42 of the 54 nations in Africa had live Definition: Host site public access to the Internet in the capital city, while eight had Any computer that can function as the countrywide local dial-up access. These were Burkina Faso, beginning and end point of data trans- Malawi, Mali, Mauritius, Morocco, Senegal, Chad, and Zimba- fers. An Internet host has a unique bwe (Jensen, 1998). Competition (where allowed) can be fierce, Internet address and a unique domain and the price for "all you can eat" web access dropped below name,such as worldbank.org. US$30 I month for some countries in the region. In Mozambique, one of the least developed nations of the continent, it is possible to make a telephone Figure 1.1 African Internet Host Sites call over the Internet today. Largely because of the efforts 140000 of private operators, the 120000 number of host sites in Afri- 100000 can countries has increased from 290 in five countries in 80000 -+-Africa Total 1995 to 6,510 in 32 countries 60000 ------- Africa minus SA in 1998 (see Annex 3). These 40000 figures exclude South Africa, which alone has 129,000 sites. 20000 However, the Internet has been growing so rapidly 1995 1996 1997 1998 worldwide (at a rate of 12 1 Economic Toolkit for African Policy Makers Definiffon:Bandwidth percent a month), that Africa's share of host sites has been fall- A networks carrying capacity. The rate ing over the last year. Africa's share of Internet host sites world- at which information can move between wide was a mere 0.025 percent in 1997, and fell to 0.022 percent computers. Bandwidth is measured in by the beginning of 1998. Excluding South Africa, the entire con- bits of data per second tinent with its population of well over 650 million has about as many Internet sites as Croatia with its population of five mil- lion.Twenty-two countries across the world, with populations of over one million, have no Internet host sites at all. Of these, 16 are African: Zaire, Chad, Somalia, Sierra Leone, Sudan, Rwanda, Malawi, Mauritania, Mali, Lesotho, Guinea, Gambia, Eritrea, Congo, the Central African Republic, and Burundi. In fact, outside South Africa, only one out of every 5000 Africans have access to the Internet. Infrastructure and development Further growth of the Internet in Africa is closely tied to the quality and availability of telecommunications infrastructure in this vast continent. A major component in this process is the lib- eralization of the sector and private sector investment. In sub- Saharan Africa, change is already underway; 25 countries have begun reform programs in telecommunications. However, how much these reforms will immediately impact the growth of the Internet is yet to be gauged. The Internet places large demands on infrastructure with its requirements of high quality and high speed connections. Ser- vice providers need cheap and reliable access to international communications lines to link with the web, as well as equally reliable local access for their customers. This need for high band- width infrastructure creates serious pressures on the less devel- oped telecommunications networks of the world. This is certainly true of most African nations, which have only two percent of the world's telephones to offer to 12 percent of the world's popu- lation. While there are very advanced networks in some African countries like Rwanda and Botswana, others, like Madagascar and Uganda have unreliable analog systems. The proportion of digital lines on the continent is 56 percent as compared to a glo- bal low income average of 90 percent. Advanced technologies such as ISDN, mobile telephony and leased lines are still not fully developed in most African countries. Despite this, Africa has a great opportunity to leapfrog such constraints by using cheaper technology such as wireless local loop, low earth orbit- ing satellites, and the ability to send data over the electricity grid. 2 Figure 1.2 Internet Access and ISP Costs· Costs for Local Internet Access Telephone Access 15% Internet Service 43% Computer 37% Modem 5% Costs of Regional Internet Access Telephone Access 72% Modem 2% Computer 12% ISP Costs Yearly Internet and Telecoms 48% Equipment 32% 3 Economic Toolkit for African Policy Makers Table 1.1 Lost Profits to International Email Cost structure and development Substitution Another, no less serious, challenge to the de- Country Lost profit/yr Telecom Lost profit as velopment of the Internet in Africa is the ex- US$m (high) Revenue/yr % of revenue isting cost structure for access to the network. US$m At the moment telephone charges represent an insignificant cost for those who have lo- South Africa 31.9 36747.0 0.87 cal access to an ISP used only for electronic Zimbabwe 1.2 146 ..2 0.80 mail. For any user accessing the World Wide Mozambique 0.4 62.6 0.65 Web, and for email users accessing the Ghana 0.6 65.0 0.89 Internet from outside a local dialling area, telephone charges become very significant. For instance, the percentage of costs for local Internet access that go toward telephone charges and ISP charges is 58 percent of the total user cost. For a user who accesses an ISP from out- side a local calling area, telephone and ISP charges skyrocket to 86 percent of total user cost (see Figure 1.2). Furthermore, In- ternet Service Providers (ISPs) pay a large percentage of their costs for connectivity: 48 percent of ISP costs are accounted for by Internet backbone connection and international leased line costs. While the Internet is threatened by inefficient telecommu- nications infrastructure, unreformed African telecommunication companies in turn feel threatened by the impact that the Internet might have on their revenue stream. In fact, while the Internet is likely to divert traffic from high revenue-generating international voice Figure 1.3 Comparison of Forces on PTT Revenue communications and adversely affect the profits of African telephone companies over the short term, the relative impact of the Internet on revenue is not signifi- cant. In markets that have not yet seen any reform, and where there is greater dependence on overpriced international call charges, the Internet is likely to be utilized to bypass these costs. Estimates of the resulting losses are presented in Table 1.1, based on a low-level use of about one email in and out a day. How- ever, the Internet is only one of a num- ber of forces that will have an impact on telecommunications companies in the re- gion. There are many others,such as in- ternational pressures on regulating accounting rates charges, and the grow- South Africa Zimbabwe Mozambique Ghana Ethiopia ing presence of callback technology. Even countries with advanced Internet provi- a Lost profit as % of revenue El US settlement as % of revenue sion are probably experiencing only 4 about one percent of revenue reduction directly resulting from "Internet access is more essential for email substitution. U.S. accounting rates settlement payments Africa than for any other region," to African countries are significantly larger - 2.4 percent of argues Ernest Wilson, Director of the Mozambique's revenues, for instance, or 14.3 percent of Ghana's Center for International Development and Conflict (see Annex 4 and Figure 1.3). Furthermore, while in the longer Management at the University of term, Internet telephony probably represents the larger Inter- Maryland, ''because of the net-based threat to telecommunications company revenue, at the deterioration of African libraries and moment the level of capacity in most African countries is insuf- universities which lack current ficient to support this technology. journals, periodicals and books. To be genuinely competitive in the global marketplace, Af- Internet access could help keep rican telecommunications companies need to rapidly integrate scholars and students connected the changes that are reshaping telecommunications across the with current developments in their world, and transforming it into a commodity business. By em- field" bracing Internet technology and expanding the number of users who can access the network, these companies have a consider- able amount more to gain in the long run. At the same time, the growth of the Internet offers new opportunities for businesses and communities on the continent. The Internet is good business I 1 \ Models presented in this toolkit suggest that expansion of the I Internet in Africa cah provide opportunities for significant re- duction in the communications cost of a wide range of African telecommunications users. These savings would come almost exclusively from international calls. There are two reasons for this conclusion. First, at the moment, 80 to 90 percent of email are sent to and received from outside of the continent. Second, international calls from and to the continent attract revenues far in excess of the costs of completing the call. Ironically, given the pressures on rate rebalancing which are unconnected with the Internet, it is likely that emailing will become relatively less attractive as a substitute for voice or fax as international call costs drop and local call costs increase. At the moment, however, Internet access is a profitable investment, purely from the point of view of direct savings on communications for companies that do a lot of international business. Talking to the United Kingdom for about an hour each business day over the period of a year would cost a Mozambican businessman approximately US$38,250. Faxing that same information would cost US$7,650. All of the yearly costs of a regional Internet connection - a computer, a modem, and Internet access - used for international email alone as a substitute for fax traffic, would together amount to US$1,328. The yearly savings over fax use would thus be US$6,322. 5 Economic Toolkit for African Policy Makers Benefits for the commu1i1ity Figure 1.4 Telecoms Liberalization and the Costs of Internet Access The Internet is far more than just a cost saving 250 substitute for voice and fax communication. Al- ready in Africa, the technology is being used y;; CJ) 200 for a wide range of other applications with a :::> ..._, direct influence on the quality of life. Crafts- ....... 150 men in Uganda, Botswana, and Senegal are Ul 0 (,) marketing their products worldwide through >. 100 the UN International Trade Center's Virtual .c ....... c Handcraft Exhibition Center; national newspa- 0 50 2 pers in Cote d'Ivoire, Kenya, and Zambia pub- 0 lish daily on the web; doctors share diagnostic 1 2 3 4 5 data; and more than 400 students in six African countries are participating in the World Bank Liberalization Index World Links for Development program aimed r:i 1 o pages/day • 50 pages/day at connecting schools around the world to the Internet for collaborative distance learning. Spreading the benefits of Internet access If these benefits are to be enjoyed by the many rather than the few, it is vital that costs of Internet access are reduced. As we have indicated earlier, the major costs of Internet access for end users is ISP and Telecommunication charges. The model of Internet costs and benefits suggests that there are economies of scale in Internet service provision. However, Sta- tistical analysis points to a stronger influence of liberalization in general, and ISP competition in particular. Figure 1.4 lists av- erage monthly costs of two levels of Internet access in African countries sorted by a measurement of liberalization of the tele- communications sector in the region (see Annex 3 for details). In the figure, 1 represents a low level of private activity in the sector, while 5 indicates a very high level. Clearly, more liberal markets enjoy cheaper access to the Internet. Table 1.2 shows that countries with ISP competition Table 1.2 African Internet Users and Liberalization also have more users, more interna- tional bandwidth, and more users in Users International Users/ proportion to their population than bandwidth population(/m) countries without competition. The importance of cost is further Average 1681 336 188 illustrated by a model of the potential number of Internet users in the region. ISP competition 2643 553 246 Monopoly ISP 842 112 190 This model suggests that a country like Tanzania could increase its potential Source: Michael Jensen, http://demiurge.wn.apc.org:BO/africa/users.htm user base seven times over by reduc- ing the cost of accessing the Internet All statistics exclude South Africa (equipment, phone charges and ISP 6 costs) to South African levels. This increase, however, depends on these sites being accessible to a paying public. While the private sector will likely compete for users in the urban areas, there will be an increased role for government to insure access to those users in rural areas. Unless Internet sites proliferate in community centers, libraries, schools, and telecenters, access will be restricted to a small urban elite. While financial considerations are clearly central, results from a survey of Internet usage in Africa point to other impor- tant factors in spreading the technology to a larger number of people. First, usage was severely restricted by a lack of knowl- edge and adequate training. Further, even those with the skill to access the technology were discouraged by the paucity of rel- evant content and the unreliability of the network connections. Policy conclusions The Internet is likely to continue to revolutionize the means in which people communicate and access information. Because the Internet represents such a- powerful new communication tool, the environment in which it operates must be regulated differ- ently from traditional information and communication media. The toolkit stresses three general principles: the importance of not trying to fit the Internet into existing regulatory structures, the power of competition on Internet growth, and the necessity of allowing the Internet to flourish without the burden of un- necessary regulation. The models and data presented in the toolkit suggest a num- ber of policy conclusions. To expand access and use of the Inter- net in Africa, it is necessary to provide the following: • Low cost and reliable access to international band- width. • Low cost and reliable local bandwidth connectivity. • Countrywide reliable local cost access to ISPs. • Low cost access to network equipment. • Widespread public access to networked computers. • An educated and trained user and provider base . . • Support for the development of national and Afri- can Internet content. The policies that may help to meet these needs are: • Liberalization of the telecommunications network. • Liberalization of Internet service provision. • Lowering of tariffs on computer and telecommuni- cations equipment. • General tariff rebalancing with possible support for local cost ISP access. • Support for community access to the Internet. • Support for training in the use of the Internet. 7 Economic Toolkit for African Policy Makers • Support for local content development. • An Africa-wide backbone. The Internet represents a new opportunity for increased knowledge and information for development. In the long term, it will significantly alter the structure of telecommunications. In the short term, the main policy conclusions from this toolkit are not very different from those of studies that concentrate on telecommunications in general. The Internet is heavily depen- dent on an efficient telecommunications sector, which is usually market-based. The Internet is itself another pressure acting to- ward the market option of further competition and liberaliza- tion. However, while the market forces will expand access to a section of the population, there will still remain a role for gov- ernment to reach the poorest. 8 2 Global Telecommunications, Emergence of the Internet, and Africa Not long ago, there were less than 200 telephone companies in the world. Now there are perhaps 5,000 companies, and soon there might be 100,000. Technological advances across the world are making it difficult for monopolies to survive. This changing environment is having diverse effects on the international tele- communications system - not least on the costs and prices of international phone calls (see box). International calls account for a large percentage of Afri- can post and telecommunications (PTT) profits, and these prof- its have been artificially maintained by agreements between governments and the big national telephone companies. Chang- ing technology and global policy reforms are now threatening this arrangement, most visibly though the combined pressures of accounting rates reform and call back services. This is only the beginning, for there are other factors waiting to transform the old telecommunications paradigm. Technological advances have also ushered in changes in the economics of local supply (see box). Wireless local loop tech- nology, for example, eliminates the need to run cables to every home, by operating a short-distance wireless link that greatly Pfrss~fes:o~ftntir~ati&"J~i.calz·P~i;;~ .<:' , '• , '1:1,titomati.s~HYi':'a · fo.. ., .~line afW8f:~tes ...· the1~~Xce }.;i..a ·countryfs tar: 01ft,t:.8~1i~e.~itl1tJg 9}flt&es · · . •coJ~~,i~~ii~W~;~<;~~J~~,f~~i}\:~~~i~i:~1U!~~::. ; ·•·: · fiica'rJ>ptrs; ·· VallleY(V; ,~; ;·:y,, ;- '. ;;,;:=~~;~ ;.;-·, ,, 9 Economic Toolkit for African Policy Makers Table 2.1 Telecommunications Liberalization in Sub-Saharan Africa, Late 1997 Liberal Separate Independent New Sector Private Tel co Competition Equipment Posts and Regulator Law Cellular Privatized in Basic Trade Telecoms Service .Angola x x n.a. Benin x Botswana x x x 97 98 Burkina Faso x 98 Burundi x x 97 x Cameroon x x 97 Cape Verde x x x x CAR x x Chad x 97 Comoros x x Congo-Brazzaville x x 97 x x 97 Congo-Kinshasa x x Cote d'Ivoire x x x x x x 2004 Djibouti x n.a. Equ. Guinea n.a. n.a. Eritrea x n.a. Ethiopia x x x Gabon x 98 97 98 98 Gambia x x x Ghana x x x x x x x Guinea x x x x x Guinea Bissau x x x x Kenya x 97 97 97 97 98 Lesotho x x x Liberia n.a. x n.a. Madagascar x x x x Malawi x x Mali x x x n.a. Mauritania x Mauritius x x x x x Mayotte Mozambique x x x Namibia x x x x Niger n.a. n.a. n.a. n.a. n.a. n.a. n.a. Nigeria x x x Rwanda x x S. T & Principe x x n.a. x Senegal x x x x 98 97 Seychelles x x x n.a. x x x Sierra Leone x x n.a. Somalia n.a. n.a. n.a. n.a. n.a. n.a. n.a. South Africa x x x x x 2002 Sudan x x x x x Swaziland x Tanzania x x x x x 98 Partial Togo x x 97 97 98 99 98 Uganda x 97 97 97 x 97 97 Zambia x x x x Zimbabwe x X = reforms already carried out. Numbers =years reforms start. 10 ·;The~i:J;rei:lint/~'~kY bf¢8m"vi«~fr:qti. ',C.. ::.· Rik'~Joptics, fii~'~;;in,trod~c~:~lfJ;~~e . s/ ~~ie~·:;g;~~'- :?nd ailow for:s~~~;;~:~~i~es sue . . ca11~~1lb,:and effi- 'ejcrfosi'On in,,!l;r,e ca,.pa~ify, OJ ba~9.~t9..th£8f tran~mission ·. ciency gains through "intelligent networking". §Y§tems1,AJilj~r tRJnnert1lfl!;l hi.ih\atj··pajJ.qm:(:arty 1llany Wireless has become even more affordable through thousand~e~·teik!?P:~we cort:\?etsations/s'&'~l{~f.':!he:co~tp~r. the introduetion oJ ne"?",techfaologi~s for using the radio · · \TOice'brcuif:,pe<:;oine~·.in.fiJ]itesimal. A milJtt.t~bn aJta~s,~ spectrum, induding Time Division Muttiple Access:··. ·aHai1tic cap~~ l.f~d 40Y~.~!~ ago cost a hefty US$2A4; bl.\!i.•; ~sm~A~~~~7:~1i~~~~1~~s~~i~~[~e[i~~~~i~wr~~~1~~,.· that the basic. Ccisfof •a internationaL:link:iS,.n · ·· ·vial. ;,/: · · "1.ms sucli ~s iD.tegrated circ · · and infriia ...... ·zatiotL itch-········ . . ·, pute.r~'p~c?Ine'faster,,t •. These cables. can now.be·cormect .. · ..• fogc61-T1EBn~~tstha~,~·?st. a';fi:~,fJ.t\i'· . 1 ;ceof~~&1V~DJeveLdf .. ,. •,e:2WA)!ns power ~~,~-rease~;;~.0·; esame. =4 3.29 2.57 1750 406 0.15 21 266 Liberal lndex<4 2.18 2.06 1538 327 0.38 8 225 Private Index= 0-1 1.53 1.40 717 95 0.46 6 217 Private Index= 2-4 3.75 3.13 2494 614 0.13 22 269 Source: Michael Jensen, http://demiurge.wn.apc.org:80/africa/users.htm All average statistics exclude South Africa. 29 Economic Toolkit for African Policy Makers 4 Policy Issues By changing the way connections are utilized, the Internet chal- lenges present profit centers and network configurations. In particular, Internet telephony and substitution threaten profits from international voice communication. However, there are at least three arguments that suggest this is not reason enough for PTTs to control or discourage Internet expansion: • The Internet is only one of a number of technology- driven forces that are initiating changes in the structure of the network and pricing. Discouraging its expansion on the continent cannot preserve the status quo. • The Internet offers potential new revenue sources - increased local call revenue and greater demand for international bandwidth, for example. Evidence suggests that an increased user base will generate profits for the PTT. • There is more than enough evidence to demonstrate the value of the Internet and its ability to encourage economic growth in Africa. Fear of change in tele- communications companies is not reason enough to restrict the opportunities that the Internet presents. Even if official attitudes toward the Internet change for the better across Africa, there remain policy issues. One of the ma- jor ones, already discussed, deals with rate rebalancing which may lead to an increase in the cost ot local calls and connection to the network. This presents a challenge to policies encourag- ing universal service or at least universal access to telephones, and may also restrict Internet usage. To meet this challenge, and others involving costs and access, the toolkit offers the two- pronged strategy of liberalization and competition coupled with public access. There are other issues that may affect decision making in this sector. The toolkit offers some general principles for gov- ernment policy: • Think differently and stress openness: governments need to treat the Internet as a new technology and not a mere addition to the existing telecommunica- tions or broadcasting regimes. Policies and regula- tion need to be open and flexible, to adjust to new environments. 30 • Promote competition: governments should work to encourage competition at all levels, a tested tool for increasing efficiency and lowering costs. • Avoid unnecessary regulation: governments need to work to avoid unnecessary interference with the Internet's development. The Internet has grown so rapidly primarily because it has been allowed the freedom to do so. Think differently It is important that government regulatory policies encourage innovation and new service applications, and to do so, very of- ten those policies have to undergo major redesign. The existing definitions of broadcasting, publishing, and communications are changing all over the world. Government policy must be sensi- tive to these changes, and realize that the Internet landscape a few years in the future may look very different from that of to- day. Also, convergence is changing the way that voice, data, and video content is delivered. Opportunities to create and deliver this information will come from many sources, both domestic and international. The regulatory infrastructure must be sensi- tive to this and create conditions to allow many new entrants to participate in and contribute to the development of a new infor- mation society. Promote competition Competition in telecommunications network services, Internet service provision, and content development should be stressed in order to bring the greatest efficiency gains to the country. Wherever possible, market forces should be harnessed to take the place of direct regulatory intervention. Although new ser- vices like Internet telephony and streaming audio and video may create regulatory problems, these developments are positive ones which governments should encourage. NewCqmmunjcat,ions.LaW'i~~i;~~a~W,e:;',Exampl.ep£,'f!n ·,;~:;;i~~rl.]Joli91{' · Media worke~·s andci~il s.ociety r~presentatives iAZim~ .... ti9p9 :e~Ji~~s);;itl-t!h§ e~~eptich{;~~:~~Ii~Iarphdneser\Tices,. babwe have, call~d on' ,the government t,o aine,nd 'forasuccessoi- cofn:par1yofthe existingP.osts;and J'elecom- Zimbabwe' s draft comm.unicatior1s. pm in ord§r}O,• al- municatidps qirporatiori .. All s.J}q.:re§ ig.the·:successoryQfn:- • l()w for.an independenttelecoll}111un~cati.ons (lritit.tite.·of' . · have in its :possession or control; ·a broadcast station:.: 'Soutl:iep1.Africa ...:.,...MlSA.C2'/.25/9.8\ The ·d~~-~~-~,b'Pl ·'.~!Sci"i;et.~i;1:~~,~monop?~~·i1:1,,·t~~:P,~:¥!R~~p~~~-, ,-·;·:'\:=>L~c., .;~:;;~;<= :; ,· 31 Economic Toolkit for African Policy Makers Avoid unnecessary regulation Governments should regulate the Internet as little as possible. The ultimate objective should be less regulation for all, rather than more regulation for some. For instance, if competitive im- balance exists because a new technology is not subject to the same regulatory constraints as a competing older technology, the answer should be reduced regulation of the older technol- ogy. However, positive contributions to network development by governments are important, especially in the area of univer- sal service, international cooperation, content creation, training and development of education and health networks. The most common policy issues dealing with the Internet today have been categorized in five separate sections below: costs and liberalization, infrastructure, management, content, and training. The sections identify the questions raised by the chang- ing telecommunications environment, questions that the toolkit attempts to answer. Costs and liberalization Internet as a liberalized value-added service Questions to be raised and debated about the basic nature of services offered by the Internet include: • Should the Internet be considered a part of telecom- munications network services or value-added services? • Should the Internet fall into the existing regulations for telecommunications or for broadcasting? • What are the benefits of liberalizing the Internet in my country? . • Should organizations be allowed to run private Internet networks? Although some African countries have already liberalized the Internet as a value-added service, many others have not. The toolkit clearly indicates that the Internet should be fully liber- alized as a value-added service and the government should en- courage the growth of the Internet by allowing private sector initiatives to flourish. The potential benefits of this liberaliza- tion are many. It will lower costs to consumers, generate more communication access points for the country, provide opportu- nities to create new services and job opportunities, and moti- vate entrepreneurs to construct communications networks. International gateways The next set of questions are concerned with restrictions that 32 relate to international Internet gateways: • If the Internet is liberalized, should there be any restrictions on the number of international gate- ways licensed? • Do many different international gateways present inefficiencies in traffic sharing within a country? • With a number of cable and satellite ventures offering expanded bandwidth, should participation in such a system be a matter of government policy or should decisions be based on the business needs of the operators? Policies vary widely across the African continent. In Ghana, ISPs have access to three gateways. There is no official limit to gateways in Mozambique or Uganda (although this is under consideration), but South Africa and Zimbabwe have monopoly providers. A number of countries have allowed companies and organizations to bypass international gateways and construct their own private networks. The toolkit has outlined that al- though there are economies of scale in Internet service provi- sion, the efficiency gains from competition outweigh the gains from economies of scale. The continued increase in international bandwidth suggests that costs and prices will continue to fall. Governments, therefore, should not restrict access to this sup- ply. Licenses Licenses and associated costs for operating an Internet service are valid concerns for any government considering a liberal en- vironment for the Internet. Questions include: • Should governments require licenses for Internet service provision? Policies Toward International • What is the primary purpose for a license - to Gateways: the Case of Ghana generate revenue for the government for expanding Internet connectivity to underserved areas, offset Network Computer Systems (NCS) in lost government revenue from international com- Ghana was allowed a special license to munication, control the access points for informa- bypass Ghana Telecom's infrastructure tion, or ensure the quality of Internet provision? and operate a 2Mb Satellite link to • How should governments license VSAT services MAE-EAST in Virginia. Another gateway and public satellite spectrum? provider is the Africa Communications Licenses provide a government with an important tool to Group, the consortium that won the bid stay on top of developments and demands. The licensing proce- for the second network operating license.A range of other groups are now dure, however, should be as simple as possible, and should not applying for VSAT links including be high priced as these costs will ultimately be passed on to the Ashanti Goldfields, Barclays Bank and consumer. If the telecommunications sector is interested in the Shell. As part of the application, they development of the Internet, the revenue from the licenses have to provide information on the should be used for this very purpose. In Africa, countries such company, its business plans and as Kenya and Uganda display two opposite ranges of such costs. feasibility plans. 33 Economic Toolkit for African Policy Makers Kenya has a fixed license cost of US$10,000, as well as an addi- tional US$5,000 and one percent of gross turnover as yearly costs. Uganda, on the other hand, has only a single fixed cost of US$1,000. Costs of inputs. Concerns about import duties and taxes give rise to the follow- ing questions: • What is the appropriate tariff level for computers and networking equipment in Africa? • What discounts should be given to educational institutions? • What discounts should be given to other sectors of the economy? The model suggests that net density is strongly correlated to the cost of Internet inputs such as computers and networking equipment. In order to expand Internet access in a country, gov- ernments should consider reducing tariffs on computer equip- ment. For instance, a reduction from 45 percent to zero on computer tariffs in Mozambique would decrease the cost of In- ternet access by more than 11 percent. Table 4.1 describes the current tariff level of a number of countries in Africa. Some of the zeros in the table may indicate that although there is officially no import tax, there is a monopoly or quotas on the provision of equipment, so that governments control imports and prices more directly. Infrastructure Table 4.1 Telephone Equipment and Computer Import Duties and Taxes(%) Questions associated with government's role in Telephone Equipment Computers developing Internet infrastructure are listed be- low. Import Sales Import Sales • What are the consequences of allowing duty tax duty tax service providers to set up their own infra- structure? Ghana 10 7.5 10 17.5 • If there is a monopoly provision of infra- South Africa 0 5 Ethiopia 0 15 2 structure and some form of competition in Mozambique 35 45 other services, must the suppliers of the Congo 15 3 15 3 competitive services use the monopoly's Kenya 30 18 15 15 infrastructure, or can they supply their own? Nigeria 15 15 • If there is a monopoly provision of the local Senegal 61 26 loop, should the use of radio technology for Zambia 20 20 Internet in the local loop be considered part Zimbabwe 0 20 of the monopoly, or just one of the alterna- Uganda 30 17 tives to the local loop? • What are the rules under which private Source: World Trade Organization networks can operate? 34 Universal access Universal Service in Developed Countries African countries have been far from successful in achieving Developed countries have been dealing universal service, and despite efficiency gai:r:is and falling costs, with the· concept of universal service there is little hope of their reaching this goal in the near future. diversely. In Canada, the Telecommunications Act requires that This emphasizes the importance of setting a more realistic goal reliable and affordable of universal access. Here are some of the concerns linked with telecommunication services be universal access to the Internet as opposed to universal service. accessible in all regions of of the • What is the definition of universal access? country, which allows flexibility to revise • How does Internet affect opportunities for univer- universal service obligations in light of sal access as interim measure to reaching the goal of changing technologies and services. In universal service? Denmark, the definition is more • Who should be made responsible for telecenter specific, citing in particular telephony development and how should it be financed? services, ISDN, and leased lines with a • How can telecenters be made sustainable? maximum of 2 Mbit capacity. The concept of universal access for the Internet stipulates that all citizens should be within a certain distance of a communication facility. Many countries have begun to encour- Telecenters in South Africa The Universal Service Agency is a age the development of infrastructure called telecenters to meet statutory body established by the this goal. A telecenter is a community access structure where Telecommunications Act of 1996. The the fusion of telecommunications, information, multimedia and job of the agency is to promote computing functions help address a variety of community prob- universal access to telecommunications lems and needs. Most telecenters charge the community a nomi- for all in South Africa. The Agency is nal fee for use of the facility. An example of a telecenter project working with a number of different is the one being undertaken by the Universal Service Agency in organizations such as schools, libraries, South Africa. churches, and existing community organizations to establish the PTT competition telecenters. In the first year of the project, 80 are being planned and an additional 300 are forecast for the The success of the Internet is closely linked with competition in following year. the provision of service. The questions that follow have direct significance for African PTTs. • What is the impact of local telecommunication service competition on Internet growth? • How can competition be encouraged? • What are the negative consequences of opening up to competition too quickly? The toolkit demonstrates that competition in telecommu- nications services lowers the costs of Internet access to end users. This suggests that competition should be encouraged as a mecha- nism to reduce costs, provide technical innovation, improve quality of service to customers, and provide a greater variety of services. Local competition is likely to be a key policy instru- ment used by governments to expand affordable access to the Internet. 35 Economic Toolkit for African Policy Makers Internet telephony Full duplex telephone to telephone connections are available over the Internet today, but the regulatory questions about this have not yet been answered in all countries. Here are some con- cerns that may help to ensure best practice. • How should Internet telephony be defined? • Should this service be considered voice traffic or data traffic? • What is government's role in assuring quality of service? Countries around the world have yet to determine how to, or whether to regulate Internet telephony. Many countries have existing voice monopolies and as such would view Internet te- lephony as a breach of this monopoly. Others have taken no position, either because Internet telephony was not foreseen in current legislation or because of their already liberal markets. International cooperation The international presence of the Internet and its ability to en- hance global solidarity and cooperation make the following ques- tions pertinent for African countries. • What role should governments play in developing regional cooperation and network extension across countries in Africa? • What are the advantages a nation might find in synergy, greater participation in regional econo- mies, improved personal and business communica- tions, and expanded trade? • What are the opportunities for regional standards structures to develop? • How much could a regional backbone reduce costs to the end users? One of the most important regional issues in which gov- ernments need to work together is the development of a regional Internet backbone. At the moment a majority of African ISPs must connect through a high bandwidth connection to the In- ternet backbone in the United States as there is no Internet back- bone in Africa. In essence, Africans are paying United States ISPs so that the world is able to access Africans. Substantial sav- ings and efficiency gains are possible if the African continent develops its own Internet backbone. Governments of African na- tions, therefore, should consider creating, with international co- operation, their own African Internet backbone. 36 Management The Internet is a decentralized and distributed network of net- works and thus inherently difficult, if not impossible, to con- trol. There is no single organization that manages the Internet; rather, many individual networks cooperate to guarantee its success. The issue of domain name registration, however, is an area of Internet management in which governments might need to intervene. Domain name registration A controversial policy question in Internet management is the control of domain names - the brand names of the Internet. In the U.S., Network Solutions Inc. (NSI) currently charges US$50 per year to register a domain name. A portion of this money goes to NSI to recover administrative costs and a portion goes into an "Internet Intellectual Infrastructure fund". But the existing registration process for generic top-level domains has generated substantial controversy. Some parties have objected to what they consider to be NSis monopoly control over a valu- able resource, especially since an entity in the U.S. is respon- sible for assigning addresses with international ramifications. The International Ad Hoc Committee, comprising representa- tives from the Internet Society, International Telecommunications Union, the World Intellectual Property Organization and other groups, has issued a wide-ranging proposal to restructure the generic top-level domain name system. In all countries outside of the U.S., the root domain has a country designation at the end, and one entity in the country is charged with distributing top level domain names. In some coun- tries other bodies are charged with assigning lower level do- main names such as .co for a private sector company, .school for an educational institution, and .org for an organization. Sub- stantial debate has revolved around control of the top level do- mains in a number of African countries. The controversy centers around the question of whether domain name provision should be handled by a private entity or a government entity. For in- stance, Malawi and Zimbabwe have recently debated this issue. Content Content on the Internet and government's role in this area are major issues that cannot be covered adequately in this section of the toolkit. J?riefly, however, here are the major issues: • What is government's role in legislating content on the Internet? • What is government's role in protecting intellectual 37 Economic Toolkit for African Policy Makers property rights? Are databases considered copy- right material? What about software programs? • Are the legal rationales for regulation of content in other media, such as scarcity of transmission capac- ity and invasiveness, applicable to the Internet? All countries will have different laws governing issues such as acceptable content, intellectual property, and privacy. One area, however, in which government should focus its attention is the creation of local content. With an increase in relevant local content, it can be hoped that more users will use the Internet and the number of individuals contributing to its growth will increase. Governments can play a significant role in assisting with the computerization of government and public informa- tion. Also, the preservation of cultural heritage is an important area in which the Internet can be deployed to computerize, and thus share traditions that otherwise would be lost, forgotten, or at best retained by only a limited audience. Training and human resource development Training is one of the most important areas of investment for the future development of the Internet in Africa; it is also a press- ing need that is not being adequately addressed. Here are some of the issues that need to be considered. • What should be the government's role in the pro- motion of telecommunications training institutes and R&D centres? • On which aspects of training should government focus its resources? African governments need to invest in raising awareness of the benefits of the Internet across a broad range of sectors , including education, trade, agriculture, and health. In addition, training for end users in how to access and efficiently use the Internet, for system administrators in managing networks, and for content creators in development of data bases and web pages for accessing local content, all need to be emphasized. In conclusion For all of the policy areas listed above, the solutions will be different for each country. The toolkit's aim has been to provide a broad overview of the issues and work through some of the areas that need to be given extra consideration. In all cases local models for local needs will develop to address each country's specific and unique reality. But the Internet continues to expand globally, Africa must take advantage of the opportunities offered by the unique properties of this network of networks. 38 Bibliography The CABECA study liberally referred to throughout the report has been documented in the following papers: Asaba, K. J. F.; Bazirake Bamuhiiga, B. "Connectivity in Africa: use, benefits and constraints of electronic commu- nication - Uganda Phase 1 and 2" (final version). Study carried out under the CABECA project of Padis as part of the IDRC sponsored research program on the Impact of information on development. Addis Ababa, UNECA/ PADIS, April 1998. Chifwepa, V. "Connectivity in Africa: use, benefits and con- straints of electronic communication - Zambia Phase 1 and 2" (final version). Study carried out under the CABECA project of Padis as part of the IDRC sponsored research program on the Impact of information on devel- opment. Addis Ababa, UNECA/PADIS, April 1998. Diop, 0. La connectivite en Afrique: "Utilisation, avantages et contraintes des communications electroniques - Senegal Phase 1, 2" (version finale). Une etude realisee dans le cadre du projet CABECA de Padis sous l'egide du programme de recherche Impact de !'information sur le developpement finance par le CRDI Addis Ababa, UNECA/PADIS, Avril 1998. Menou, M.J. "Connectivity in Africa: use, benefits and con- straints of electronic communication - Synthesis Report, Part 1 and 2." Study carried out under the CABECA project of Padis as part of the IDRC sponsored research program on the Impact of information on development. Addis Ababa, UNECA/PADIS, May 1998. (In preparation). Rorissa, A. "Connectivity in Africa: use, benefits and constraints of electronic communication - Ethiopia Phase 1 and 2" (fi- nal version). Study carried out under the CABECA project of Padis as part of the IDRC sponsored research program on the Impact of information on development. Addis Ababa, UNECA/PADIS, April 1998. 39 Economic Toolkit for African Policy Makers Rorissa, A. "Connectivity in Africa: use, benefits and constraints of electronic communication - Results Obtained from a questionnaires survey of participants in the April 1995 Re- gional Symposium on Telematics for Development in Africa" (final version). Study carried out under the CABECA project of Padis as part of the IDRC sponsored research program on the Impact of information on development. Addis Ababa, UNECA/PADIS, April1998. Other documents, Internet sites and papers used in the re- port are listed below: Africa Internet Connectivity website (1998) http:// demi urge. wn.apc.org:80 I africa/ AfricaNET website http: I I www.fas.harvard.edu/-kbgandy / africanetlO.html Antonelli, C. (1993), Investment, Productivity Growth and Key- Technologies: The Case of Advanced Telecommunications, The Manchester School, LXI, 4 (December) 386-397. Antonelli, C. (1996) The Network of Networks: Localized Tech- nological Change in Telecommunications and Productivity Growth, Information Economics and Policy, 8, 317-335. Aschauer, D. (1989) Is Public Expenditure Productive? Journal of Monetary Economics, XXIII, 177-200. Blake, P. (1997) King of the Callbacks, Telephony, May 26. Bond, J. (1997), The Drivers of the Information Revolution - Cost Computing Power, and Convergence in The World Bank, The Infor- mation Revolution and the Future of Telecommunications. Canning, D. (1997a) A Database of World Infrastructure Stocks 1950-1995, mimeo, Harvard Institute for International Development. Canning, D. (1997b) Does Infrastructure Cause Economic Growth? International Evidence for Infrastructure Bottlenecks, mimeo, Harvard Institute for International Development. Canning, D. and Fay M. (1993) "The Effect of Transportation Net- works on economic Growth," Discussion Paper Series, Columbia University Department of Economics. Easterly, W. and Levine, R. (1996) Africa's Growth Tragedy: Poli- cies and Ethnic Divisions, mimeo, World Bank. Easterly, W. and Rebelo, S. (1993) "Fiscal Policy and Economic Growth: an Empirical Investigation journal of Monetary Economics, XXXII, 417-57. Federal Communications Commission (1997) Report and Order In the Matter of International Settlement Rates, IB Docket No. 96-261, August 18. Gebhardt, G. (1997) Proposed ISP Access Fees from an Economic Perspective Appendix to Reply comments to FCC Notice of Enquiry CC Docket 96-263. Gilder, G. (1994) "The bandwidth Tidal wave", Forbes ASAP, Dec. 5 Hawkins, R. (1994) Computer Networking with Mozambique and its Relationship to the Bank's Strategic Objectives, World Bank, mimeo. Hawkins, R. (1998) The Internet in Uganda, mimeo, World Bank, Washington. 40 Hirsch, A. (1998) Computer Training and Internet Access Issues in Wa, Upper West Province, Ghana, mimeo, World Bank, Washing- ton. International Telecommunications Union (1996) African Telecom- munication Indicators, ITU, Geneva. International Telecommunications Union (1997) Challenges to the Network: Telecoms and the Internet, ITU, Geneva. International Telecommunications Union (1997) World Telecom- munication Development Report 1996/7, ITU, Geneva. Jensen, M. (1995) Telematics for Development: Discussion Paper http//www.sas.upenn.edu/African_Studies/Padis/ telematics_Jensen.html Jensen, M. (1996) Bridging the gap in Internet Development in Africa http: I /www.idrc.ca/ acacia/ studies/ir-gaps2.htm#3.0 Jensen, M. (1997) Internet Connectivity for Africa: A UNIDO Study, Conference of African Ministers of Industry, Accra, Ghana, 1997. Network Wizards: http:/ /www.nw.com/zone/WWW /report.html O'Neill, J. (1998) Africa: The New Telecoms Frontier in Interna- tional Telecommunications Review. Euromoney, London. OECD (1996) Information Infrastructure Convergence and Pric- ing: The Internet, http:/ /www.oecd.org/dsti/gd_docs/96_xxe.html Operations Evaluation Department (World Bank) (1992) 1992 Evaluation Results, The World Bank, Washington DC. Quaynor, N., Tevie, W. and Buley, A. (1998) The Expansion of Internet Backbone in Ghana, mimeo, Network Computer Systems, Ghana. Saunders, R. Warford, J. and Wellenius, B. (1998) Telecommuni- cations and Economic Development, Johns Hopkins, Baltimore. United Nations Economic Commision for Africa Economic and Social Council (1996) Building Africa's Information Highway: The Case of Mozambique. Seventh Meeting of the Technical Thirty-First Session of the Commission, Addis Ababa, 30 April to 3 May. USAID (1997) Model for Tariff Structure of an International In- ternet Gateway for Mozambique (spreadsheet and mimeo). Wellenius, B. and others (1993) Telecommunications: World Bank Experience and Strategy, World Bank Discussion Paper 19. World Bank (1996) Africa Regional Office, Internet Toolkit for Task Managers, Mimeo, March. 41 Economic Toolkit for African Policy Makers Annex One: Results of the Model for Costs and Revenues from Communications The Internet is likely to have a variety of effects - on users, ISPs, and telecommunications companies in Africa. Mere access to email would make an enormous difference to traditional data transmission and communication methods. In three minutes, a voice call can "transmit" two pages of information; a fax can transmit 10 pages of information; while, in the same time, a 28 k/ s modem is able to send an email of between 175 and 350 pages. At such rapid speeds, email is a very attractive alternative to traditional modes of communication. However, it has high fixed costs: users require a computer and a modem in addition to a telephone, and there are costs of leasing, access, personnel and equipment of the ISP which are passed on to the user through Internet tariffs. All the same, once the equipment has been pur- chased and an email account paid for, the price of sending a single page across the country or around the world is the price of a few seconds long local call to the users' ISP. To send that page by fax may cost up to US$30 in Sub-Saharan Africa. In response to these opportunities, users are likely to com- municate more than they have done so far. They will email messages that they would not have bothered with if limited to voice or fax communication. But they will also substitute email for voice and fax communication, so reducing PTT revenues. Of course, this exaggerates the scale of the problem for tele- communications companies. First, home telecommunications companies have to pay the foreign company that provides service to the recipient an ac- counting rate fee for completing the international fax connec- tion, so that gross revenues are higher than net revenues for international calls. Second, the telecommunications company will gain rev- enues from the ISP, which has to lease lines and equipment to connect to the Internet backbone in the U.S., along with tele- phone lines to connect to customers. Further, email substitution will reduce pressure on international connections and regional trunks, reducing the need to invest in new trunk lines. How- ever, as this need is chronic, and the effect of email on demand is unlikely to be that significant, this should provide little com- fort to the region's telecommunications companies. Third, this is likely to be a temporary problem, and one that 42 will lose its importance with the general increase in international traffic (outgoing traffic from Sub-Saharan Africa has increased from 322,000 minutes in 1990, to 449,000 minutes in 1994), and the changes underway due to political and technological pres- sures on the accounting rate regime. Despite these ameliorating factors, however, email substi- tution will reduce African telecommunications company profits from outgoing calls, at least in the short term. Also, the compa- nies will lose revenue they presently receive in the form of ac- counting rate charges for incoming calls when foreign customers email African countries instead of calling. These accounting rates are far above the real costs. For example, Cameroon's account- ing rate with the U.S. is US$1.80 per minute, so that for every minute of incoming calls from the U.S., Cameroon's telecom- munications company receives 90 cents (one half of the account- ing rate, called the settlement rate). This compares to a local call rate of under three cents per minute. African telecommunica- tions companies, that cross-subsidize the cost of phone installa- tion and local calls using regional and foreign phone charge revenue, will find that an increasing number of users are using telephone lines mainly to make local calls to their ISPs, leading to large per line losses. World Wide Web access presents its own set of problems. On the one hand, Web browsing involves long periods of local calling that offers revenue opportunities. Further, expanded Internet usage in the U.S. has seen approximately 10 million people ask for a second telephone line, which are cheap to in- stall and profitable to offer. If this trend is repeated in Africa, telecommunications companies can expect to earn revenue from second line installation and rental. They can also lease lines to ISPs to handle the large quantity of data transfer that web ac- cess requires. On the other hand, networks that were not designed to handle calls measured in hours rather than minutes can be over- loaded by. Internet browsing, requiring the installation of new switch and transmit equipment. Whether this balance of prob- lems and opportunities will lead to increased profitability or further losses depends very much on the assumptions made about network capacity and the pricing of telecommunications company services. The following model explores the above issues. It was con- structed to do two things. First, it measures the theoretical ad- vantages (in reduced phone and/ or fax bills) to an individual subscriber, and revenue losses to a phone company from using email rather than phone or fax. This is calculated based on the number of pages of information sent. Second, it calculates ISP cost and revenue streams (based on a monopoly international hub), telecommunications cost and revenue streams, and user 43 Economic Toolkit for African Policy Makers costs and savings at various levels of Internet usage. Table A 1.1 Phone, Internet and Equipment Because of the assumptions and simplifications made, Charges in Mozambique these estimates can be considered only very rough approximations to the real costs and benefits of the us$ Internet to African users, ISPs and telecommunications companies. • 3 Minute Local call 0.04 3 Minute Regional call 0.60 Benefits to users 3 Minute International call 7.65 I (UK) Accounting Rate (/min) 1.19 Table Al.1 lists phone and equipment charges for Web: Connection 10.00 Mozambique (the Internet charges are regional I Web: Yearly fee Fax Machine 600.00 400.00 equivalents). With these numbers, we can estimate Computer 1,300.00 the yearly cost of communication for users. In this run, Modem 175.00 we assume a four-year depreciation rate for computer I equipment and a 20 percent cost of capital for users. We also assume that each user has to purchase a I computer to access the Internet. I What are the fixed costs to each method of communication? We have assumed that a telephone line is already installed (all compared systems would require a telephone, anyway). There is no extra equipment required to make a voice call, so the "fixed cost" of phoning is zero. The fixed cost to faxing information is the yearly cost of a fax machine allowing for depreciation and cost of capital. Here, it has been estimated to be in the region of US$155 per year. For email, the fixed cost is taken to be Internet access charges and the cost of a computer and a modem, allow- ing for depreciation and the cost of capital (or US$1,172 per year). Given that computers are useful for far more than Internet ac- cess, charging the entire cost of the computer against email fixed costs effectively overestimates this cost. Internet access as de- fined here also gives unlimited access to WWW, so that charg- ing all of the cost of Internet access against fixed costs for emailing also effectively overestimates the fixed costs of emailing. The high cost of capital discriminates against emailing as a cheap communications option as well. Therefore, the fixed costs of emailing presented here are, if anything, exaggerated. What are the costs of delivering this information per page? As mentioned above, a person can read out a page of informa- tion in about one and a half minutes. A fax can transmit the same information in about 18 seconds. A standard 28 k/ s modem should take around half a second to transmit the data. On the basis of data given above, we can calculate the approximate cost of sending a page of information by multiplying the number of pages by the amount of time taken in minutes, dividing by three and multiplying by the cost of a three minute call. The actual time spent transferring data by email is prob- ably shorter than the time taken to connect and disconnect from the email server. To calculate the yearly cost of emailing, then, 44 we have assumed that each business day it takes three minutes to connect, upload and download email, and disconnect. This gives a flat amount that emailers are likely to have to pay in telephone charges. For emailers dialing into a local ISP, this will amount to 260 multiplied by the cost of a three minute local call. For emailers dialing into an ISP in a different region of the coun- try, this will amount to 260 multiplied by the cost of a three minute regional call. This is the only telephone charge that emailers will face - in that three minute period, they can down- load and upload their email for other regions in the country and for other countries in the world. Once again, charging all email telephone costs to local email (rather than dividing it up amongst local, regional and international) gives an exaggerated picture of the costs of local emailing. As is clear from Table Al.2, email is not a great money saver on local calls in Mozambique. If a user talks on the telephone for about 250 hours a year (about an hour each business day), enough time to "speak" 10,000 pages, even over a local email service the difference between email and voice costs is a mere US$190. This is far less than the difference in fixed costs between voice and email ($1,172). Given that email and voice are not perfect substitutes, then, consumers will most likely continue to use local voice whenever it is preferable. With regional calls, however, the savings are larger. Be- cause the costs of email have all been accounted for, the cost of 10,000 pages of regional email is close to zero. This compares to a cost of US$3,000 forl0,000 pages spoken (in 250 hours). This difference is enough to pay for an emailer's com- Table A 1.2 Non-Fixed Cost of puter, modem, Internet hookup and monthly fee, and Communication to User (US$) regional connection to a server (total: US$1,328). On the other hand, a user would still be better ad- Pages Sent/yr 100 1,000 10,000 vised to buy a fax machine, where the combined fixed costs and regional phone bill would come to just $755. Local Only a heavy user of regional telephony in Mozambique Phone 2 20 200 should buy a computer and hook up to the Internet purely Fax 0 4 40 to replace that service. Email (local) 10 10 10 Email (regional) 156 156 156 International communication is where the savings become impressive. Talking for about an hour each busi- Regional ness day to the U.K. would cost a Mozambican business- Phone 30 300 3,000 man approximately US$38,250 per year. Faxing that Fax 6 60 600 information would still cost US$7,650. Again, taking all Email (local) 0 0 0 of the yearly costs associated witli. a regional Internet Email (regional) 0 0 0 connection (US$1,328), and assuming that it is only used for international email substituting for fax traffic, the International yearly savings amount to US$6,322. For example, a busi- Phone 383 3,825 38,250 ness with 10 staff members buys each one a computer, Fax 77 765 7,650 Email (local) 0 0 0 modem and Internet connection. Fixed costs rise by Email (regional) 0 0 0 US$13,280 per annum. But if every employee substitutes 45 Economic Toolkit for African Policy Makers Table A 1.3 Fixed Costs of Internet Access 10 minutes of voice traffic with the U.K. for email in Mozambique, per Year (US$) each business day, and one 10-page fax to the U.K. with email per day, telephone bill savings will amount to US$37,223, increasing profits by Item Cost Yearly % of total US$23,943 per year, or over US$2,000 per employee. equivalent How can email be made even more attractive? cost Presently, it is the high fixed costs of Internet ac- cess which make the system uneconomic for small Computer 1,300 507 43 users. These fixed costs come from two sources: ISP Modem 175 68 6 Web: Connection 10 2 0 service and computer equipment. In turn, NCS Web: Yearly Fee 600 51 Ghana (an ISP) estimates that 90 percent of its costs come from the expense of international connection, and a large percentage of the cost of equipment is accounted for by import tariffs. Although telephone charges are not a particularly signifi- cant cost for email use, they do make a great deal of difference for a user's cost of full Internet access. A user, who accesses the WWW for one hour each business day, will have to pay approxi- mately US$208 dollars per year in telephone charges if there is a local connection to an ISP available, or US$3,120 if Internet ac- cess is possible only through a regional call. Clearly, allowing low-cost telephone access to ISPs will be an important tool for encouraging wider Internet use. Table A 1.4 Revenue to Telcos per Page (US$) Losses to PTTs Pages Sent/yr 100 1,000 10,000 A major benefit of emailing over voice and fax communication is the savings made on telephone Local charges, which will have detrimental effects on Phone 2 20 200 telecommunications company revenue streams. For Fax 0 4 40 10 email, the only revenue to telecommunications Email 10 10 Email (regional) 156 156 156 companies comes from the daily call-up to deliver to, and receive messages from, the ISP. Regional For international calls, however, the picture is Phone 30 300 3,000 complicated somewhat by international accounting Fax 6 60 600 rates. This is a rate set between countries to com- Email 0 0 0 0 pensate carriers in recipient countries for their part Email (regional) 0 0 of the connection costs of a call originated in the International Out other country. One half of the accounting rate (the Phone 293 2,933 29,325 settlement rate, in Mozambique's case about 60 cents Fax 59 587 5,865 a minute) is paid to the recipient carrier per minute Email 0 0 0 of voice traffic received from the originating coun- Email (regional) 0 0 0 try. This reduces telecommunications company rev- International In enues for outgoing calls, but provides extra revenues Phone 89 893 8,925 from incoming calls. Because email bypasses the Fax 18 179 1,785 accounting rate regime, the telecommunications Email 0 0 0 company receives no payment for incoming email. Email (regional) 0 0 0 For a company that sends and receives 10,000 pages 46 of email each year, the loss to the telephone company amounts to US$7,650 in revenue per year if the email is substituting for fax and US$38,250 in revenue per year if the email is substitut- ing for voice. With the accounting rate calculation, it is possible to make a rough estimate of the lost profits from email substitution. The settlement rate is meant to compensate for the cost of delivering the traffic from half way between the countries to the recipient (the half circuit). In fact, accounting rates are known to be well above cost. However, they present an upper estimate of the cost of providing the connection. Three minutes of traffic at account- ing rate cost equals US$3.57. The amount charged to Mozambican users is US$7.65, suggesting an approximate profit margin to the telephone company on outgoing calls of US$4.08, or 114 per- cent. Given that real upper-bound estimates of the costs to tele- communications companies of providing this service are estimated at closer to 30 to 40 cents per minute, profit margins rise to US$7.15 a minute. Data based broadly on the Mozambique cost and price struc- ture for Internet provision, telecommunications supply and equipment and usage levels, help to demonstrate the present and potential financial impact of the Internet on ISPs, telecom- munications companies and subscribers. · Minimizing the cost of the Internet For greater detail on the information presented in the model below, see Annex 6. Data and estimates on the number of subscribers, the average length of an email and average levels of Internet use were provided by UEM. Estimates on the average number of email sent and received and their destination come from the CABECA survey, which was also used to estimate the percentage of email that replaced other methods of Usage statistics Number of Internet subscribers 3,500 Proportion web subscribers with local access 0.90 Av. subs. Hours of web access/year 100 Average User Average number of emails sent/year 240 Average number of emails received/year 200 Average length of email (pages) 4 % email replacing fax 25 % email replacing phone 10 % email local destination 10 % email regional destination 5 % email international destination 85 47. Economic Toolkit for African Policy Makers PTT Charges to ISPs License 0 PTT charge for 64 kbs of LL/Satellite 6,000 communication as opposed to being messages that otherwise would not be sent. 1 There is no license charge for running an ISP in Mozambique PTT Details (US$) Cost per main line, initial 4,500 Cost per main line, recurring 600 Cost of installing second (and subsequent) lines 500 Cost per second main line, recurring 600 Cost of switch and transmit, initial 1,125 Cost of switch and transmit, recurring 188 Cost of 64 kbs of LL/satellite provision 4,800 and the costs of a 64 kps leased line (from within the capital to the international gateway) is a surprisingly low $6,000 per year. PTT costs are all estimates based on World Bank data and author calculations. ISP running costs are estimated from USAID data on Mozambique. It should be pointed out that these costs are for a monopoly provider. Presently, there are three ISPs in Mozambique, which together provide access for 3,500 subscrib- ers and only two of which utilize the low-cost international ac- cess provided by the Leland link. The largest provider, UEM, does not use the Leland link because of concerns about the qual- ity of the leased line to the link provided by the Mozambican PTT. It should also be pointed out that this ISP model probably underestimates the cost of labor and of marketing. Below are the cost breakdowns for three U.S. Internet providers and the comparative breakdown for the model presented in this chap- ter. The cost structure of a U.S. ISP is as follows: 1. Cost of revenues (for ISP) is comparable to the cost of .goods sold by a manufacturing company. This figure represents the cost of producing the service. Depreciation can be part of this cost, although 1 It should be noted that the traffic data is unclear. First, we do not know if the users followed the rules and ignored mass-mailings when calculating the number of emails sent and received. 48 Expenditure Breakdown of U.S. Internet Service Providers America On Line costs and expenses (US$ 000) 1996 costs of revenues 627,372 61% marketing 212,710 21% general and administrative 110,653 11% depreciation and amortization 7,078 1% equipment lease 0 0% product development 53,817 5% non recurring items r 0 0% Acquired Research and Development 16,981 2% Total costs and expenses 1,028,611 100% Netcom costs and expenses (US$ 000) 1996 costs of revenues 88,369 52% marketing 51 ,237 30% general and administrative 23,610 14% depreciation and amortization 0 0% equipment lease 0 0% product development 6,020 4% non recurring items 0 0% Acquired Research and Development 0 0% Total costs and expenses 169,263 100% Californian ISP costs and expenses (US$) 1997 costs of revenues 497,729 33% marketing 278,505 19% general and administrative 674,960 45% depreciation and amortization 37,429 0% equipment lease 0 0% product development 0 0% non recurring items 0 0% Acquired Research and Development 0 0% Total costs and expenses 1,488,623 100% African ISP (in main model) costs and expenses (US$) 1998 costs of revenues 83,520 62% marketing 5,043 4% general and administrative 17,340 13% depreciation and amortization 28,933 21% equipment lease 0 0% product development 0 0% non recurring items 0 0% Acquired Research and Development 0 0% Total costs and expenses 134,836 100% Source: Anat Bernstein-Reich, 1998 49 Economic Toolkit for African Policy Makers many companies list their depreciation and amorti- zation in a separate account. The cost of revenues includes direct labor, direct overhead and any other direct cost. Most U.S. ISPs and OLSPs (online service provider) have about 40 to 60 percent of their costs accounted for by costs of revenues. African ISPs are likely to face higher costs because of their higher costs of equipment, financing and communications in the region. 2. The sales and marketing cost is a component that should not be neglected. In the U.S., marketing expenses account for 20 to 30 percent of total costs. A large percentage of these costs is attributed to the high comp~tition that exists among ISPs. A market- ing budget is important for Africa not necessarily because it can help overcome the competition (although this might become increasingly impor- tant), but because it may be used as a tool to en- courage Internet use and increase Internet penetra- tion. These costs are not properly accounted for in the main model, and the model might, therefore, underestimate the long term costs of Internet ser- vice provision in Africa. 3. The research and development account of the OLSP is much higher than that for the simple ISP in the U.S. Unlike ISPs that provide simple connectivity, the OLSP provides content and other Internet applications as well. This requires the companies to invest in product development and content. African ISPs are also likely to have to develop content because of the language issues and the need for local content. Again, this is a cost not accounted for in the main model. Despite these problems, we believe that the model offers a ISP Details (US$) Administrative computer cost 55,000 Other basic computer costs 1,000 Setup for Leased line/VSAT 0 Other setup costs 17,800 Foreign telco charge for 64kbs VSAT/LL 24,000 Internet backbone connection 0 Recurrent salary and maintenance 21,000 50 Assumptions Number of subscribers to each computer 3 Email is accessed every business day for X minutes 3 Average second line for web per subscriber= X 0.05 In hypothetical ISP income streams, user support (per user) costs $X per year 10 In hypothetical ISP income stream, one extra ISP phone line per X web subscribers 20 In hypothetical ISP income stream, one extra ISP computer per X web subscribers 1,000 In hypothetical ISP income stream, 64kbs of outside link for each X subscribers 1,000 In hypothetical ISP income stream, each new subscriber requires X new switches 0.01 Replacement of computer goods after X years 4 Replacement of telecoms equipment after X years 10 Replacement of other goods after X years 10 Real cost of capital for telco = X% 0.10 Real cost of capital for subscriber/ISP = X% 0.20 Telco YRC rate for telco.s equipment= 0.16 User YRC rate for computer equipment= 0.39 User YRC for other equipment= 0.24 YRC = yearly return to capital reasonably good estimate of ISP costs and telecommunications company revenues from ISPs. Assumptions necessary to the model are based on a range of sources and estimates. Here, it should be noted that, as op- posed to the results above, the assumption is that subscribers share computers. In the Mozambican case, the primary source of telecommu- nications company revenues from the Internet is the extra user traffic. This new traffic will be driven by the use of WWW, and not email. In Mozambique, the total user access to the Internet adds US$760,000 to phone bills. A monopoly ISP would pay a surprisingly small amount to Mozambique Telecom (US$36,600). This is both because there· is no ISP license charge in the coun- try, and because most payments for international access go to the Leland link, rather than the Mozambique PTT. Further, leased lines in Mozambique, at only US$6,000 for a connection to the Leland link, are very cheap. In Kenya, on the other hand, where a 64k international leased line connection costs US$177,000, ISP costs would be a more significant percentage of total telecoms revenues (although still smaller than user access). This also has a significant impact on ISP profits. The relatively low cost of international Internet connectivity, and the relatively low need for bandwidth because of low usage, reduce dramatically the most significant portion of an ISP' s costs in Mozambique. Although telephone, leased line, satellite ari.d Internet connections account for about 48 per- cent of the ISP's budget when it has 3,500 subscribers, this is 51 Economic Toolkit for African Policy Makers Mozambique International ISP Telco Revenue Gained from Internet Use Equipment 89,616 Subscriber New lines 12,600 Yearly inet + telecoms 134,490 Subscriber Extra traffic 759,360 Salary and maintenance 56,000 ISP License 0 ISP Subscriber lines 12,600 Web revenue 2, 107,000 ISP lntnl. Connection 24,000 TOTAL 808,560 Profit/loss 1 ;826,894 Telco Revenue Lost From Internet Use International Calls Out 732,832 Subscriber Costs, Savings International Faxes Out 366,416 International Calls In 179,303 Equipment 664,739 International Faxes In 89,652 Internet Access 2,107,000 Regional Calls 4,410 Subscriber New Lines 12,600 Regional Faxes 2,205 Subscriber Extra Traffic 759,360 Local Calls 588 Local Faxes 294 Telephone savings 1,441,298 TOTAL t,375,700 significantly below what might be expected in Africa as a whole. This allows for the possibility of making huge profits at a US$600 yearly ISP subscription charge. A monopoly provider would be making excessive returns if it faced this hypothetical environ- ment. On yearly outlays of US$280,000, such a company would be making revenues of over US$2 million. In part, this is be- cause of returns to scale, but only in part. The estimated fixed costs for a Mozambican ISP would amount to approximately US$77,000 a year. More significant is that a company with such comparatively low costs of international access, low demand for bandwidth (an average of only 64kps for each 1,000 customers), and an equally low demand for ISP computing power, faces very low costs overall. This makes it a strong argument for competi- tion to force down prices charged by an ISP. The primary source of revenue loss through substitution for telecommunications companies is from the decline in outgo- ing international telephone traffic. Email substituting for voice and fax calls deprive the Mozambique PTT of more than a mil- lion dollars each year, or about 1.5 percent of revenue. Lost rev- enue from settlement rate payments not made because of incoming communication substitution adds a little over another quarter of a million to the revenue losses. The major cost for subscribers is Internet access which in- cludes both ISP and PTT charges. Together, at a cost of approxi- mately US$3,000,000 (or US$820 per head), these figures dwarf that for the annual cost of equipment (assuming three subscrib- ers to each computer): US$190 per subscriber per year. Again, it 52 Scenario Web Web Ave user Foreign ISP ISP 64 kps connection yearly fee hrs of charge for computer for X users 1 10 600 100 24,000 1,000 1,000 2 10 600 100 24,000 400 400 3 10 600 100 170,000 1,000 1,000 4 10 600 100 170,000 400 400 5 10 600 260 170,000 150 150 6 0 168 100 24,000 400 400 7 0 168 100 150,000 1,000 1,000 appears that lowering the cost of telephone and Internet access is probably the most significant thing that countries can do to encourage users to access the new technology. By changing some of the underlying assumptions of the model, it is possible to see the impact on ISP revenues. The first scenario presents the base scenario, which suggests very large profits for the hypothetical ISP. The second scenario is that the ISP brings up its Internet access standards to international norms, and that a 64 kps line is sufficient for 10 people to be on line at any one time. Given average use of 100 hours per year, the aver- age user is online 0.115 percent of the time, suggesting that, at average loads, a 64 kps bandwi.dth is satisfactory for 874 sub- scribers. Figures from NCS in Ghana suggest that peak hours see up to three times the number of people on line than the av- erage load. To provide a good quality of Internet access, there- fore, 64kps of bandwidth is required for at least every 400 subscribers. Altering this assumption reduces ISP profits, but because international bandwidth is so cheap, profits remain ex- tremely healthy. In the third scenario, the cost of international bandwidth is raised to US$170,000 for 64 kps. This is similar to prices paid by ISPs in other countries, including Kenya. Raising the price of connectivity at the present level of service reduces prices more significantly, bringing Scenario ISP Profits (US$) at Number of Users down ISP profits to 1.2 million at the present level of usage. O users 1,000 users 3,500 users 10,000 users If both the level of service and the cost of connectivity 1 -76,961 558.982 1,826,894 5,820,509 are increased (scenario 4), 2 -76,961 456,490 1,570,665 5,051,820 profits for a monopoly pro- 3 -222,961 412,982 1,242,894 4,360,509 vider charging US$600 per 4 -222,961 18,490 256,665 401,820 5 -222,961 -770,493 -2,702,024 -6,882,507 year for access fall to 0.26 mil- 6 -76,961 -25,510 51,665 231,820 lion, or about 12 percent of 7 -202,961 -49,018 -196, 106 -259,491 revenues. If subscribers begin 53 The Costs and Revenues from Internet Communication (C:\charles\cost2.xls) Phone Charges, US Dollars Installation (business) 108 Yearly rental (business) 61 3 Minute local call 0.04 3 Minute regional call 0.60 3 Minute International call 7.65 (UK) Accounting Rate (/min) 1.19 Internet Charges Web: Connection 10 Web: Yearly fee 600 Equipment Fax Machine 400 Computer 1,300 Modem 175 Usage statistics Number of Internet subscribers 3,500 Proportion web subscribers with local access 0.90 Av. subs. Hours of web access/year 100 Average User Average number of emails sent/year 240 Average number of emails received/year 200 Average length of email (pages) 4 % email replacing fax 25 % email replacing phone 10 % email local destination 10 % email regional destination 5 % email international destination 85 PTT Charges to ISPs License 0 PTT charge for 64 kbs of LUSatellite 6,000 55 Economic Toolkit for African Policy Makers PTT Details Cost per main line, initial 4,500 Cost per main line, recurring 600 Cost of installing second (and subsequent) lines 500 Cost per second main line, recurring 600 Cost of switch and transmit, initial 1, 125 Cost of switch and transmit, recurring 188 Cost of 64 kbs of LUsatellite provision 4,800 ISP Details Administrative computer cost 55,000 Other basic computer costs 1,000 Setup for Leased line/VSAT 0 Other setup costs 17,800 Foreign telco charge for 64kbs VSAT/LL 24,000 Internet backbone connection 0 Recurrent salary and maintenance 21,000 Assumptions: Revenue to PTT is phone charges to user minus half of IAR for intnl. calls Number of subscribers to each computer 3 Email is accessed every business day for X minutes 3 Average second line for web per subscriber = X 0.05 In hypothetical ISP income streams, user support (per user) costs $X per year 10 In hypothetical ISP income stream, one extra ISP phone line per X web subscriber 20 In hypothetical ISP income stream, one extra ISP computer per X web subscribers 1,000 In hypothetical ISP income stream, 64kbs of outside link for each X subscribers 1,000 In hypothetical ISP income stream, each new subscriber requires X new switches 0.01 Replacement of computer goods after X years 4 Replacement of telecoms equipment after X years 10 Replacement of other goods after X years 10 Real cost of capital for telco = X% 0.10 Real cost of capital for subscriber/ISP = X% 0.20 (Telco YRC rate for telco.s equipment = 0.16 (User YRC rate for computer equipment= 0.39 (User YRC for other equipment= 0.24 56 User Savings and Telco Revenues per Page LOCAL REGIONAL (Extra) Pages SenVyr 0 100 1,000 10,000 100 1,000 10,000 Cost to user, per year Phone 0 2 20 200 30 300 3,000 Fax 155 155 159 195 6 60 600 Email (local) 1,172 1,182 1,182 1,182 0 0 0 Email (regional) 1,172 1,328 1,328 1,328 0 0 0 Revenue to PTT from user per year Phone 0 2 20 200 30 300 3,000 Fax 0 0 4 40 6 60 600 Email (local) 0 10 10 10 0 0 0 Email (regional) 0 156 156 156 0 0 0 INTERNATIONAL (Extra) Pages SenVyr 100 1,000 10,000 Cost to user, per year Phone 383 3,825 38,250 Fax 77 765 7,650 Email (local) 0 0 0 Email (regional) 0 0 0 Revenue to PTT from user per year Phone 293 2,933 29,325 Fax 59 587 5,865 Email 0 0 0 Email (regional) 0 0 0 Pages Rec.d/yr 100 1,000 10,000 Revenue to PTT from sender per year Phone 89 893 8,925 Fax 18 179 1,785 Email 0 0 0 57 Economic Toolkit for African Policy Makers Telco. Revenue Gained Subs. 0 1,000 3,500 10,000 Subscriber New lines 12,600 ISP, hypothetical cost/revenue Subscriber Extra traffic 759,360 ISP License 0 cost 76,961 91,018 280,106 679,491 ISP Subscriber lines 12,600 ISP lntnl. Connection 24,000 revenue 0 650,000 2, 107,000 6,500,000 TOTAL 808,560 profit -76,961 558,982 1,826,894 5,820,509 Telco. Revenue Lost International Calls Out 732,832 Telecom. Income, Revenue International Faxes Out 366,416 International Calls In 179,303 revenue 6,072 230,160 808,560 2,301,600 International Faxes In 89,652 cost 6,132 76,247 270,267 766,080 Regional Calls 4,410 Regional Faxes 2,205 substit. 0 393,057 1,375,700 3,930,570 Local Calls 588 Local Faxes 294 profit b. -60 153,913 538,293 1,535,520 TOTAL 1,375,700 profit a. -60 -239,144 -837,407 -2,395,050 Total Web Revenue Effect Users, Costs, Benefits TOTAL -567,140 cost 0 1,012,485 3,543,699 10, 124,855 International ISP substit. 0 411,800 1,441,298 4,117,995 Equipment 89,616 Yearly inet + telecoms 134,490 Salary and maint 56,000 Web revenue 2,107,000 Profit/loss 1,826,894 Subscriber Costs, Savings Equipment 664,739 Internet Access 2,107,000 Subscriber New Lines 12,600 Subscriber Extra Traffic 759,360 Telephone savings 1,441,298 Total Cost 2,102,401 I 58 Annex Two: Calculation of Telecommunications Revenue Lost to Substitution Using FCC Data A second simple way to estimate the approximate losses to tele- communications companies from international substitution is presented below. We have already presented estimates for the amount of incoming and outgoing telephone and fax traffic the average user substitutes with email. The FCC, as part of its at- tempts to put downward pressure on international accounting rates, has estimated the real cost per minute for countries to ter- minate an international call (the "half circuit cost") for a range of countries. This will be broadly the same as the cost of origi- nating the call (both halves of the circuit are identical). In Af- rica, numbers are presented for Kenya (42.6 cents/minute) and South Africa (16.9 cents/ minute). Despite methodological prob- lems with the FCC' s calculations, these numbers should approxi- mate the real costs of completing a call. We also know the accounting rate and the international call charges of a range of African countries. With all of this data, we can estimate the amount of profit (revenue minus cost) that telecommunications companies lose from each email user each year (using the Kenya number as an upper bound of costs and South Africa as a lower bound). This leads to the figures below: Table A2.1 Lost Profits to International Email Substitution Country Estimated 3 minute Accounting Lost profit/yr Lost profit/yr Revenue Lost profit as number of international rate with (US$) (low) (US$) (high) /yr ($m) %revenue subscribers call (US$) US$/min South Africa 700,000 4.58 1.00 31,943,473 n.a. 3674.7 0.87 Zimbabwe 8,000 7.50 0.75 913,070 1, 172,742 146.2 0.80 Mozambique 2,750 7.65 1.19 316,384 405,646 62.6 0.65 Ghana 4,000 7.50 1.00 450,785 580,621 65 0.89 Ethiopia 1,000 4.80 1.80 46,086 78,545 72.3 0.11 59 Economic Toolkit for African Policy Makers Annex Three: The Relationship between Communications Liberalization and Internet Pricing and Penetration A recent study by ITU, Cha1¥!1g~§_jo the N~tivork: Telecoms and t_he In~~rnet (1997), arguec!, that the following factors are signifieantly-::-- co:i;related with global Internet :E'enetra!ion: wealth, the qua!}-- tit_z_and the qu~lity of telecommunications eg,uiEment, the num- ber of PCs, low telephone and Internet charges, language, literacy and trai!i.ing. -- ~--Concentrating on the link between policy and Internet growth, an attempt was made to explore in different ways the relationship between telecommunications liberalization,~­ nology friendly" policy~gim-es, and the ~nd-extenLof Interne±_E~_c_es_s in Afr_ica. Three differen!sets of data are pre- sented below, which suggest the existence of such a relation- ship, and its positive nature. The first takes two different measures of sector liberaliza- tion: one from World Bank datc.i._ (Biirfill,1997), ari:d'''TI!le1rom O'Neill's article onAfrica: "The New TelecomsFrontier" (1998). TaBie A3.1 lists this data for all Sub-Saharan African countries wherever information was available. The first column presents a private involvement index, based on O'Neill's report on the state of private involvement in African countries at the start of 1998. Countries receive one point for each of the following: new telecommunications laws and regulations; introduction of pri- vate investment into non-basic services excluding cellular; in- troduction of private investment into basic services (planned by the end of 1998 or already carried out); introduction of private investment into cellular services. The second index measures a broader definition of liberal- ization from World Bank data at the start of 1997. Countries are given one point for each of seven stages: • liberal CPE trade; • separate posts and telecommunications; • independent regulator; • reform in progress (or announced); • new sector law/ private cellular; and • privatized telecommunications company. The table then lists two basic features of the economy that are likely to be connected with the scale of Internet penetration - population and GDP per capita (data from the ITU Yearbook). The next column estimates the number of host sites in the coun- try, using information from the Network Wizards site, 60 www.nw.com. However, it should be noted that Network Wiz- ards have changed their method of counting host sites, so that the data between 1998 and previous years are not strictly com- parable. Access prices come from the list of recommended ISPs in African countries, from the 1997 UNIDO study of Internet con- nectivity in Africa. The next two columns measure the monthly cost of web access per page as calculated for two levels of usage (measured in terms of pages per day) by Jensen. The next col- umn measures the age of Internet access in a country, with a score of one indicating that there was TCP /IP connection at the end of 1995 (from the ITU yearbook). The final column indicates Table A3.1 African Telecoms Liberalization and Internet Access Liberalization Index Economy Hosts Monthly Cost (US$) Internet Access number of 1998 private 1997 10 50 yes/no providers involvement liberalization population GDP/capita 1998 pages/day pages/day 1995 1997 Angola 1 1 10.8 997 4 149 149 0 4 Benin 1 1 5.5 290 13 25 25 0 1 Botswana 2 4 1.4 2905 550 23 23 0 4 Burkina Faso 0 2 10.4 153 45 24 24 0 3 Cameroon 1 3 13.3 454 2 20 20 0 3 CAR 1 2 3.3 303 0 27 27 0 1 Chad 1 2 6.5 124 0 47 47 0 2 Congo-Brazzaville 1 5 2.7 627 4 139 139 0 1 Cote d'Ivoire 4 7 14.2 487 253 33 33 1 3 Djibouti 1 2 0.6 835 0 198 198 1 1 Ethiopia 2 4 57.0 86 78 100 100 1 4 Gabon 2 2 1.3 3812 0 137 137 0 1 Ghana 3 7 17.1 320 252 51 51 1 7 Guinea 2 5 6.6 539 0 27 27 1 1 Kenya 3 3 27.0 267 458 75 75 1 10 Lesotho 1 0 2.8 444 0 61 152 0 1 Madagascar 4 5 13.5 215 17 20 20 0 3 Malawi 1 2 9.7 120 0 32 32 0 2 Mali 0 3 9.8 196 0 166 778 0 4 Mauritius 2 6 1.1 3134 201 17 17 0 6 Mozambique 3 4 17.4 87 69 20 20 0 3 Namibia 1 5 1.6 1923 640 56 56 1 2 Nigeria 3 4 111.3 412 49 17 39 0 7 Senegal 4 6 8.4 479 117 20 20 1 4 Sierra Leone 1 3 4.5 178 0 238 238 0 2 South Africa 3 5 41.5 2923 122025 14 14 1 46 Swaziland 0 1 0.9 1077 330 57 58 1 1 Tanzania 4 6 29.7 124 25 50 50 0 5 Togo 0 3 4.1 243 37 95 95 0 2 Uganda 3 3 19.0 244 30 29 29 1 4 Zambia 4 4 9.4 330 181 25 25 1 2 Zimbabwe 2 2 11.0 444 599 6 23 1 10 Avera ge 1.9 3.5 14.8 774 3937 62 86 0.41 5 61 Economic Toolkit for African Policy Makers Table A3.2 Telecoms Liberalization and Internet Access in Sub-Saharan Africa Liberalization Index Economy Hosts Monthly Cost ($US) Internet Access number of 199 8 private 1997 10 50 yes/no providers involvement liberalization population GDP/capita 1998 pages/day pages/day 1995 1997 Average 1.9 3.5 13.9 705 128 64.00 87.97 0.39 3.35 0 2.3 6.3 417 103 85.50 238.75 0.25 2.50 1 to 2 2.9 8.2 1013 123 76.59 82.94 0.29 2.71 3 to 4 4.9 26.7 297 145 34.00 36.20 0.60 4.80 0 to 3 8.3 599 89 81.53 123.94 0.29 3.06 4 to 7 20.8 833 174 42.71 44.29 0.50 3.71 ':tf""fi?e number of providers present in 1997, again from Jensen (1997). If we take average values for less and more liberal coun- tries excluding South Africa (see Table A3.2), it becomes clear that more liberal regimes have, on average, more and cheaper access to the Internet. Although, countries such as Burkina Faso, with cheap access but a very illiberal telecommunications mar- ket, suggest the relationship is far from perfect. According to both liberalization measures, countries with a more open tele- communications sector have more host sites, lower monthly nternet charges and a greater number of providers. Table A3.3 suggests that competition in Internet service provision is also associated with higher rates of Internet penetration and lower costs. Of course, this correlation between liberalization in tele- communications and Internet provision and the extent and cost of access does not necessarily prove causation. In particular, it is difficult to disentangle the separate impact of liberalization of Internet access and that of the telecommunications sector more generally. Further, countries with more liberal telecommunications and Internet markets tend to be more populous (with an aver- age population of 21 million as compared to 8 million for less .JiQei;a~arkets). We have seen elsewhere that there does appear Table A3.3 ISP Competition and Internet Access in Sub-Saharan Africa Number of ISPs Liberalization Index Economy Hosts Monthly Cost ($US) 1998 private 1997 population GDP/capita 1998 10 pages/day pages/day involvement liberalization One 1.13 2.25 3 991 43 83.88 95.38 More than One 2.13 3.87 18 605 157 57.09 85.39 62 ,,,..--------~~ . to be some economies of ~,e;.in Internet provision, Table A3.4 Telecoms Liberalization and so it is possfole-tharsmaller countries are less able Costs of Internet Access, 1997 to take advantage of the benefits of competition without losing these economies of scale. One ar- Liberalization Monthly Cost ($US) gument against this interpretation is that countries 1998 private 1O pages/day 50 pages/day with the greatest private involvement (scoring three involvement 1to four on O'Neill's index) have the lowest GDP 1 per capita of the group ($297 per year). Despite their 0 85.50 238.75 larger populations, they have smaller total GDPs 1 90.18 98.45 than the group that scores one or two on the pri- 2 51.67 54.50 3 38.40 42.80 vate involvement index. This suggests that scale 4 29.60 29.60 economies cannot explain lower costs or the num- ber of providers in the more liberal telecommuni- cations environments. The data provide enough evidence to suggest that liberalization is a boon to the spread of the Internet in Africa. The method of calculating the cost of Internet access above excludes one important component of the cost of access: the price of calling the ISP. Table A3.5 is based on a broadly similar meth- odology to the OECD' s Report on Information Infrastructure Con- vergence and Pricing (1996), which is the source of the OECD data reported in the table. Data on the pricing policies of African ISPs are from Jensen (1997), data on PTT pricing are taken from the ITU (1997). The price per month for Internet access is calcu- lated as follows: The ISP connection charge divided by 36, the monthly charge, and any extra charges that apply to 20 or 30 hours of use. The PTT charge is the telephone connection charge divided by 60, the monthly charge for a telephone and the cost of 20 or 30 hours of local calls (multiplied up from the three- minute local call rate reported by the ITU). African countries in gray in the table have comparatively "liberal" telecommunica- tions sectors, scoring more than three on the liberalization in- dex based on World Bank data. In the OECD, seven out of the eight countries with tele- communications infrastructure competition had cheaper than average total access charges while 12 of the 17 countries with- out competition were above the same benchmark (OECD, 1996). .These results are mirrored in estimates for African Internet pro- vision. Some African countries provide very competitive Internet access as compared to the OECD (although it should be noted that the calculation does not follow exactly the same methodol- ogy). Botswana, Ghana, Lesotho, Namibia, Senegal and South Africa, all provide~cheaper overall Internet access than the OECD average, for example. On the other hand, closed telecommuni- cations regimes in Africa provide 30 hours of Internet access at more than twice the OECD average cost. Only one closed re- gime provides cheaper Internet access than the OECD average - Lesotho. Lesotho appears to benefit from its proximity to the 63 Economic Toolkit for African Policy Makers Table A3.5 The Price of Internet Access in Sub-Saharan Africa and the OECD ISP charge/month PTO charge/month Total (ISP+PTO) 20 hrs 30 hrs 20 hrs 30 hrs 20 hrs 30 hrs AFRICA Angola 155 223 147 165 301 387 Benin 127 181 63 89 190 270 '"·: .:;;~:;"':''~' ;- Kenya Lesotho ',::·{~~~~ii0s'•'',' :, ,"'iff~Ahl2~' ':;:::·f~,~:~.~rii$.·,··•·•.,.::•,y· Togo Uganda Zambia.<·· Average 115 137 60. 84 175 221 Closed 139 176 74 101 213 277 Liberal 97 109 50 71 147 mo OECD France 61 91 45 64 107 156 Japan 110 110 33 44 142 153 Mexico 80 80 32 112 112 192 Portugal 42 67 56 75 98 142 UK 15 15 66 90 81 104 us 21 21 11 n 32 32 OECD Average 66 82 42 56 108 138 64 efficient (and large) South African telecommunications market. Open telecommunications regimes in Africa appear to provide both cheaper ISP charges and cheaper PTT charges. This sug- gests that market efficiencies rather than rebalancing lead to cheaper access. Statistical analysis of the determinants of Internet penetra- tion in Africa (this time including South Africa) confirms this finding. Running a regression of Internet hosts in each country against a constant and population (regression 1), there is a posi- tive relationship, but it is insignificant. Replacing population with GDP produces a strong positive relationship which is highly significant (regression 2). Undoubtedly, then, wealth plays a large role in determining the extent of Internet penetration in Africa. However, replacing GDP with number of computers (re- gression 3) produces an even better fit - the number of com- puters in a country explains 85 percent of the variation in the number of Internet hosts. This is a result that echoes one found by the OECD (1996) for richer countries and by the ITU with a global sample (1997). When entered with GDP per capita, the number of computers in a country remains significantly corre- lated with Internet penetration, whereas GDP does not (regres- sion 4). The number of personal computers and population, when entered together in a regression (5), account for 98 percent of the variation in the number of Internet hosts in the 16 African countries for which we have data. It has to be assumed that the number of computers per capita is proxying for a technology- friendly environment in general. This suggests that, while wealth is undoubtedly one important determinant of such an environment, even poor African countries can encourage far wider Internet penetration by becoming more technology friendly. However, it should be pointed out that these results will be driven largely by South Africa. Further, regressing the number of Internet hosts per capita against variables, found GDP per capita a stronger determinant of Internet penetration than com- puters per capita. Regressions of Internet hosts per capita against GDP per capita and the cost of a local call, percentage of lines that were digital, main lines per capita, modems per capita or minutes of outgoing calls as percentage of incoming calls, found all variables except GDP per capita insignificant. 65 Economic Toolkit for African Policy Makers Table A3.6 Determinants of Internet Penetration in Sub-Saharan Africa Independent variable is number of Internet hosts in each country, 1995 1 2 3 4 5 Constant 49.9 -2904 -3946 -4425 2522 0.02 -2.30 -1.42 -1.35 2.12 Population 0.00014 -0.00038 1.41 -9.84 GDP 5.90E-07 7.80E-08 11.30 0.30 Personal Computers 0.081 0.072 0.1 8.77 2.29 26.06 N 49 35 16 16 16 A-Squared 0.04 0.76 0.85 0.85 0.98 Numbers in italics are t-statistics. Source: latest available year data from ITU. 66 Economic Toolkit for African Policy Makers 1. Annex Four: The Importance of Accounting Rates. to African Telecommunications Companies Table A4.1 presents a list of African countries and an "account- ing index". The index is designed to be a very rough measure of a country's reliance on income from accounting rates. It is cal- culated as follows: ((no of minutes, international telephone traffic in) - (no. of minutes, international traffic out))* (one half the accounting rate with the United States) all expressed as a percentage of GDP. This measure can only be an approximation for a number of reasons. First, if we take the example of Swaziland, it is prob- able that the huge negative index score, based on more outgo- ing than incoming traffic, is because of a disparity of calls with South Africa, not a disparity with the US. To make the index a true measure of reliance on Table A4.1 African Countries and Dependency on 1 international accounting rate revenues, it International Accounting Rates would have to be an accounting rate average weighted by origin of international calls mi- Country Accounting Rate Payment (% GDP) nus an accounting rate average weighted by destination of international calls, all multi- plied by international traffic imbalance in turn Swaziland -0.74 expressed as a percentage of GDP. Because we Zimbabwe -0.36 only have accounting rate figures for the U.S., Kenya -0.17 the U.K. and New Zealand, we cannot con- Botswana -0.17 -0.06 struct this index. It has to be hoped that the Cameroon Guinea 0.00 U.S. rate is a fair proxy (see Table A4.2). Burkina Faso 0.03 The table is significant because it suggests Angola 0.06 the scale of the potential threat posed by both Sudan 0.06 accounting rate changes and also substitution Zambia 0.07 from voice and fax telephony to email. If the Benin 0.11 Gambia gains anywhere near 1.18 percent of Uganda 0.11 GDP per yea:t from accounting rates, any de- Mauritius 0.12 velopment that reduces the number of incom- Tanzania 0.13 ing international telephone calls (such as email Nigeria 0.13 substitution) is likely to have a serious effect Equatorial Guinea 0.41 Mozambique 0.56 on foreign exchange revenues and telecommu- Senegal 0.66 nications company profitability. Of course, this Guinea Bissau 0.70 measure does not capture the risks from re- Cape Verde 0.79 duced outgoing traffic, which is even more Sao Tome and Principe 1.15 profitable and likely to fall as a result of email Gambia 1.18 ·and callback services. 68 Table A4.2 African Accounting Rates Revenue from the US Country US net US settlement US settlement rate settlement as% revenue Proposed (9/96) B-mark Angola 102,191 0.6 73 19.1 Benin 717,345 3.0 50 23.4 Botswana 663,030 1.0 75 19.1 Burkina Faso 366,390 1.4 59 23.4 Cameroon 5,340,070 6.7 90 23.4 CAR 562,121 3.6 146 23.4 Chad 468,289 5.4 250 23.4 Congo 350, 129 0.8 87.5 23.4 Djibouti 343,543 1.6 75 19.1 Ethiopia 11,000,000 20.6 90 23.4 Gabon 713,322 0.9 87.5 19.1 Ghana 8,779,732 14.3 50 23.4 Guinea 973,649 7.8 58.5 23.4 Kenya 11,800,000 4.6 70 23.4 Lesotho 122,465 1.0 75 23.4 Madagascar 490,234 1.5 302.5 23.4 Malawi 307,174 1.2 50 23.4 Mali 1,518,875 91 23.4 Mauritius 642,746 0.9 75 19.1 Mozambique 1,518,692 2.4 79.5 23.4 Namibia 825,387 1.5 85 19.1 Nigeria 8,465,816 1.8 75 23.4 Senegal 12, 100,000 11.7 110 23.4 Sierra Leone 3,615,028 54.8 75 23.4 South Africa 27,000,000 1.0 50 19.1 Swaziland 334,872 1.5 75 19.1 Tanzania 3,905,555 5.6 75 23.4 Togo 1,516,662 13.9 87.5 23.4 Uganda 1,252,754 3.0 60 23.4 Zambia 1,659,572 1.5 60 23.4 Zimbabwe 2, 166,959 1.8 75 23.4 69 Economic Toolkit for African Policy Makers Annex Five: Potential Internet Penetration in Africa Taking GDP and percentage of income by quintile from the World Bank, the model presented here estimates the potential expen- diture per person in each income quintile on communication con- nectivity. This is based on the common assumption that communities are willing to spend around one percent of their income on access to telephones. Taking this number, estimates for the costs of computer equipment, and the costs of connectiv- ity from Jensen and the ITU, we can estimate achievable tele- phone and Internet densities (per 100 people) on the basis of costs and willingness to spend. There are three sets of estimates. The first assumes that all expenditure will go to telephone con- nectivity, the second that 75 percent will go to telephone con- nectivity, but consumers will be willing to spend 25 percent of their connectivity budget on the Internet. The third set assumes that the connectivity expenditure will be evenly divided between the Internet and the telephone. The estimates measure expected teledensity, or telephones per 100 people (teledens), and expected Internet density, or computers hooked to the Internet per 100 people (netdens), for each of the five income groups, as well as an average. Below that, the total number of telephones and com- puters hooked to the Internet is estimated, and this number used to calculate an average density in terms of telephones or Inter- net sites per square kilometer. The model, as it stands, appears to overestimate the actual number of telephones in the countries. There are two reasons for this, both connected with policy choices. The first is the prob- lem of waiting lists. In countries where political pressure keeps the costs of access low without concomitant transfers or subsi- dies, telephone companies are unable to meet the demand cre- ated by below-cost service provision. This leads to huge waiting lists and low levels of satisfied demand. In Tanzania, the wait- ing time for a telephone is more than 10 years and satisfied de- mand is but 43 percent of the total (figures for other countries are: Ghana, 67.9 percent; Cote d'Ivoire, 67.6 percent; Senegal, 85 percent). The second and more significant reason for the model over- estimating telephone demand is that although models based on the "one percent of expenditure" assumption do quite well in estimating the number of connections to telephone networks, they do this by assuming that any household whose willingness to 70 spend is below the cost of a tele- phone connection spends nothing Table A5.1 Actual versus Predicted Penetration at 25 percent Expenditure on Internet on connectivity. To take the South African example, assuming an Predicted Actual Predicted Actual Actual average household size of five, telephone telephone Internet Internet telecom the costs of a telephone connec- density density density density revenues, tion will be beyond all but those %GDP households in the top half of the fourth quintile and those in the South Africa 13.78 9.45 0.84 0.17 2.2 fifth quintile (who will also con- Tanzania 1.59 0.30 0.01 1.9 trol, directly or indirectly, tele- Ghana 3.70 0.35 0.07 0.003 1.1 Cote d'Ivoire 2.74 0.81 0.14 1.7 phones operated in businesses). Senegal 4.64 0.98 0.15 2.7 Given this, an actual teledensity of 9.45 is fairly respectable. In Tanzania, only households in the very top of the top income quintile will be able to afford a telephone or own telephones in their businesses. The issue with such an assumption is that stud- ies also suggest that if given access to public telephones, poorer people will spend one percent or more of their income on con- nectivity. This suggests that achievable teledensity, given pro- active efforts to install public telephones, should be higher than present-day teledensities. If Tanzania was to raise the number of payphones per inhabitant from 0.02 to China's level of 0.70, for example, this would increase the country's overall teledensity by nearly 25 percent (from 0.30 to 0.37). While the model overestimates actual telephone densities, it allows us to estimate potential order of magnitude telephone and Internet densities under the assumption of expanded pub- lic access to telephones and the Internet. Perhaps the first (and most obvious) thing to be noticed is that achievable teledensity or netdensity is highly dependent on two factors - the wealth of the community and the cost of telephone or Internet access. Looking at telephone access to begin with, we can see that po- tential density varies enormously between countries. If no ex- penditure is assumed to be going to Internet connectivity, countrywide potential teledensity varies between 18.38 in South Africa to 2.12 in Tanzania. Potential teledensity also varies im- mensely between income brackets - for the poorest quintile in Sou th Africa, the potential number of telephones per 100 people is little higher than the average for Tanzania, at 3.04. If we now look at potential Internet density, it is clear that, if expenditure on Internet access substitutes for access to tele- phones, increasing netdensity will have an impact on teledensity. Because access to the Internet is so much more expensive than access to a telephone, this will reduce average teledensity. To take the Tanzanian case, if Internet access expenditures substi- tute for 50 percent of expenditures that presently go toward tele- phone access, teledensity will drop from a potential 2.12per100 71 Economic Toolkit for African Policy Makers people to 1.07 per 100 people (1.06 telephones plus 0.01 tele- phones attached to an Internet access point), while net density will reach only 0.01per100 people. In terms of average number of potential telephones and Internet access points per square km, the number of potential telephones would fall from one every one and a half km 2 to one every three km 2, while the number of Internet sites would reach one every 333 km2 • Of course, the picture is not that bleak. First, because of the potential of the Internet, it is likely that expenditure on access would be as much additional as it was substitutive (especially for rich users). Second, potential teledensity at 50 percent ex- penditure on the Internet is still higher than actual teledensity achieved in Tanzania (1.06 compared to 0.30 per 100 people). This suggests that the problem of substitution expenditure on the Internet driving down teledensity is likely to be largely theo- retical. It is clear, however, that the price of communication access, especially access to the Internet, is prohibitive for many. Even if 50 percent of potential expenditure on communications access was harnessed to provide Internet access to the poorest 20 per- cent in Tanzania, the number of Internet access points that they could afford per head at present prices would be one for each 25,000 of them, or one every 800 km2 at average population den- sities. Obviously, this is not a sustainable level of access. Despite this, policy changes can be made to increase Inter- net access to a larger segment of the population. Tanzania's yearly Internet access costs (including equipment), at US$5,425, are more than five times the South African yearly Internet ac- cess costs of US$793. If Tanzania were to bring down the costs of Internet access to South African levels, potential Internet con- nectivity in the country would skyrocket. At a rate of 50 percent of potential communications expenditure being used for Inter- net access, average net density per 100 people would rise from 0.01to0.07 (or one per 10,000 to one per 1,500). At average popu- lation density, this suggests that the maximum walk to the near- est Internet access point would drop from about three hours long to around one hour (10.5km to 3.9km). The primary reason for cheaper Internet service provision in South Africa is cheaper access to leased lines and the Internet backbone. South Africa Telkom has been able to afford this cheaper backbone access in part through rate rebalancing be- tween value added services and basic services. While greater efficiency undoubtedly plays a large role, it is also clear that South Africa's basic provision of telephone services is more ex- pensive than Tanzania's (a yearly cost of US$178.2, as compared to US$57.33). Perhaps, then, it is fairer to take the whole bundle of South African costs and apply them to Tanzania. As before, we see far higher Internet access (although a little lower than 72 the SA Internet prices, because of the extra telephone charges). Yet, potential teledensity drops from 1.06 in the base case to 0.34 in the South African prices model. While overstating the case, this raises an important policy issue. If the costs of Internet ser- vice provision are reduced by lowering the costs of broadband access, this will reduce a source of cross-subsidy for basic tele- phone provision. In an environment of inefficiencies in public telephone monopolies, where demand cannot be met for basic telephone provision anyway, it should be possible to both re- duce the costs ofinternational connectivity for Internet service providers and continue provision of basic telephone services at reasonably low cost. However, there might be at least short term costs in terms of basic telephone service provision. Table A5.2 South Africa: Achievable Telephone and Internet Densities GDP (USDm) 136035 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (m) 41.5 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 81 teldens teldens netdens tel dens netdens Telephone rental/mth ($) 12.6 Internet eqp. + instal ($) 1875 1st 3.3 5.41 3.04 2.28 0.14 1.52 0.28 Internet charge/mth ($) 14 2nd 5.8 9.51 5.33 4.00 0.24 2.67 0.49 Area (1,000 km2) 1221 3rd 9.8 16.06 9.01 6.76 0.41 4.51 0.83 4th 17.7 29.01 16.28 12.21 0.75 8.14 1.49 5th 63.3 103.75 58.22 43.66 2.67 29.11 5.34 Tel. year cost ($) 178.2 Int year cost($) 793 Average 32.75 18.38 13.78 0.84 9.19 1.69 Total Phones/Site (k) '7626 5720 350 3813 700 Site Density (no/km2) 6.25 4.68 0.29 3.12 0.57 Table A5.3 Tanzania: Achievable Telephone and Internet Densities GDP (USDm} 3602 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (m} 29.6 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 46 teldens teldens netdens tel dens netdens Telephone rental/mth ($) 3.5 Internet eqp. + instal ($) 1875 1st 6.9 0.42 0.73 0.55 0.00 0.37 0.004 Internet charge/mth ($) 400 2nd 10.9 0.66 1.16 0.87 0.00 0.58 0.01 Area (1,000 km2} 945 '3rd 15.3 0.93 1.62 1.22 0.00 0.81 0.01 4th 21.5 1.31 2.28 1. 71 0.01 1.14 0.01 5th 45.4 2.76 4.82 3.61 0.01 2.41 0.03 Tel. year cost ($) 57.33 Int year cost ($) 5425 Average 1.22 2.12 1.59 0.01 1.06 O.D1 Total Phones/Site (k) 628 471 2 314 3 Site Density (no/km2) 0.66 0.50 0.002 0.33 0.003 73 Economic Toolkit for African Policy Makers Table A5.4 Ghana: Achievable Telephone and Internet Densities GDP (USDrn) 6315 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (rn) 17.1 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 196 tel dens tel dens netdens teldens netdens Telephone rental/rnth ($) 0.8 Internet eqp. + instal ($) 1975 1st 7.9 1.46 1.95 1.46 0.03 0.97 0.05 Internet charge/rnth ($) 51 2nd 12 2.22 2.96 2.22 0.04 1.48 0.08 Area (1,000 krn2) 239 3rd 16.1 2.97 3.97 2.98 0.06 1.98 0.11 4th 21.8 4.03 5.37 4.03 0.07 2.69 0.15 5th 42.2 7.79 10.40 7.80 0.14 5.20 0.29 Tel. year cost ($) 74.93 Int year cost($) 1270.33 Average. 3.69 4.93 3.70 0.07 2.46 0.14 Total Phones/Site (k) 843 632 12 421 23 Site Density (no/krn2) 3.53 2.64 0.049 1.76 0.098 Table A5.5 Cote d'Ivoire: Achievable Telephone and Internet Densities GDP (USDrn) 10069 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (rn) 14 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 148 tel dens tel dens netdens tel dens netdens Telephone rental/rnth ($) 12.3 Internet eqp. + instal ($) 1926 1st 6.8 2.45 1.24 0.93 0.05 0.62 0.10 Internet charge/rnth ($) 35 2nd 11.2 4.03 2.05 1.53 0.08 1.02 0.16 Area (1,000 krn2) 322 3rd 15.8 5.68 2.89 2.16 0.11 1.44 0.23 4th 22.2 7.98 4.05 3.04 0.16 2.03 0.32 5th 44.1 15.86 8.05 6.04 0.31 4.03 0.63 Tel. year cost ($) 196.93 Int year cost ($) 1062 Average 7.20 3.66 2.74 0.14 1.83 0.29 Total Phones/Site (k) 512 384 20 256 40 Site Density (no/krn2) 1.59 1.19 0.062 0.79 0.124 74 Table A5.6 Senegal: Achievable Telephone and Internet Densities GDP (USDm) 4867 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (m) 8.5 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 112 teldens teldens netdens teldens netdens Telephone rental/mth ($) 4.6 Internet eqp. + instal ($) 1905 1st 3.5 1.00 1.08 0.81 0.03 0.54 0.05 Internet charge/mth ($) 20 2nd 7 2.00 2.17 1.62 0.05 1.08 0.10 Area (1,000 km2). 197 3rd 11.6 3.32 3.59 2.69 0.09 1.79 0.17 4th 19.3 5.53 5.97 4.48 0.14 2.99 0.29 5th 58.6 16.78 18.13 13.60 0.43 9.07 0.87 Tel. year cost ($) 92.53 Int year cost($) 875 Average 5.73 6.19 4.64 0.15 3.09 0.30 Total Phones/Site (k) 526 394 13 263 25 Site Density (no/km2) 2.67 2.00 0.064 1.33 0.128 Table A5.7 Tanzania: Theoretical Achievable Telephone and Internet Densities (all SA costs) GDP (USDm) 3602 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (m) 29.6 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 81 teldens teldens netdens tel dens netdens Telephone rental/mth ($) 12.6 Internet eqp. + instal ($) 1875 1st 6.9 0.42 0.24 0.18 O.Q1 0.12 0.02 Internet charge/mth ($) 14 2nd 10.9 0.66 0.37 0.28 0.02 0.19 0.03 Area (1,000 km2) 945 3rd 15.3 0.93 0.52 1.39 0.02 0.26 0.05 4th 21.5 1.31 0.73 1.55 0.03 0.37 0,07 5th 45.4 2.76 1.55 1.16 0.07 0.78 0.14 Tel. year cost($) 178.20 Int year cost ($) 793 Average 1.22 0.68 0.51 0.03 0.34 0.06 Total Phones/Site (k) 202 152 9 101 19 Site Density (no/km2) 0.21 0.16 0.010 0.11 0.020 75 Economic Toolkit for African Policy Makers Table AS.8 Tanzania: Achievable Telephone and Internet Densities (SA Internet costs) GDP (USDm) 3602 Inc. quintile Expend Percentage of expenditure to telephone and Internet Population (m) 29.6 %Nat. inc. /person 100% tel 75% tel 25% lnet 50% tel 50% lnet Telephone instal ($) 46 teldens teldens netdens tel dens netdens Telephone rental/mth ($) 3.5 Internet eqp. + instal ($) 1875 1st 6.9 0.42 0.73 0.55 0.01 0.37 0.02 Internet charge/mth ($) 14 2nd 10.9 0.66 1.16 0.87 0.02 0.58 0.04 Area (1,000 km2) 945 3rd 15.3 0.93 1.62 1.22 0.03 0.81 0.05 4th 21.5 1.31 2.28 1.71 0.04 1.14 0.08 5th 45.4 2.76 4.82 3.61 0.08 2.41 0.16 Tel. year cost ($) 57.33 Int year cost ($) 793 Average 1.22 2.12 1.59 0.04 1.06 0.07 Total Phones/Site (k) 628 471 11 314 21 Site Density (no/km2) 0.66 0.50 0.011 0.33 0.022 76 Annex Six: the Internet Cost-Benefit Model The first two pages of the model list the data required to esti- mate the effects of Internet penetration on users, ISPs and tele- communications companies (C5-C65 and E71 to E83). Much of the data required is self-explanatory, but a few points should be made, and information on sources given. Phone charges The cost of a three-minute call has been used for the sake of convenience. It is the rate published by the ITU. When used as a basis for calculating the cost of a 30-minute or longer call (which is the more likely length of a connection to the Internet), it is obviously going to produce inaccuracies, usually overpricing the cost of a call of that length. In Ghana, for example, the price of a three-minute local call is 200 cedis, the price of a 30-minute lo- cal call is only 1,400 cedis. The cost of a regional call will have to be an average, if there are many different regional call rates. Table A6.1 provides ap- proximations from available data. The cost of an international call would preferably be weighted based on international call destinations. Here, we have taken estimates from available data. We have chosen the U.K. or the U.S. accounting rates to act as an estimate for the average accounting rate charged and received by the African telecommunications company. Table A6.1 Phone Charges for .Selected African Countries, US Dollars Installation Yearly 3-min 3-min 3-min Accounting (business) rental local call regional international rate/min Mozambique 108.0 61.0 0.04 0.60 7.65 1.19 Ghana 52.3 12.0 0.10 0.30 7.50 1.00 Ethiopia 55.8 37.2 0.10 4.80 1.80 Cameroon 54.0 86.4 0.07 1.80 South Africa 74.8 130.8 0.06 0.64 4.58 1.00 Mali 91 .9 40.8 0.15 20.00 0.91 Uganda 127.6 18.0 0.05 11.00 1.00 Zimbabwe 206.4 22.8 0.04 7.50 0.75 77 Economic Toolkit for African Policy Makers Internet Charges Table A6.2 ISP (SLIP/PPP) Flat Rate Service Charges for business in Africa (US dollars) Internet connectivity charges are as- sumed to be a flat fee for full service. Country Connection Yearly Fee A connection charge plus a yearly flat fee provides unlimited Internet Ivory Coast (Africa Online) 18 1,584 and email access for the customer. Ivory Coast (Africom) 53 423 The Internet charges data allow for Djibouti 282 2,373 only one price regime for a country Ghana 100 613 (obviously already false for much of Namibia 0 671 the continent). It would make the Senegal (ENDA) 40 240 model much more complicated if Senegal (Metissacana) 0 1,056 this were allowed to take multiple Sou th Africa 0 168 values. Further, the hypothetical rev- Swaziland 22 761 enue streams are based on a mo- Tanzania 0 4,800 nopoly ISP provider. Togo 0 1,144 Uganda 50 780 Equipment Mozambique 10 600 The figures for the price of a fax machine, computer and modem are estimated from data for Mozambique (fax machine US$400; com-: puter US$1,300; modem US$175). The prices of these goods will clearly vary according to the cost of importing them. Usage statistics In South Africa, the number of users is estimated at 700,000, while Zimbabwe is estimated to have 10,000 users. Information from UEM in Mozambique provides an estimate of approximately 2,000 dial-up links for UEM and a total of 750 dial-up links for the other two ISPs. Because many dial-up links are accessed by more than one user, the total user-base in the country is esti- mated at 5,000 to 6,000 (up from zero in 1995). Estimates for Ghana are around 5,000 users (up from 60 in 1995). Ethiopia has approximately 1,600 users. In order to be consistent, however, we will concentrate on the number of subscribers rather than the num- Table 6A.3 Use of Email in Uganda and Ethiopia: User Survey ber of users. Nearly all users in Mozam- Uganda Ethiopia bique are based in Maputo, and so Sent Received Sent Received have local access to an ISP. Esti- (%) (%) ('Yo) (0/o) mates for Ghana also suggest about 90 percent local access rates. UEM 0-1 per month 4.7 1.6 8.4 18.3 data point to an average access time 1-4 per month 3.1 12.5 15.7 12.2 to the web in Mozambique of about 1-5 per week 43.8 46.9 39.8 50.0 >1 per day 48.4 39.0 36.1 19.5 eight hours per month, or approxi- mately 100 hours per year. In 78 1 I Ghana, figures from NCS suggest that an average of 10 out Table A6.4 of 1,200 subscribers are accessing the Internet at any one Frequency of Use of Email in Uganda: time, suggesting an average usage level of 73 hours a year. Data from Node Average user Number of emails No. of subscribers sent in two months The average user statistics ask for the average number of email sent and received. Looking at data from the CABECA 0-10 73 survey for Uganda and Ethiopia, it is possible to estimate 11-20 32 the average number of email sent by each user. The data 21-30 27 from user surveys suggest fairly high levels (and these fig- 31-40 19 41-50 3 ures exclude the receipt of mass-mailings). 51-60 5 The data suggest that users send somewhat more email 60+ 4 than they receive, and (on average) send over one email a day. There are problems of selection bias with this data, however (it is likely to have excluded those with an ac- count which they do not use). CABECA also asked for node data on email usage which is likely to be more reliable. The two tables, A6.4 and A6.5, enumerate the number of outgoing email per subscriber from the node. The number of email sent over a month for all subscribers, suggests an average of 20 email per month, or about one each business day - a fair estimate for present African usage levels. Annually, this might amount to 240 email. As the number of in- coming email appears to be lower, we might estimate this at about 200. The average length of email in Mozambique can be estimated at about three and a half pages. 1 Next, the model asks for the percentage of these email that are replacing telephone calls and faxes, excluding "new" email traffic. "New" email consists of email that contain information which otherwise would not have been sent by fax or phone. It is assumed that much email traffic will be new, because of the very low marginal cost of sending an email. However, empirical evi- dence suggests that there is also substitution occurring. In all of the countries of the CABECA survey, respondents reported that email had saved money on their communications bud- gets. The CABECA study in Uganda reveals that the method Table A6.5 of communication most commonly replaced by email is sur- Frequency of Use of Email in face and air mail, followed by fax, telephone and telex. Tele- Ethiopia: Data from Node phone and fax service savings were the most significant Number of emails No. of subscribers sent in two months 1 This estimate is, admittedly, back of the envelope. UEM estimated that 60 percent of their outgoing traffic was emails. For the month of February, this amounted to 2.037 mb of emails. UEM has an estimated 2,000 users, who send approximately one email per day, or a total of <30 143 40,000 emails per month. The average length of an email is then 50.93 30-61 16 kb, or about three and a half pages. Senegal's average email length is 61-122 20 estimated by ISPs there to be about 50 kb, although a Namibian ISP >122 27 estimated the average length at just five kb, or under half a page. 79 Economic Toolkit for African Policy Makers source of reductions in communication budgets brought on by email (also mentioned were savings on publications, courier and postal fees and travel). It is fair to assume, then, that 35 percent of email traffic is substituting for either phone or fax. World wide, fax is estimated to account for 40 percent of international phone traffic. This is because fax is considered a more efficient means of communication, transferring informa- tion five times faster than voicer. Given that email is also a sub- stitute for fax, estimates from the CABECA study suggest that about 25 percent of email traffic is substituting for faxes and 10 percent for voice calls. The remaining 65 percent substitutes for other forms of communication (primarily post) or is new traffic. The statistics also ask for the destination of outgoing email. In a user survey conducted in Uganda during the CABECA study, 80 percent of respondents had not yet used email for local ex- change of information, but had done so to communicate with partners abroad. Eighty-two percent of Ethiopian users said that less than 10 percent of email traffic they sent stayed in-country, although in the more Internet-dense country of Zambia the fig- ure was 26 percent. The CABECA study suggests that approxi- mately 56 percent of email from Uganda have destinations outside the continent, 27 percent have destinations in Africa but not within the country, and only 17 percent are for recipients within Uganda. Given that the great majority of users in Africa appear to reside in capital cities, this implies that the percent- age of email with a different region in the same country is very low indeed. All of the above lead to the assumption that, per subscriber, email replaces about 68.9 minutes a year of outgoing interna- tional telephone calls (voice and fax), 4.1 minutes of regional calls, 12.2 minutes of local calls and 57.4 minutes per year of incoming international calls. PTT charges to ISPs We already know that licensing costs are a significant percent- age of revenues for ISPs in some countries. Another major charge is the cost of leased line or VSAT connection to the Internet. The yearly charge is the portion paid to the PTT (any portion paid to foreign telephone companies or other providers should be listed below at C63). In Mozambique, the PTT charges US$250 a month for a 64 kbs leased line to a satellite connection provided by the Leland initiative. Leland charges US$2,000 a month for a con- nection from there to the U.S. Internet backbone (this would be covered by "Internet Backbone connection" charges, C64, with no separate foreign telco charge for 64 kb VSAT /LL, line C61). The major Internet provider, the UEM, pays US$4,982 a month to Transtel, a South African company, for a leased line to South 80 Africa's Internet backbone plus a usage-based charge to UNINET for access. They are willing to pay this premium for a more reli- able service. Their costs will fall, however, if they are able to negotiate a ~iirect connection to the U.S. provided by Lyman Brothers for US$3,000 a month for 128 kbs. This suggests the advances that scale, technology and competition can make in cutting costs and increasing reliability for the Internet. On the spreadsheet, not all points to enter information are necessarily relevant in all cases. In Mozambique, for example, payments to the PTT (C55) cover only the local leased line. For the international connection portion, the Leland Initiative link covers all costs. The charge by Leland includes Internet back- bone access, and so there is only one lump sum to count (thus the zero charge for backbone access). Data for setup costs in- clude leased line setup (C61), so, again, this is not counted as a separate charge. If the local telecommunications company is not providing the leased line service, as in Ghana, the cost to it of provision (C55) should be zero. PTT details Estimations of the cost of a main line in Africa vary greatly be- tween countries. The ITU has found that costs per line vary be- tween US$1,000 and US$8,000 per subscriber line, with an average of around US$4,250. This compares to a world average cost of US$1,500. It should be pointed out that this high cost will largely be the result of two factors: inefficiencies in African telecommunications companies and difficulties due to subscriber density and geography. A recent World Bank study has estimated a reasonable capital cost per line of rural telephones to be US$5,000 in Kenya, US$6,000 in Zimbabwe and US$8,000 in Uganda. Fifty-nine percent, 51 percent and 71 percent of tele- phones are in the largest cities in these countries, however. There- fore, the average cost of a line should be much lower. Nonetheless, in Mozambique, the average cost of a main line has been esti- mated at US$4,500 (estimates for South Africa are closer to US$1,000). The recurrent costs of a telephone line are estimated at about US$600 for South Africa. The cost of installing a second line is - at least on more modern lines - considerably below that of installing the first. For Mozambique, it may be assumed as one-sixth of the cost, or US$500. According to a World bank study in Tanzania, exchange equipment amounted to about 25 percent of the cost of a line, suggesting a cost in Mozambique of about US$1,125. The cost of providing the satellite and/ or leased line has to be calculated on a yearly basis. The cost of leased line provision to the PTT is likely to be significantly below the rates which it charges. Mozambique is charging US$250 per month for local 81 Economic Toolkit for African Policy Makers leased line provision, suggesting cost to the telecommunications company of perhaps US$200 per month. USAID estimates sug- gest costs of US$2,000 a month for international leased line pro- vision. For example, last mile costs will be little different from those of telephone connections. A 1.2 meter ground station with 64 kbs capacity costs US$10,000. Renting space on a satellite and a hub station costs about US$1,750 per month. Mike Jensen esti- mates a monthly cost of operating a 64 kbs VSAT at about US$4,000 to US$5,000 (of which about US$2,000 would go to the satellite company and hub station). There are large economies of scale to international commu- nication. Intelsat charges US$9,975 per month for a 2mps con- nection to an 11 meter dish, or about US$320 for each 64 kbs. This suggests that PTTs should be able to provide one half of a , 64 kbs satellite circuit at a cost to them of US$30,000 a year, with perhaps a further US$2,500 per month to foreign providers - a total of US$60,000 a year. ISP details All of these details are designed to be for a small "bare bones" operation. A small sized ISP is estimated to require a US$12,835 administrative computer with US$1,000 of associated costs (USAID estimates total computer costs for a larger system at US$55,000). Other setup costs are estimated at US$17,800. Con- nection charges to the Internet are mentioned above. Recurrent salary and maintenance are estimated by USAID to be US$21,000, and by Ruth to be US$11,000 for a smaller operation. Assumptions The first assumption is about the number of subscribers to each computer terminal. In the first part of the model, the assump- tion is that each user has his or her own computer. In the second part of the model, it is assumed that subscribers share comput- ers. Evidence from the CABECA study suggests between four to eight users per terminal (or around two to four subscribers). This reduces user cost of access to the Internet. Another assumption considers the amount of time it takes to access an email account (as opposed to using the Web). A three-minute link up every business day could theoretically transmit 91,000 pages of infor- mation each year. Thus, the model assumes that the length of time spent accessing email is determined not by the amount of data transferred, but by the number of times during the day the email account is accessed, and the amount of time taken each time to do so. The CABECA survey suggests that the average African user accesses his or her account at least once a day. The time taken will be brief; here we have assumed three minutes. 82 As for the number of subscribers who are likely to obtain a second line dedicated to web access (measured in terms of ter- minals), in the U.S., this figure is as high as 40 percent of Inter- net users. In Africa, with lower incomes and long waiting times for installation, and between four to eight users per terminal, this figure is likely to be far lower. We assume here that five percent of subscribers manage to obtain a second line for Inter- net access. It is assumed that an increased number of subscribers will increase the number of telephone lines, computers, size of leased line/VSAT connections and user support personnel required by the ISP. UEM has five technicians working for an estimated 5,000 users. In Uganda, the charge to users for training is US$8 ini- tially with access to free courses every week. This suggests an annual per subscriber cost of perhaps US$10 to cover expenses. In the U.S., ISPs tend to have one phone line for 10 to 15 sub- scribers. NCS in Ghana has 100 dial-in lines for 2,500 subscrib- ers. Here we have assumed one phone line for every 20 subscribers. USAID estimates suggest that a new computer will be needed for 1,000 subscribers. UEM is just now considering upgrading to a 128kbs link, at a level of approximately 2,000 subscribers, suggesting that at the moment 64kbs in Africa can provide the minimally acceptable level of access for around 3,000 users. As the comfortable capacity of a 64kbs line is only about 10 active Internet users at any one time (and figures for Mozambique already suggest that the average number of sub- scribers on line at any one time must be about 15), providing reasonably fast access even at this low level of usage will re- quire greater bandwidth - perhaps 64kbs per 1,000 users. It is also assumed that new web subscribers put more pres- sti.re on the telephone network, demanding extra switches. This is a matter of some debate in the U.S.: we have high and low estimates (0.1 and O) for the number of new switches required per subscriber, and the figure is likely to be toward the low end in Africa because of low levels of usage. Here we assume 0.01. Finally, there are assumptions about the real cost of capital for telecommunications companies, ISPs and users, and depre- ciation rates for telecommunications equipment, computers and other items. In order to provide figures on savings and costs per year which include capital expenditures and depreciation rates, all capital expenditures and one-off payments on the debit side and all one-off receipts on the credit side have been converted into yearly costs using the formulae for perpetuities and annu- ities. YRC (yearly return to capital) rates equal the cost of capi- tal for goods with long lives (E82 to E83) and the annuitized cost of capital for goods with shorter lives (calculated E84 to E86)). The model assumes that telecommunications companies have access to cheaper capital than ISPs and Internet users. It 83 Economic Toolkit for African Policy Makers also assumes differing replacement rates for telecommunica- tions, computer and other equipment. Based in part on USAID data, we have estimated a four-year replacement period for com- puter equipment and 10 years for other items, with a real cost of capital at 10 percent for telecommunications companies and 20 percent for· ISPs (in nominal terms, NCS Ghana pays 25 to 45 percent per year). Equations for the model Subscriber savings and telco revenues per page The first section of the model estimates the costs to the subscriber and revenue streams to the telephone company based on send- ing a number of pages of information by voice, fax, locally con- nected email and regionally connected email. Across the top of the section are the number of pages sent locally (G7 to J7), re- gionally (L7 to N7) and internationally (H26 to J26), down the side the different methods of sending information, and the rev- enues received from the user by the PTT. In the case of interna- tional communication, there is a further section (F42 to J47) on the amount of revenue received by the PTT from overseas users contacting the home user utilizing different forms of communi- cation (any telephone-based communication that is received from abroad is subject to accounting-rate charges by the home PTT). The local section includes the fixed costs of operating dif- ferent communications systems (ignoring the rental and instal- lation charges for a telephone line, assumed already there), along with the extra costs associated with local calling charges. Under the 0 Pages sent/yr column (G9 to G20), the fixed costs of the various communications methods are calculated. For voice com- munication (G9), the additional fixed cost (above the rental and installation charges of the 'phone) is zero. For the fax (GlO), the additional cost per year (over the price of a basic phone connec- tion) is the cost of a fax spread over its depreciation period. For email (Gll and G12), the additional cost per year is the annuitized or perpetuitized sum of connection fees and com- puter and modem costs and the yearly web connection charge. Given that computers are useful for far more than the Internet, this can be argued to over-estimate the equipment costs that should be associated with user access to the Internet alone. The additional cost per local page sent by phone (HlO to J10) is the fixed cost plus the number of pages sent multiplied by the time taken to 'speak' one page in turn multiplied by the unit cost of calling time. The additional cost per local page sent by fax (Hll to Jll) is the number of pages sent multiplied by the time taken to fax them in turn multiplied by the unit cost of calling time. The additional cost per local page sent by local email 84 (H12 to Jl2) is fixed. It is the local phone call cost of the daily connection time times the number of business days in a year (260). The additional cost per local page sent by regional email (i.e. the user has to make a regional call to access his email) is calculated in the same way (H13 to Jl3), but substituting the cost of a regional call for the cost of a local call in the daily con- nection time. There is no additional cost to users for sending email to regional or international recipients (L19 to N20 and H31 to J32). For,voice and fax, the extra cost is calculated as above. For voice (L17 to N17 and H29 to J29), then, this is number of pages mul- tiplied by time taken to speak a page multiplied by the unit cost of a regional or international call. For fax (L18 to N18 and H30 to J30), this is number of pages multiplied by time taken to speak (fax) a page multiplied by the unit cost of a regional or interna- tional call. The phone company receives no revenue from fixed costs of communication (beyond the installation and rental of the phone line, which is excluded from the calculation). Revenue from local (G17 to J17) and regional (L17 to M17) voice calls is based purely on the length of connection time and the charges for calls made. This is the same with local (G18 to J18) and re- gional (L18 to ¥18) faxes. In the case of international calls, the picture is complicated a little by accounting charges. The phone company has to pass on half of the accounting rate value of the call to the telco that provides the other half of the circuit. For voice (H29 to J29)and fax (H30 to J30), then, the calculation is the time taken to speak (fax) the number of pages multiplied by the unit cost of the call minus the time taken to speak (fax) the number of pages multi- plied by half of the accounting rate paid to the foreign telecom- munications company. The telecommunications company also receives accounting rate payments from foreign telecommunications companies originating incoming voice and fax communication with the home user. For voice (H45 to J45) and fax (H46 to J46), this amounts to the time taken to speak (send) the number of pages multiplied by one half of the accounting rate. For email, the amount that the telecommunications com- pany receives from the user is fixed, as we saw above. For local access email (H12 to J12), and regional access email (H13 to Jl3), · the receipts equal the by minute cost of the call multiplied by the time taken each day to access email multiplied by the num- ber of business days in a year (260). This is the only incremental cost of email, whatever the destination or source of communi- cation. Therefore, the extra costs associated with regional or in- ternational emailing is zero (L19 to N20, H38 to J39 and H37 to J37). 85 Economic Toolkit for African Policy Makers Full impact on telecommunications companies, ISPs and subscribers The second part of the model looks at the detailed effects on telecommunications and ISP revenue and user costs and sav- ings of some fixed level of Internet usage. The ISP is assumed to be a monopoly provider, to simplify the model. This is a clear over-simplification. Already, much of Africa has multiple Inter- net service providers. The effect on the model of this simplifica- tion is to overestimate ISP profits and reduce telecommunications company revenue from ISPs (because of multiple leased lines, more phone lines and more licenses). This will also slightly over- estimate telecommunications company revenue from users, be- cause it is likely that, with more ISPs, a larger percentage of users can make a local call to connect to the Internet. The section is divided in two. On the left (F49 to G93) is a static illustration of revenue gains and losses for Telecommuni- cations companies and ISPs and summed costs and savings for users, based on the number of Internet users entered at C24 ("number of Internet users"). The second section is a model of the potential impact of the Internet on ISPs, telecommunications companies and users given different numbers of Internet users. Telecommunications company revenue gained Starting on the left, the first calculation is of telecommunica- tions revenue gained from the present number of Internet users. Telecommunications companies will gain revenue from second Internet user lines. This will be based on the number of new lines (number of users multiplied the percentage who are esti- mated to take second lines) multiplied by the revenue (rental and installation) that these lines provide (G51). Telecommunications companies will also receive revenues from expanded Internet traffic. The (smaller) part of this will come from users accessing email. As above, this is assumed to be a fixed length call every day. The first part of the equation calculates the cost of calls for accessing email for local and then regional users based on the number of users, the amount of time taken to access email, the number of business days and the cost of a local (regional) call of the duration inputted in E71 ("Email is accessed every business day for X minutes"). The second part of the equation calculates the cost of calls involved in accessing the web - average number of hours users access the Web (in- putted in C30), multiplied by the cost of a local (regional) hour- long call. Telecommunications companies are also assumed to make money from licensing ISPs. As the model assumes a monopoly provider, this equals the input from C44 ("License"). Telecom- 86 munications companies will also receive revenue from the rental costs of phone lines into ISPs. This is based on the assumption of the number of subscribers that can be handled per line (E75) multiplied by the revenue (rental and installation) that these lines provide (G55). It is assumed that the telecommunications company pro- vides the ISP's international connection. No matter how many web subscribers, it is assumed that email alone will not require the provision of an expanded leased line. The theoretical maxi- mum data that could be transmitted by a 64 kbps line over one year is equivalent to 210 million minutes of voice communica- tions (total outgoing telephone traffic from Mozambique totaled 16.4m minutes in 1995). This suggests that email (without the web) does not require large capacity (further evidence is that the NSFNET backbone in the US was only 56kbps as recently as five years ago). It is the increasing use of the web that will drive the need for ever-larger international Internet access capacity. As above, the equation (G55) is based on the assumption of the number of subscribers that can be handled per 64kbs of VSAT or leased line connection (E77) multiplied by the charge for that line. The model assumes a flat fee for each 64kbs of connection to the backbone by the telecommunications company and for- eign providers. This is an over-simplification - telecommuni- cations companies will charge less for a 128kbs connection than they would for two 64 connections. Telecommunications company revenue lost In calculating telecommunications revenue lost, we have as- sumed that there is no substitution between WWW access and phone or voice usage, the only substitution is via email. This is an obvious oversimplification. Accessing websites provides an alternative way of obtaining data, which might well substitute for a voice or fax communication to find the information. How- ever, it is very difficult to estimate what percentage of web us- age time is substituting for voice or fax, and it is even more difficult to calculate a "conversion rate" between Web down- loads and fax or voice. It is, for example, questionable that a Website that uses frames and moving graphics imparts far more valuable information than a site created without these features. The difference in terms of data transfer can be immense, how- ever. The effect of ignoring Web substitution while fully account- ing for Web use will be to underestimate the revenue losses on telecommunications companies of Internet access. Looking first at substitution of outgoing international calls (G60) this equation calculates the amount of international email traffic that is substituting for international outgoing voice traf- fic. First, the number of "pages" of email sent internationally in 87 Economic Toolkit for African Policy Makers place of a telephone call is calculated. This is equal to the sum of the proportion of email that replace phone calls multiplied by the proportion of email going to an international destination in turn multiplied by the average number of email sent per year, their average length and the number of users. This number of pages estimate is multiplied by the revenue generated by one "page" of an international voice call -the cost of a call long enough to "speak" a page minus the accounting rate payment. The international faxes out (G61) calculation follows the same procedure as above, only substituting the percentage of information sent by fax for the percentage of information sent by telephone, and using the number of kb in a three minute fax (144 kb) for the amount of information in a three minute voice call. The international calls and faxes in (G62 to G63) calculates accounting rate losses, assuming that international email com- munication into a country has the same percentage substitution rate for fax and voice (which might be an over-simplification). This calculation is a repeat of the ones above, except that it cal- culates the accounting rate charges lost by the home telecom- munications company based on the amount of incoming email. The amount of incoming email is calculated by subtracting the average number of email sent locally and regionally from the average number of email received (each email sent locally or regionally must be received locally or regionally, so any residual from this calculation must be email received from abroad). The rest of the equations follow the pattern set above, with the replacement of international data with regional and local data (C23 to G27 and G35 to G38) and the removal of the ac- counting rate adjustments. International ISP Next, international ISP costs and revenues are calculated (G77). The first equation calculates equipment costs. This includes num- ber of computers and the size of the VSAT I leased line connec- tion required to service the number of users (G76 to G77). The cost of administrative computers, fixed other computer costs and setup costs for the VSAT /LL connection (C59-C61) is multiplied by the yearly return requirement for computer equipment while other setup costs (assumed fixed) are multiplied by the return requirement for other equipment. Internet and telecom charges (G78) include the cost of the yearly license fee, the backbone connection, the charge for the number of phone lines the ISP needs, the home charge for the VSAT or leased line connection required and the foreign coun- try charge for the VSAT or leased line connection required. Sal- ary and maintenance charges (G79) includes recurrent costs and maintenance costs (C65) and a calculation of the cost for user 88 support multiplied by the number of users (E74 and C24). User costs, savings User costs and savings are calculated for the total user base. Equipment costs (G87) amount to the yearly return cost value of the amount spent on modems and computers. Internet access (G88) paid to ISPs is thus the annualized cost of initial connec- tion multiplied by the yearly fee. The charge for the "fixed costs" of web access and accessing email (G89) include the installation and rental of second lines and the calculation of regional and local call charges for 260 days of email access, as above, multi- plied by the number of users. Web telephone charges (G90) amount to the cost of a local/ regional three minute call multi- plied by twenty (to give an approximate cost of an hour-long call) multiplied by the number of hours on line, calculated for local and then regional users. For the user, telephone savings equal the revenue lost to the telecommunications company ex- cluding the calculations made for accounting charges (G60 to G61 minus the accounting rate adjustment and G64 to G67). The final part of the model calculates hypothetical cost and revenue streams from expanded Internet access (J51 to N75). These calculations are largely based on those above. We have assumed that future users will utilize the Internet as much as the first users on whom email substitution rates and Web access rates are based. Given that we would expect the first Internet users to be those with the most to gain in communications sav- ings and Web access, this is an oversimplification. Later users will be those who make fewer international calls and have less interest in accessing the Web. These estimates exaggerate ISP equipment needs, telecom income losses from substitution and new equipment needs, telecom gains from time spent on the Web and user costs and benefits. Given the fixed costs of email, this will mean that user benefits from substitution are overestimated. As ISP revenues are (in this model) independent of usage, their profits will be higher. The negative effect of email substitution will be lower for telecommunications companies. The revenues from web surfing will be somewhat lower as will the costs of providing new equipment. Under usual circumstances, this sug- gests that the model will overestimate the negative effects of an expanded Internet on telecommunications companies. ISP, hypothetical cost, revenue ISP costs, as above, have a fixed component (K53). This consists of a license, phone installation and rental, basic computer equip- ment, other setup costs, fixed setup costs for the leased line/ VSAT, local telecommunications company charges for the VSAT I 89 Economic Toolkit for African Policy Makers LL link, foreign charges for the VSAT I LL link and the cost of connection to the Internet backbone. As more users are added, costs go up (L53 to N53). New subscribers require more user support, more telephones, more computers and a larger VSAT / LL connection. ISP revenues (K54 to N54), as above, are a fixed connection fee and a fixed yearly fee multiplied by the number of users. Telecommunications company income, revenue The telecommunications company revenue calculation (K62 to N62) is made as above ((G51 to G55), taking account of new phone lines installed at the ISP and by users, ISP licenses, VSAT I LL charges and user telephone call charges. The cost to telecom- munications companies (K63 to N63) is based on a new formula (above, we only calculated revenues). The calculation is based on the cost of installing and operating the telecommunications equipment involved in servicing ISPs and user. These costs in turn are: the cost of running the VSAT or leased line, the initial user line into the ISP, the cost of subsequent lines into the ISP and out of the users who order a second line, the cost of ex- panding VSAT connections or leased line connections and the cost of providing extra switch and transmit equipment to handle the longer calls made to access the Internet. The total of present actual phone company revenue lost from users switching to email (K65 to N65) calculated is scaled up from the earlier rev- enue lost calculation (G60 to G67) by the number of hypotheti- cal users over present day users in the model, to give a straight-line estimate of revenue lost to email substitution. This is an over-simplification - as the number of email users in the country increases, the percentage of email with a local destina- tion will increase. A straight-line estimate will thus over-esti- mate the scale of substitution losses to telecommunications companies. User costs, revenue User costs (K73 to N73) are calculated as above (at G87 to G90). User substitution (K75 to N75) is calculated as above (at G91). This does not include all user benefits. Unaccounted for are the benefits of email that is not substituting for fax or voice (i.e. the majority of email traffic) and the benefit of web access. Obvi- ously, there is a large consumer surplus here. 90 Calculations Assumptions E82-83: To calculate the yearly return to capital required to pay the cost of capital (for goods for which do not need to be re- placed) and depreciation (for goods that will have to be re- placed), the following equations were used: The formula for a perpetuity is PV= C/ r, where PV is the present value, C the yearly cash payment and r the real cost of capital. The yearly cash pay- ment required to cover a present value expenditure is C= PV*r. For one off payments for goods which have a very long life (for example, telephone connection for the recipient) or one-off re- ceipts (for example, the payment for initial telephone connec- tion for the telephone company), this formula was used to calculate the effective yearly cash cost or income of the present payment or receipt. Thus, the YRC (yearly return to capital) rate is equal to r, the cost of capital. E84 to E86: The formula for an annuity is PV= C[l Ir -1 I r(l + r)At], where PV is the present value, C the yearly cash pay- ment, r the real cost of capital and t the number of years over which the annuity is paid. The yearly cash payment required to cover a present value expenditure for a good that will need to be replaced int years is C= PV /[1/r - 1/r(l + r)At]. For one-off purchases of goods that have a short life (for instance, comput- ers), this formula was used to calculate the effective yearly cash cost or income of the present purchase. Thus, the YRC (yearly return to capital) rate is equal to [1 I cost of capital - 1 /(cost of capital* ((1 + cost of capital) I\ number of years depreciation))] User savings and telecommunications company revenues per page Fixed cost to user per page GlO: 0 GH: (Cost of fax machine)* (yearly return requirement for user computer equipment) 2 G12: ((initial connection fee)* (user cost of capital))+ ((cost of computer plus modem) * (yearly return requirement for user computer equipment))+ (yearly web connection fee) 2 As calculated above, taking into account depreciation and the cost of capital. 91 Economic Toolkit for African Policy Makers Local cost to user per page: (HlO to J10): (fixed cost) + (number of pages sent) * (cost of a three minute local call) / 2 3 H11 to J11: (fixed cost)+ (number of pages sent)* (cost of a three minute local call) / 104 H12 to J12: (fixed cost)+ (cost of a three minute local call)* (num- ber of minutes each business day that email is accessed I 3) * 260 5 H13 to J13: (fixed cost) +(cost of a three minute regional call)* (number of minutes each business day that email is accessed I 3) * 260 Regional and international cost to user per page (voice/fax) L17 to N17 and H29 to J29: (number of pages sent)* (cost of a three minute regional/ international call) I 2 L18 and to N18 and H30 to J30: (number of pages sent)* (cost of a three minute regional/ international call) I 10 Regional and international cost to user per page (Email) L12 to N13 and H31 to J32: 0 Fixed revenue to PTT from Email H12 to J12: (cost of a three minute local call)* (number of min- utes each business day that email is accessed I 3) * 260 H13 to J13: (cost of a three minute regional call) * (number of minutes each business day that email is accessed / 3) * 260 Revenue to PTT per page local/regional (voice/fax) G17 to J17 and L17 to M17: (number of pages sent)* (cost of a three minute local/ regional/ call) I 2 G18 to J18 and L18 to M18: (number of pages sent)* (cost of a 3 From above, a user can "speak" two pages in a three-minute call. 4 From above, a user can.fax 10 pages in a three-minute call. · 5 From above, it is assumed that there is no additional cost per number of pages sent by email; it is the fixed daily time cost of accessing that accounts for the number of minutes on line. This total is multiplied by 260 (the approximate number of business days) to calculate a yearly total for access. 92 three minute local/ regional call) / 10 Revenue to PTT per page from home user, international (voice/fax) H29 to J29: (number of pages sent * cost of a three minute re- gional/ international call) I 2 - (one minute accounting rate* 3 *number of pages sent) / 4) 6 H30 to J30: ((number of pages sent) * (cost of a three minute regional/ international call) / 2) - (one minute accounting rate* 3 *number of pages sent) / 20) 7 Revenue to PTT per page from overseas sender H45 to J45: (one minute accounting rate* 3 *number of pages sent) I 4 H46 to J46: (one minute accounting rate * 3 * number of pages sent) I 20 ·1 Revenue to PTT from regional and international Email L19 to N20, H38 to J39 and H37 to J37: 0 Telecommunications company revenue gained G51: (ROUNDUP( average no. of second lines per user))* (num- ber of web users)* ((yearly rental) +(installation*real cost of capi- tal for telco)) G52: ((proportion of users with local access*number of web users*cost of 3 minute local call*length of email access each day(in minutes)/3*260)+((1-proportion of users with local access)*number of web users*cost of 3 minute regional call*length of email access each day /3*260)+( proportion of us- ers with local access *average user hours of web access*number of users*20*cost of three minute local call+(l- proportion of us- ers with local access)*average user hours of web access*number of users*20*cost of three minute regional call)) 8 6 The accounting rate calculation is divided by four rather than by two, because the telco only passes on half of the accounting rate to its interna- tional partner. 7 The accounting rate calculation is divided by 20 rather than by 10, because the telco only passes on half of the accounting rate to its international partner. 8 The cost of a three-minute call is multiplied by 30 in this calculation to give a cost per hour of the telephone call. 93 Economic Toolkit for African Policy Makers G53: (ISP licence) G54: (((ROUNDUP((Number of users/ISP subscribers per line))))*(phone installation*telco cost of capital+phone yearly rental)) 9 G55: ((ROUNDUP((Number of users/ISP subscribers per 64 kbs of VSAT /LL connection))))*(charge for 64 kbs of VSAT /LL con- nection) Telco revenue lost G60: (((cost of three minute international call-1.5*accounting rate)*0.5)*(%email replacing phone/100)*(%email with interna- tional destination/lOO)*average no. email sent/yr*average length of email*no. of users) G61: (((cost of three minute international call -1.5* accounting rate)*O.l)*( % email replacing fax/100)*( %email with interna- tional destination /100)* average no. email sent/yr*average J length of email*no. of users) )' G62: (((1.5* accounting rate)*0.5)*( %email replacing phone I 100)*( average no. of email received/year-(( average no. email sent/yr*(% email local destination +%email regional destina- tion)/100)))* average length of email*no. of users) G63: (((1.5* accounting rate)*O.l)*( % email replacing fax I 100)*( average no. of email received/year -((average no. email sent/ yr *( % email local destination+ %email regional destination) I 100)))* average length of email*no. of users) G64: (((regional call cost)*0.5)*( %email replacing phone /100)*( %email regional destination /100)* average no. email sent/ yr*average length of email*no. of users) G65: (((regional call cost)*0.1)*( % email replacing fax /100)*( %email regional destination /100)* average no. email sent/ yr* average length of email*no. of users) G66: (((local call cost)*0.5)*( %email replacing phone /100)*( % email local destination I 100)* average no. email sent/ yr* average length of email*no. of users) 9 The "Round Up" command rounds up the result in brackets follow- ing tothe nearest whole number (since it is not possible to have half a tele- phone). 94 G67: (((local call cost)*0.1)*( % email replacing fax /100)*( % email local destination I 100)* average no. email sent/ yr* average length of email*no. of users) International ISP G77: (((ROUNDUP((Number of users/ISP subscribers per computer)))*ISP administrative computer costs+ISP other basic computer costs)* yearly return requirement for user computer equipment+ ISP other setup costs* yearly return requirement for user other equipment+(ISP setup for VSAT /LL)* yearly return requirement for user computer equipment) 10 G78: (ISP licence+ ISP backbone connection+(((ROUNDUP((Number of users/ISP subscribers per line))))*(phone installation *user cost of capital+phone yearly rental)) + ((ROUNDUP((Number of users/ISP subscribers per 64 kbs of VSAT /LL connection))))* (charge for 64 kbs of VSAT I LL connection+ Foreign charges for 64 kbs of VSAT /LL connec- tion)) G79: (ISP recurrent salary and maintenance+(ISP support per subscriber*Number of users)) User costs, savings G87: ((cost of computer+cost of modem)* yearly return require- ment for user computer equipment*(Number of users I number of users for each station)) G88: (Number of users*(web connection*user cost of capital+web yearly fee)) G89: (G51 -user new lines) G90: (G52 - user extra traffic) ISP hypothetical costs, revenue K53: ((ISP licence+phone installation*user cost of capital+phone yearly rental)+((ISP administrative computer costs+ISP other basic computer costs)* yearly return requirement for user com- puter equipment+ISP other setup costs* yearly return require- 10 As above, the Round Up command rounds up the number of computers and VSAT /leased line connections required. to service the number of users assumed to a whole number. 95 Economic Toolkit for African Policy Makers ment for user other equipment+(ISP setup for VSAT /LL* yearly return requirement for user computer equipment)+charge for 64 kbs of VSAT /LL connection+ Foreign charges for 64 kbs of VSAT /LL connection+ISP recurrent salary and maintenance)+ ISP backbone connection) L53 to N53: (ISP support per subscriber*no. users+((ROUNDUP((no. users /ISP subscribers per line))- l))*(phone installation*user cost of capital+phone yearly rental)+((ROUNDUP((no. users/ISP subscribers per computer))- l))*(ISP administrative computer costs* yearly return require- ment for user computer equipment)+((ROUNDUP((no. users/ ISP subscribers per 64 kbs of VSAT /LL connection))-l))*(For- eign charges for VSAT /LL+charge for VSAT /LL)) K54 to N54: (no. users*(web connection*user cost of capital+web yearly fee)) Telecommunications company, income, revenue K62 to N62: ((phone installation*telco cost of capital)+(charge for VSAT /LL+ ISP licence+phone yearly rental)) + ((ROUNDUP((no. users/ISP subscribers per line)))-l)*(phone installation*telco cost of capital+phone yearly rental)+((ROUNDUP((no. users/ISP subscribers per VSAT / LL)))-l)*(charge for VSAT /LL)+((ROUNDUP((no. users*Average second line per user)))-l)*(phone installation*telco cost of capital+phone yearly rental)+(Users with local access*no. us- ers*(3 min local call*number of minutes each business day that email is accessed I 3*260+user web access yr*3 min local call*20))+((1-Users with local access)*no. users*(3 min regional call*number of minutes each business day that email is accessed /3*260+user web access yr*3 min regional cal1*20))) K63 to N63: (Telco cost of VSAT /LL)+((Telco cost of main line, initial* yearly return requirement for telco equipment+ Telco cost of main line, recurring)) +(Telco cost of second main line, ini- tial* yearly return requirement for telco equipment+ Telco cost of second main line, recurring)*((ROUNDUP((no. users /ISP sub- scribers per line))-l)+(ROUNDUP((no. users*Average second line per user))-l))+(ROUNDUP((no. users/ISP subscribers per 64 kbs of VSAT /LL))-l)*Telco cost of 64 kbs of VSAT / LL+((ROUNDUP((no. users*no. of switches for each user))- l)*(Telco cost of switch and transmit, initial* yearly return re- quirement for telco equipment+Telco cost of switch and transmit, recurring))) K65 to N65: ((telco revenue lost in fixed model*no. users in hy- pothetical model)/Number of users in fixed model) 96 Users, costs, benefits K73 to N73: (hypothetical no. users*(((cost of computer+cost of modem) I number of users per terminal))* yearly return require- ment for user computer equipment+web connection*user cost of capital +web yearly fee+telco revenue from user extra traffic in fixed model/Number of users in fixed model+user new lines in fixed model/Number of users in fxed model)) K75 to K95: ((savings in fixed model*no. users in hypothetical model)/Number of users in fixed model) 97 I I !. I