February 2022 Using Geospatial Analysis to Overhaul Connectivity Policies How to Expand Mobile Internet Coverage and Adoption in Sub-Saharan Africa Using Geospatial Analysis to Overhaul Connectivity Policies © 2022 International Bank for Reconstruction and not have been possible, as well as government representatives Development / The World Bank from the National Communications Authority of Ghana, Rwanda Utilities Regulatory Authority, Rwanda Ministry of 1818 H Street NW Information Communication Technology and Innovation, Washington DC 20433 Ministry of Information and Communication of Sierra Leone, Telephone: 202-473-1000 Universal Access Development Fund of Sierra Leone, Federal Internet: www.worldbank.org Ministry of Communications and Digital Economy of Nigeria, This work is a product of the staff of the World Bank with and Broadband Implementation Steering Committee of external contributions. The findings, interpretations, and Nigeria for their comments on the underlying country reports. conclusions expressed in this work do not necessarily reflect Vivien Foster (Chief Economist, Infrastructure Vice the views of the World Bank, its Board of Executive Directors, Presidency), Albert Zeufack (Chief Economist, Africa Vice or the governments they represent. Presidency), Mark Dutz (Lead Economist, Africa Chief The World Bank does not guarantee the accuracy, Economist Office), Izak Atiyas (senior consultant), Tim Kelly completeness, or currency of the data included in this work (Lead Digital Specialist), Rajendra Singh (Senior Digital and does not assume responsibility for any errors, omissions, Development Specialist)—all from the World Bank—and or discrepancies in the information, or liability with respect Guillaume Xavier Touchard (Investment Officer, Global to the use of or failure to use the information, methods, Upstream Unit, IFC) provided essential comments for the processes, or conclusions set forth. The boundaries, colors, preparation of the report. The Digital Development regional denominations, and other information shown on any map team, Xavier Decoster (Senior Digital Development Specialist), in this work do not imply any judgment on the part of the Rita Oulai (Digital Development Specialist), Kaoru Kimura World Bank concerning the legal status of any territory or the (Senior Digital Development Specialist), Marc Jean Yves endorsement or acceptance of such boundaries. (Lead Digital Development Specialist, Program Leader), Isabella Hayward (Digital Development Specialist), Wilson Nothing herein shall constitute or be construed or considered Muyenzi (Extended Term Consultant), Ida S Mboob (Senior to be a limitation upon or waiver of the privileges and Digital Development Specialist), Doyle Gallegos, Lead Digital immunities of the World Bank, all of which are specifically Specialist, Hyea Won Lee (Junior Professional Officer), and reserved. Tim Kelly (Lead Digital Specialist) provided comments and Rights and Permissions support for the preparation of background country studies. The material in this work is subject to copyright. Because Estefania Vergara Cobos (Economist) contributed to the the World Bank encourages dissemination of its knowledge, review and preparation of this report and background studies. this work may be reproduced, in whole or in part, for The preparation of the report was guided by Isabel Neto, noncommercial purposes as long as full attribution to this Michel Rogy, and Mark Williams, Practice Managers from work is given. the World Bank’s Digital Development Global Practice, under Any queries on rights and licenses, including subsidiary rights, the general leadership of Nicole Klingen (Acting Global should be addressed to World Bank Publications, the World Director, Digital Development Global Practice) and Boutheina Bank Group, 1818 H Street NW, Washington, DC 20433, USA; Guermazi, former Global Director, Digital Development Global fax: 202-522-2625; e-mail: pubrights@worldbank.org Practice and current Director, Regional Integration Africa, Acknowledgements Middle East and North Africa. This report was prepared as part of the studies on GSMA broadband connectivity that will inform the Africa Digital The GSMA is a global organisation unifying the mobile Transformation Report in support to the Digital Economy ecosystem to discover, develop and deliver innovation for Africa Initiative.* Tania Begazo (Senior Economist, Digital foundational to positive business environments and societal Development, World Bank) coordinated the preparation of change. Our vision is to unlock the full power of connectivity the study as part of the Broadband Access for All Global so that people, industry, and society thrive. Representing Solution Group and Digital Infrastructure Initiative (led by mobile operators and organisations across the mobile Doyle Gallegos, Lead Digital Specialist, Digital Development). ecosystem and adjacent industries, the GSMA delivers for its The underlying country studies and substantial sections of members across three broad pillars: Connectivity for Good, this report were prepared by GSMA and GSMA Intelligence Industry Services and Solutions, and Outreach. This activity team: Genaro Cruz (GSMA), Kalvin Bahia (GSMA Intelligence), includes advancing policy, tackling today’s biggest societal Federico Agnoletto (GSMA Intelligence), Robert Wyrzykowski challenges, underpinning the technology and interoperability (GSMA), Pau Castells (GSMA Intelligence), Bryce Hartley that make mobile work, and providing the world’s largest (GSMA), and Caroline Butler (GSMA Intelligence). platform to convene the mobile ecosystem at the MWC and Key inputs on network characteristics used in this report M360 series of events. We invite you to find out more at were provided by GSMA and funded by the UK’s Foreign, gsma.com Commonwealth and Development Office and by the Swedish GSMA Intelligence International Development Cooperation Agency (SIDA); GSMA Intelligence is the definitive source of global however, the views expressed in this report do not necessarily mobile operator data, analysis and forecasts, and publisher reflect the UK or Swedish governments’ views or official of authoritative industry reports and research. Our data policies. We would also like to thank mobile operators in the covers every operator group, network and MVNO in every seven countries covered by this study for providing underlying country worldwide — from Afghanistan to Zimbabwe. It is data on mobile infrastructure without which the study would the most accurate and complete set of industry metrics available, comprising tens of millions of individual data * For more information on the Digital Economy for Africa initiative, points, updated daily. For more information, please visit please visit https://www.worldbank.org/en/programs/all-africa-digital- www.gsmaintelligence.com. transformation Using Geospatial Analysis to Overhaul Connectivity Policies Contents Executive Summary 2 Definitions of Terms Used 6 1 Introduction 8 2 Connectivity has Been Delivered by Market Forces 16 3 Innovation is Key to Unlocking Rural Markets and Boosting the 28 Impact of Policy Change 4 Enabling Policies Can Increase Network Coverage and Adoption 34 5 Public Investment is Needed to Achieve universal Connectivity 46 6 Conclusions 52 Appendix A: Methodology 54 Appendix B: Acronyms and Abbreviations 61 2 Executive Summary Using Geospatial Analysis to Overhaul Connectivity Policies Executive Summary Increasing access to the internet is one of the great challenges of our time and has grown in importance since the outbreak of the COVID-19 pandemic. Around half the world’s population remains online, while in Sub-Saharan Africa just over a quarter of the population uses the internet. The region also accounts for almost half of the 450 million people around the world who do not live in areas covered by 3G or 4G mobile networks. Connectivity gaps are a consequence of fundamental economic challenges around supply and demand. In a market-led environment, mobile operators will provide coverage where there is existing or expected demand for connectivity. Expanding mobile broadband coverage will partly depend on lowering costs and investments risks, but the main driver will be enhanced demand for connectivity services. Mobile technology is particularly important to drive connectivity forward in Africa, as it accounts for more than 98 percent of broadband connections.1 1 Source: International Telecommunication Union (ITU). Using Geospatial Analysis to Overhaul Connectivity Policies Executive Summary 3 Ghana Nigeria Sierra Leone Benin DRC Rwanda Tanzania To gain a better understanding of the policies and interventions needed to accelerate connectivity in Sub-Saharan Africa, this study focuses on mobile connectivity, although it recognizes the importance of enabling the use of various broadband technologies that respond to demand. This study has two main objectives: • To map mobile coverage and adoption in Africa at the highest possible resolution (that is, at the settlement level). This was done across seven countries: Benin, Democratic Republic of Congo, Ghana, Nigeria, Rwanda, Sierra Leone, and Tanzania. Together, these account for 40 percent of the population in Sub- Saharan Africa. They also reflect the diversity of the continent, ranging from the largest countries in terms of both population (Nigeria) and land area (Democratic Republic of Congo) to smaller countries in Western Africa (Benin, Ghana, and Sierra Leone) and Eastern Africa (Rwanda and Tanzania). • To simulate the effects of different policies using granular data on both the location of infrastructure and demand for mobile and internet services. This enables more precise calculations and therefore a deeper understanding of the impacts policy reforms can have on coverage and adoption, as well as the additional investment needed to achieve universal connectivity by 2030. 4 Executive Summary Using Geospatial Analysis to Overhaul Connectivity Policies Key Findings • Mobile operators are very close to the ‘market operators have access to sufficient and affordable frontier’ for 2G coverage, with at least 87 spectrum in sub-1 GHz bands, including the percent of the population covered across the refarming of the existing spectrum so that it is seven countries. There is a limited amount of technology neutral. Across four countries where additional 2G coverage that can be provided by not all operators have access to sub-1 GHz the private sector that is financially viable and spectrum for 4G (in Ghana, Nigeria, Sierra Leone, sustainable in current market conditions. Almost and Tanzania), an operator could increase rural all the uncovered areas are in rural, often remote, 4G coverage by more than 5 percentage points (or locations. 7.5 million people) if it could use 700 or 800 MHz spectrum for 4G (assuming existing spectrum fees • Extending mobile broadband to areas with are applied). no coverage presents a substantial economic challenge. It will require efforts to reduce • Infrastructure sharing at the site level deployment costs and, more importantly, increase would enable coverage to expand while demand. Both are contingent on continued, maintaining service-level competition. Active collaborative action by all stakeholders, building sharing would allow 2–10 percent of the rural on private sector innovation, which in recent years population (depending on the country) to have has driven significant cost reductions in rural mobile broadband coverage from more than one network deployments, as well as in handset and operator. Policy makers that are considering data prices. single wholesale networks (SWNs) should also consider that active sharing can deliver similar • While 3G and 4G coverage are lagging at 74 levels of coverage to SWNs while maintaining a percent and 48 percent of the population, greater degree of service competition (so long as respectively, they could catch up with 2G competition safeguards are in place). coverage in the coming years if spectrum management is updated, for example, to ensure Using Geospatial Analysis to Overhaul Connectivity Policies Executive Summary 5 • Aligning tax policy with best-practice principles 5 percent in every country by 2025. Lack of and removing distortive sector-specific taxes demand, given affordability and other barriers solely applied to the mobile sector would to adoption, is the fundamental reason why improve investment incentives for operators as universal 4G coverage will be challenging well as affordability for consumers. This includes without further policy reforms and public the removal of excise duties on handsets and investment, and why ‘leapfrogging’ to 4G is mobile services that are not applied to other unlikely to occur in rural areas. However, this goods and services, as well as the reduction could change if further reform and/or investment of taxes levied on mobile operators but not on can stimulate large increases in demand. If other firms in the economy. This reform, even expected 4G penetration was 20 percent in considering partial pass-through on prices, could uncovered (mostly rural) areas, operators would expand mobile broadband coverage by up to 4 extend 4G coverage to more than 89 percent percentage points (depending on the country) and of the population in all seven markets without could bring more than 30 million people online by the need of supply subsidies. If 4G penetration 2030 (increasing mobile internet adoption by 6 increased to 40 percent, 4G coverage would percentage points). Since taxation in the sector is almost reach the same levels as would be part of broad national taxation policy, expected achieved with a pure infrastructure subsidy. benefits for the sector and digital transformation Further research to better identify and evaluate of the economy have to be considered when interventions that can influence mobile internet assessing tradeoffs. An encompassing analysis of demand will therefore be important going forward effects in the whole economy is beyond the scope to enable widespread 4G coverage in Sub-Saharan of this report. Africa. These could include policies to enhance digital skills and literacy and the availability • While these policy reforms would drive important of relevant content and mobile applications, gains in connectivity, by 2030 it would still leave interventions to expand access to electricity, or 9 percent of the population without mobile interventions to improve access to mobile devices. broadband coverage and 17 percent of the population ages 10 and above offline. Additional The findings from this study can be used to directly interventions on the supply and demand sides inform policy making in the seven study countries, are therefore still needed to make mobile while the analysis can be leveraged to understand broadband coverage and adoption universal the impact of policies and investments in different by 2030. In the seven countries studied, with geographic locations. Many of the policy findings the policy reforms in place, around US$1.7 billion are likely to apply in other countries with similar of additional investment would cover the vast characteristics. However, given the extent to which majority of populations with 4G networks, of mobile connectivity varies by location, it is important which almost 40 percent could be funded by to carry out granular analysis in other countries to the private sector, leaving an investment gap ensure interventions are targeted at where they are of US$1.1 billion. Without the policy reforms needed and ensure their effectiveness. This report highlighted above, the investment gap would be uses a model developed by GSMA based on detailed US$1.3 billion; policy reforms can therefore save mobile network data collected from operators; but around 15 percent of the additional investment as coverage and usage expand, simulations need needed to achieve near-universal coverage. to be updated. Finally, it is also worth noting that addressing connectivity challenges in Africa require • An alternative, or at least complementary, the use of various technologies that respond to approach to expanding 4G coverage would be demand needs in a policy environment that allows to focus on additional policy reforms and for innovation on a level playing field. investment that stimulate demand. Over the next five years, current forecasts indicate that 4G penetration will not exceed 15 percent in any of the seven countries and in rural areas, it is expected that 4G penetration will be less than 6 Executive Summary Using Geospatial Analysis to Overhaul Connectivity Policies Definitions of terms used • Coverage. Population coverage refers to the share • Mobile internet user. A person who uses internet of a country’s population that lives in an area services on a mobile device. Mobile internet where the signal provided by a mobile network services are defined as activities that use mobile is strong enough to use telecommunication data. The data presented in this report on the services (voice, short messaging service [SMS], number of unique mobile internet users are and data). The signal strength is calculated with sourced from GSMA2 Intelligence and national standard signal propagation algorithms that use regulatory authorities. Granular estimates of the location of every tower in the country for all demand are calculated at a settlement level based operators combined (see Appendix A for further on an estimation model that uses population detail). The countrywide coverage footprint for settlement data, household survey, and night each technology (2G, 3G, and 4G) is obtained light data (see Appendix A for further detail). The by combining the coverage provided by each model is calibrated such that the total number individual site for every operator in the country. of unique mobile subscribers and mobile internet The coverage status for each individual settlement subscribers are consistent with estimates by in the country is estimated by overlaying these GSMA Intelligence as of Q3 2020. This ensures countrywide coverage footprints with population that the aggregate level of demand is accurate distribution data. and that it is distributed across the country It is possible that the estimates of coverage based on the relevant demand drivers and where presented in this report may differ from other coverage exists. sources due to differences in population data, the As the report refers to unique subscribers, rather propagation model used, or because the analysis than connections or SIM cards, the number of looks at the combined coverage for all mobile users may be lower than data published by mobile operators. Furthermore, the network infrastructure operators and national regulators. This is because data were collected in Q3 2018 (for Ghana), Q4 individuals can own and use multiple SIM cards. It 2018 (for Rwanda), Q3 2019 (for Nigeria), Q1 is also possible that the number of mobile internet 2020 (for Benin, Democratic Republic of Congo, users may have increased since Q3 2020. Sierra Leone), and Q4 2020 (for Tanzania). It is • Connected. ‘The connected’ or ‘connected therefore possible that population coverage in population’ refers to people who use mobile each country has since increased. This report does internet. not analyze coverage by geographic area. • Usage gap. Populations that live within the • Mobile broadband. 3G, 4G, or 5G technologies footprint of a mobile broadband network but do that enable high-speed access to the internet. not use mobile internet. • Coverage gap. Populations that live outside the • Mobile internet penetration (or adoption). This footprint of a mobile broadband network. refers to the percentage of a country’s population • Mobile owner/subscriber. A person who that uses mobile internet services. This report subscribes to a mobile service. They do not refers to the total population as the base (that necessarily use mobile internet. is, mobile internet penetration is calculated as the number of unique mobile internet subscribers divided by the total population). 2 GSMA = Global System for Mobile Communications. Using Geospatial Analysis to Overhaul Connectivity Policies Executive Summary 7 • Universal adoption. As it is not realistic to expect • Settlement. The population geographic every person to use mobile internet, ‘universal distribution data used for this report is sourced adoption’ is defined as when a country achieves from the High Resolution Settlement Layer 90 percent penetration of the population ages database,3 which provides estimates of human 10 years or older. This is consistent with the population distribution based on census data and Connecting Africa Through Broadband Initiative high-resolution satellite imagery. Settlements (Broadband Commission 2019) and means are defined as clusters of houses and buildings young children are excluded from the target. It located within a close distance to each other. The also considers segments of the population that clustering process used may result in settlements choose not to use information and communication that differ from local administrative or cultural technology (ICT), are prevented from doing so, or denominations of towns and villages. those that use shared facilities. In each of the • Rural/urban. For each country, we classify a seven countries considered in the study, achieving settlement as rural if the population is less than this definition of universal adoption would 5,000. We also applied an approach set out in the mean they would need to reach 65–70 percent recommendation to delineate cities and urban and penetration based on the total population. rural areas by several international organizations.4 • Feature phone. A mobile handset that allows For the majority of countries, the results were basic access to internet-based services but on a very similar (that is, rural/urban settlements were closed platform that does not support a broad mostly classified consistently regardless of the range of applications. The handset supports approach used). Where there were differences, additional features such as a camera and the the approach used in this report produced rural ability to play multimedia files such as music and population estimates that were more consistent video. A ‘smart feature phone’ is a feature phone with those published by national statistical that has an operating system that supports authorities and the United Nations (UN).5 We a range of applications created by third-party therefore applied a consistent approach across all developers and that are formatted to work on a seven countries. smaller screen and accessed through a nine-key • Market frontier. The level of coverage and layout and not a touch screen. adoption that will be provided by the private • Smart feature phone. A feature phone that has sector over the next five years based on current an operating system that supports a range of market conditions. It represents the coverage applications created by third-party developers and that operators can achieve with sites that are that are formatted to work on a smaller screen profitable. The market frontier can change with and accessed via a 9 key layout not a touch policy reforms that have an impact on network screen. deployment costs and/or adoption. • Smartphone. A mobile handset enabling advanced access to internet-based services and other digital functions. Smartphone platforms, such as Android and iOS, support a range of applications created by third-party developers. 3 Facebook Connectivity Lab and Center for International Earth Science Information Network - CIESIN - Columbia University. 2016. High Resolution Settlement Layer (HRSL). Source imagery for HRSL © 2016 DigitalGlobe. 4 ‘A Recommendation on the Method to Delineate Cities, Urban and Rural Areas for International Statistical Comparisons – prepared by the European Commission – Eurostat and DG for Regional and Urban Policy – ILO, FAO, Organisation for Economic Co-operation and Development (OECD), UN-Habitat, World Bank, 2020. 5 See, for example, https://population.un.org/wpp/. 8 Introduction Using Geospatial Analysis to Overhaul Connectivity Policies 01 Introduction Increasing access to the internet is one of the pronounced in Sub-Saharan Africa, where just over a great challenges of our time and has grown in quarter of the population uses the internet (Figure 1). importance since the outbreak of the COVID-19 The region also accounts for almost half of the 450 pandemic. Around half the world’s population do million people around the world who do not live in not have access to the internet (GSMA 2020; ITU areas covered by 3G or 4G mobile networks. Within 2020), meaning they are unable to access the the continent, the coverage gap is notably larger jobs, education, health care, financial services, and in Central Africa, while the usage gap is highest in information needed to fully participate in social, Eastern Africa (Figure 2). political, and economic life. This digital divide is most Figure 1 Figure 1. State of mobile internet connectivity, by region (2019) State of mobile internet connectivity, by region (2020) 1% 3% 2% 100% 4% 7% 5% 6% 90% 19% 22% 26% 80% 34% 40% 43% 70% 48% 61% 60% 53% 50% 40% 77% 72% 30% 64% 56% 51% 20% 45% 34% 28% 10% 0% North Europe & East Asia Latin Middle East South Sub- Global America Central & Pacific America & & North Asia Saharan Asia Caribbean Africa Africa Connected Usage Gap Coverage Gap Source: State of Mobile Internet Connectivity Report 2020, GSMA Using Geospatial Analysis to Overhaul Connectivity Policies Introduction 9 Figure 2 Figure 2. State of mobile internet connectivity, by region in Sub-Saharan Africa (2019) State of mobile internet connectivity in Sub-Saharan Africa, by sub-region (2020) 100% 12% 90% 18% 20% 80% 35% 70% 60% 49% 49% 66% 50% 40% 44% 30% 20% 33% 31% 10% 22% 21% 0% Southern Western Eastern Central Africa Africa Africa Africa Connected Usage Gap Coverage Gap Source: GSMA Intelligence. Note: Countries are initially classified based on whether they are in the Southern African Development Community, the West African Economic and Monetary Union, the East African Community, or the Economic Community of Central African States. Countries not in one of these economic unions are classified based on UN geographic classifications. Connectivity gaps are a consequence of fundamental to other barriers, including awareness, affordability, economic challenges around supply and demand access to electricity, digital skills, and literacy (Figure 3). In a market-led environment, mobile and the availability of relevant online content and operators will provide coverage where there is applications. Expanding mobile broadband coverage existing or expected demand for connectivity. will partly depend on lowering costs and investments Indeed, two-thirds of the unconnected population risks, but the main driver will be enhanced demand in Sub-Saharan Africa already have 3G/4G for mobile services. coverage but are not able to use the internet due 10 Introduction Using Geospatial Analysis to Overhaul Connectivity Policies Figure 3. Supply and demand of mobile connectivity Demand-side parameters Supply-side parameters captured by model captured by model Accessibility to mobile networks, Terrain and topography handsets and electricity* Population density and distribution Affordability of data plans and handsets Existing infrastructure (electricity, roads) (income levels, prices, consumer taxes)* Technology innovation* Digital skills levels** Infrastructure sharing* Availability of relevant content and services** MNO taxes and fees* Safety and security and other Spectrum costs* cultural norms** Costs Revenues of rolling out and operating from selling voice/sms and data networks (Capex and Opex) services (ARPU and penetration) Individual sites business case MNOs site by site roll-out decisions based on long-terms profitability Connectivity Nationwide coverage and adoption resulting from aggregate of individual decisions * Parameters for which we modelled the impact of different policies ** Parameters that are indirectly captured by the model using demand elasticities Source: GSMA. Note: ARPU = Average Revenue Per User; Capex = Capital Expenditure; MNO = Mobile Network Operator; Opex = Operating Expenditure. Using Geospatial Analysis to Overhaul Connectivity Policies Introduction 11 To address the digital gap in the region, mobile connectivity and identify appropriate policy reforms operators, governments, and international to expand digital access, the study uses a supply- organizations have committed to a number and-demand model for seven countries in Sub- of international targets to achieve universal Saharan Africa: Benin, Democratic Republic of Congo coverage and connectivity.6 These recognize the (Democratic Republic of Congo), Ghana, Nigeria, transformative impact that internet access can Rwanda, Sierra Leone, and Tanzania. Together, these have. Studies have shown that expanding mobile account for 40 percent of the population in Sub- broadband coverage and connectivity in Africa Saharan Africa. They also reflect the diversity of reduces poverty (GSMA and World Bank 2020, Bahia the continent, ranging from the largest countries et al. 2020 and 2021) and increases sustainable in terms of both population (Nigeria) and land area development (Rotondi et al. 2020) and economic (Democratic Republic of Congo) to smaller countries growth (GSMA 2020a; ITU 2019, Calderon and in Western Africa (Benin and Sierra Leone) and Cantu 2021). Eastern Africa (Rwanda and Tanzania). The countries are also at different stages in their development However, while there are common elements to the of digital connectivity (Table 1), which allows the barriers that stop people from using the internet, analysis to consider the broad range of challenges each country is unique and presents different that governments and the private sector face in the challenges in terms of geography, socioeconomic region. development, and the regulatory and policy framework. To better understand the state of Table 1. Key metrics on seven study countries UN Human Mobile Mobile Mobile Population GDP per Development Connectivity Country Area (km2) Internet Broadband (million) Capita (US$) Index Index Score Adoption (%) Coverage (0–1) (0–100)* Benin 12.12 114,760 1,219 0.545 39.1 27 89 Democratic Republic of 89.56 2,344,860 581 0.480 26.2 18 54 Congo Ghana 31.07 238,540 2,202 0.611 52.0 37 88 Nigeria 206.14 923,770 2,230 0.539 49.1 33 76 Rwanda 12.95 26,340 820 0.543 42.8 24 88 Sierra 7.98 72,300 528 0.447 38.6 28 77 Leone Tanzania 59.73 947,300 1,122 0.524 40.1 21 83 Source: GSMA analysis of data sourced from UN, World Bank, and GSMA Intelligence. Note: GDP = Gross Domestic Product. * Reflects the latest score in the GSMA Mobile Connectivity Index, which measures the performance of countries against the key enablers of mobile internet adoption. 6 Examples of such targets include the UN Sustainable Development Goals (SDG 9 includes Target 9c to “Significantly increase access to information and communications technology and strive to provide universal and affordable access to the internet in least developed countries by 2020”); the Broadband Commission for Sustainable Development 2025 targets for “Connecting the Other Half”; the World Bank’s Digital Economy for Africa, which aims to ensure that all Africans have universal and affordable access to information and communication technology (ICT) by no later than 2030 and the African Union’s Digital Transformation Strategy for Africa (2020–2030). 12 Introduction Using Geospatial Analysis to Overhaul Connectivity Policies In each of the seven countries, we carried out hyper- results from this study should therefore help support granular supply and demand analysis, leveraging policy makers make informed decisions to accelerate data from the GSMA Mobile Coverage Maps7 as connectivity (especially in rural areas)—not just well as country survey data, to understand current in the seven study countries but also in the wider levels of coverage and adoption in each settlement. region. The analysis also calculated the expected level of The policies covered in this report are not exhaustive. connectivity in current market conditions, that is, When considering policies that might increase the additional amount of coverage and adoption adoption of mobile internet services, focus was that would be delivered by the market, given primarily given to affordability and access to devices. expected growth in demand over the next five years This is because there were insufficient data to (Box 1).8 model the impact of policies that might influence The next step was to assess the impact of different other enabling factors, especially those related policy levers that can affect either supply (that is, to digital skills and literacy and the availability of costs) and/or demand (that is, adoption or revenue). relevant content. These are areas that will benefit The study first considered different aspects of an from further research and analysis, as they remain enabling policy and regulatory framework, ranging important barriers to adoption in many countries. from infrastructure sharing and spectrum policy Furthermore, the models developed for this study to taxation and SIM registration. If these policy have the capability to assess the impact of other reforms were not sufficient to meet the goal of policies that affect operational costs or prices when universal connectivity, we estimated the amount the relevant data become available. of additional investment needed to achieve it. The The structure of the rest of this analysis is as follows: • Chapter 2 provides provides an assessment of • Chapter 4 shows which enabling policies the current state of connectivity for 2G, 3G, and considered in the study can increase coverage and 4G networks in the seven countries based on adoption. granular estimates with latest available detailed • Chapter 5 highlights the additional interventions country-level information. needed to make connectivity universal. • Chapter 3 discusses the recent innovations that • Chapter 6 draws conclusions from the analysis. are already enabling increased coverage in rural and remote areas using mobile technologies and that are essential to unlocking the impacts of further policy reforms. 7 https://www.mobilecoveragemaps.com/ 8 For simulations of the effects of changes in market structure (as a proxy for stronger competition) on coverage and adoption, see Dutz, Begazo and Blimpo (forthcoming). Using Geospatial Analysis to Overhaul Connectivity Policies Introduction 13 Box 1 Analytical approach In each of the seven countries, the model used geospatial techniques to carry out hyper-granular supply and demand analysis, leveraging data from the GSMA Mobile Coverage Maps as well as data from household surveys, night-light imagery, and geospatial population distributions. This allowed to calculate, for each population settlement in a country, coverage by technology (2G, 3G, and 4G) and the level of adoption for mobile and mobile internet services. As the mobile infrastructure data are sourced directly from mobile operators, based on a comprehensive list of site locations, the analysis provides the most accurate appraisal of network coverage currently available in the public domain. Based on this data, a model was developed to emulate the decision-making process of mobile operators when they consider whether to invest in 2G/3G/4G network expansion. The model is focused on the ‘last mile’ of infrastructure, that is, the mobile site that connects with the end user as well as the backhaul link that connects sites to the core network. While investments in the ‘first mile’ (for example, international cables) and ‘middle mile’ (for example, backbone and internet exchange points [IXPs]) are important in terms of increasing network capacity, especially in urban areas, based on the current and expected levels of data usage in rural areas across the seven countries, we found that operators have sufficient network capacity to meet demand in uncovered areas. The most significant barriers to expanding coverage are in the last mile, so they are the focus of this study. The analysis was based on a net present value (NPV) approach at the level of individual sites, where operators decide whether to invest based on the expected revenues and the associated capital and operating costs from either upgrading an existing site (for example, from 2G to 3G/4G) or deploying a new site (where no coverage exists). For each site, we assess profitability considering the relevant country weighted average cost of capital (WACC) over five years. The ‘market frontier’ represents the aggregate number of sites that are profitable (with non-negative NPV), that is, where supply (costs) is equal to demand (revenues). This gives the expected level of coverage that will be provided by the private sector over the next five years, which we define as the current market frontier. The next step was to assess the impact of De different policy levers that can affect costs ma rve and/or adoption. This in turn can affect the nd cu cur ly ve pp profitability of new sites, which can then Su increase (or decrease) expected coverage and therefore change the market frontier. Lastly, 150 K$ Market frontier for sites that remain unprofitable even after policy reform, the model calculates the level of additional investment needed to achieve Cost (P) near-universal access. There were then some remaining population segments where, given Coverage (Q) the high cost, alternative technology solutions 60% 0% 100% are likely to be needed. An illustration of the outcomes at each stage in the analysis is Urban population Rural population provided in figure 4, while further detail on the modelling is provided in Appendix A. 14 Introduction Using Geospatial Analysis to Overhaul Connectivity Policies Figure 4 Illustration of mobile connectivity analysis Figure 4 Illustration of mobile connectivity analysis 01 02 03 04 05 Market-led Market Connectivity Connectivity Unattainable connectivity innovation with policy with public population changes intervention using mobile Private Public + private investment investment Market-led The level of coverage and adoption expected in prevailing market conditions with connectivity. no policy reform or innovation. This is discussed in Chapter 2. Market innovation. The additional coverage and adoption expected due to the deployment of recent Market-led The level of coverage and adoption expected in prevailing market conditions, with technology innovations for mobile connectivity, particularly lower-cost mobile sites. connectivity: no policy reform or innovation. This is discussed in Chapter 2. This is discussed in Chapter 3. Market innovation: Connectivity with The additional coverage and adoption expected due to the deployment of recent The additional coverage and adoption that could be achieved by implementing the policy changes. technology innovations for mobile connectivity, particularly lower-cost mobile analyzed policy reform. This is discussed in Chapter 4. sites. This is discussed in Chapter 3. Connectivity with The additional coverage and adoption that could be achieved with a subsidy Connectivity with public intervention. The additional (subsidizing coverage and infrastructure adoption and/or that could handsets be achieved and mobile usage).by implementing is discussedthe This in policy changes: analysed Chapter 5.policy reform. This is discussed in Chapter 4. Connectivity Population with The additional The proportion coverage and adoption of the population that unlikely tocould be obtain achieved mobile with a coverage subsidy even with a public intervention: unattainable using (subsidising infrastructure and/or handsets and mobile usage). This is discussed public subsidy, as the costs are too high; other technologies might be more suitable. mobile. in Chapter This 5. is discussed in Chapter 5. Population The proportion of the population unlikely to get mobile coverage even with unattainable using a public subsidy, as the costs are too high; other technologies might be more mobile: suitable. This is discussed in Chapter 5. Using Geospatial Analysis to Overhaul Connectivity Policies Introduction 15 16 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies 02 Connectivity has Been Delivered by Market Forces 2G coverage has almost reached the current market frontier Across all the countries studied, mobile operators are conflict areas and so are unlikely to materialize until very close to the level of population coverage for 2G these are resolved.9 that is commercially viable (Figure 5). For example, The extent of 2G population coverage varies across 2G population coverage stands at 75 percent in the countries, ranging from almost universal in Democratic Republic of Congo and this analysis Rwanda and Benin to 75 percent in Democratic suggests that when looking at new potential macro- Republic of Congo. In all seven countries, coverage site deployments, those that are profitable would in urban areas is either currently or expected to provide coverage to an additional 1.2 percent of the be more than 99 percent by 2025. The remaining country’s population. Deploying networks in the vast populations that will not have coverage by 2025 are majority of areas that are not covered by a mobile almost entirely in rural areas, many of them being network is therefore not financially sustainable. The sparsely populated (Figure 6). notable Figure 5 exception is Nigeria, though it is worth noting that the almost halfin 2G coverage potential seven coverage gains countries, are in and current market frontier existing Figure 5. 2G coverage in seven countries 100% 0.0% 0.6% 0.0% 0.3% 6.7% 99.4% 97.3% 90% 94.1% 1.4% 93.7% 88.1% 86.7% 80% 1.2% 70% 75.2% 60% 50% 40% 30% 20% 10% 0% Rwanda Benin Tanzania Ghana Nigeria Sierra Leone DRC Network coverage Market frontier Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data, and Group on Earth Observations. 9 In Nigeria, three states—Adamawa, Borno, and Yobe—are identified as high-security risk areas by the regulator. It is therefore likely that no additional coverage will be deployed in these areas until there is a resolution to military and political conflicts. Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 17 6. Figure 6. Figure Coverage maps population coverage 2G Population Maps Benin DRC Nigeria Ghana Sierra Leone Rwanda Percent Tanzania 0–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80 80–90 90–100 Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data and Group on Earth Observations. Coverage is provided for the lowest-level administrative unit that was available for analysis (usually Administrative level 3). 18 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 19 3G and 4G coverage are expected to approach 2G coverage In the case of 3G and 4G, the market frontier These improvements have been enabled by supply analysis shows that coverage is expected to and demand factors. In terms of the former, many increase over the next five years, most notably in regulators have allowed operators to refarm the Nigeria (and Benin in the case of 4G) due to the 900 MHz spectrum for 3G use and have assigned higher expected demand. Almost all the expected sub-1 GHz spectrum for 4G, which has enabled more gains in mobile broadband coverage will come cost-efficient deployment in more sparsely populated from upgrading existing 2G-only sites (given that areas (GSMA 2019a). additional coverage from new ‘greenfield’ sites10 Similar to 2G coverage, the majority of urban areas is likely to be limited, as highlighted by the market have or are expected to have 3G/4G coverage by frontier for 2G). This is consistent with the broader 2025, meaning most of the uncovered populations trend seen in Sub-Saharan Africa over the last will be sparsely populated areas with lower levels few years, as 3G population coverage in the region of socioeconomic development—for example, in the increased from 63 percent in 2017 to 75 percent in north of Ghana and outside of the western regions 2019, while 4G population increased from 25 percent of Sierra Leone (Figure 8). to 49 percent over the same period (GSMA 2020). Figure77. Figure 3G and 4G Coverage coverage by by technology technology countries in sevencurrent in seven countries, and market frontier 100% 0.0% 1.8% 0.3% 90% 8.2% 3.7% 17.0% 2.8% 80% 28.6% 8.9% 70% 51.6% 60% 6.5% 12.5% 27.4% 50% 88.5% 88.4% 5.1% 40% 84.0% 88.1% 83.2% 75.5% 76.8% 67.7% 30% 61.5% 55.3% 54.0% 20% 41.2% 41.3% 42.2% 10% 0% 3G 4G 3G 4G 3G 4G 3G 4G 3G 4G 3G 4G 3G 4G Rwanda Benin Tanzania Ghana Nigeria Sierra Leone DRC Network coverage Market frontier Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Note: For Rwanda, an SWN operator provides 4G services, so no market analysis was carried out to determine the additional coverage that might be provided by mobile operators. This is because the model is built to emulate the decision-making process of mobile operators when they consider whether to invest in network expansion. Given that operators are unable to deploy 4G in Rwanda, and that the objective of an SWN is to meet a public policy goal of wide coverage (as opposed to a profit maximization objective), we are not able to use the model to determine 4G coverage by operators. Recent announcements also indicate that the SWN provider in Rwanda (KT Rwanda Networks) has achieved 4G population coverage above 98 percent.11 10 ‘Greenfield’ sites refer to the deployment of new mobile sites in their entirety, including passive (for example, tower and mast) and active (for example, base station or radio network controller) elements. They provide coverage to populations that previously had no network coverage for any technology. ‘Brownfield’ sites refer to the upgrading of existing sites to provide 3G and/or 4G connectivity. Upgrades can either involve the installation of new hardware and equipment or, if single radio access networks are deployed, they can be upgraded simply by activating the 3G and/or 4G radio bearers. 11 See ‘Network upgrade and new product launch’, Korea Telecom Rwanda Networks. 20 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies Figure Figure 8. 8. 3G and 3G 4G population and 4G coverage maps population coverage maps Benin 3G 4G DRC 3G 4G Ghana 3G 4G Nigeria 3G 4G Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 21 Rwanda 3G 4G Sierra Leone 3G 4G Tanzania 3G 4G Percent 0–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80 80–90 90–100 Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data and Group on Earth Observations. 22 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies ‘Leapfrogging’ to 4G is unlikely to happen in rural areas The analysis above shows that in none of the proportion is even greater in rural areas. Furthermore, seven countries is 4G (or 3G) coverage expected in four of the countries studied—Benin, Rwanda, to surpass 2G coverage over the next five years. Democratic Republic of Congo, and Sierra Leone—the While such ‘leapfrogging’ has been suggested as a majority (55–70 percent) of mobile internet users way of bypassing legacy technologies, barriers such still do not connect with 3G or 4G technology, even as the affordability of 4G devices or lack of digital though mobile broadband coverage is widespread skills limit the demand for 4G services and result in in three of these countries (and even in Democratic revenues that are insufficient to cover the costs of Republic of Congo, 3G networks cover more than half the infrastructure without the revenues coming from of the population). 2G services such as voice and SMS (Figure 9). This Figure 9 Service Figure Revenue breakdown by operator 9. Service revenue breakdown by operator 100% 90% 26% 27% 23% 33% 30% 80% 43% 70% 3% 12% 19% 16% 60% 23% 5% 50% 40% 71% 65% 30% 54% 55% 53% 20% 43% 10% 0% MTN MTN Airtel East Airtel Sonatel Vodacom Nigeria Ghana Africa Francophone Africa Voice/SMS Mobile money / fintech Data Source: GSMA analysis of operator annual reports for MTN (2020), Airtel (2020), Sonatel (2020), and Vodacom (2019). Service revenues include voice, SMS, mobile data, mobile money, and fintech. It excludes revenues from devices and wholesale services. Note: Airtel East Africa includes operations in Kenya, Malawi, Rwanda, Tanzania, Uganda, and Zambia. Airtel Francophone Africa includes operations in Niger, Chad, Gabon, Congo, Democratic Republic of Congo, Madagascar, and Seychelles. Sonatel includes operations in Senegal, Mali, Guinea, Guinea-Bissau, and Sierra Leone. Vodacom includes operations in South Africa, Tanzania, Democratic Republic of Congo, Mozambique, Lesotho, and Kenya. Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 23 There are several reasons why users are unable or cost to add the required backhaul capacity to provide unwilling to use mobile broadband services, including 3G upload and download speeds. The amount of access to 3G/4G handsets and barriers around this extra cost depends on the existing backhaul awareness, digital skills, and literacy. technology and capacity and may vary from zero, when fiber or high-capacity microwave links are in For example, estimated smartphone adoption in the place, to the full cost of deploying a fiber or a new study countries ranges from around 15 percent in microwave link to connect the site. In some cases, Tanzania to almost 40 percent in Ghana.12 However, when several microwave hops are needed to reach there is a significant rural-urban gap in smartphone a point of presence, a further cost is incurred to ownership. In Nigeria and Tanzania, adults living in upgrade those upstream backhaul links. These costs rural areas are around 35 percent less likely to own a are part of the return on investment calculation smartphone than those living in urban areas.13 made by operators when deciding the technology mix The key implication is that there is not a strong to deploy or upgrade in any given area. commercial case to expand 3G or 4G coverage Voice and SMS services, which are powered by 2G alone. Some operators therefore deploy only 2G in devices, are therefore effectively paying for the bulk new areas, delaying the decision to roll out 3G or 4G of the passive infrastructure costs and covering an until there is a sufficient return, while others have important part of the cost of deploying 3G and 4G made a strategic decision to deploy 3G alongside networks. This also explains the limited examples 2G, especially where single radio access network of commercially scalable solutions relying on Wi-Fi technology (SRAN14) is used and operators can technology in rural areas to offer only data services, reallocate spectrum in the 900 MHz band. However, such as those used in some community network even in these cases, the majority of revenues still projects. Data-first technology rollouts will become comes from voice and SMS. It is important to note sustainable when there is sufficient demand in that beyond the costs of the base station, when rural areas to cover the full costs of the underlying operators decide to deploy 3G alongside 2G or infrastructure (active, passive, and backhaul). upgrade sites to 3G, operators may incur an extra 12 Source: GSMA Intelligence (2020). Smartphone adoption estimates are calculated based on the number of SIM cards used in smartphones divided by the average number of SIM cards per unique subscriber. As there can be differences in the number of SIM cards used by smartphone owners and non-smartphone owners, these estimates may not be precise. 13 Source: GSMA Intelligence Consumer Survey. Estimates are based on surveys carried out in 2020 for Nigeria and 2018 for Tanzania. 14 Single RAN technology allows a single mobile base station to support multiple technologies (2G, 3G, and 4G) concurrently. It often relies on software-defined radios which can be activated and configured remotely. 24 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies Affordability will remain a key barrier to adoption without policy reform On the demand side, more affordable handsets and relevance as the main barriers to adoption and data plans have driven an increase in mobile (GSMA 2019). Across the seven countries included internet adoption, which has enhanced the business in this study, adult literacy ranges from 43 percent case to deploy 3G and 4G. This is particularly the in Sierra Leone to 79 percent in Ghana.15 In terms case in Ghana and Nigeria; both countries now meet of affordability, Nigeria is the only country in the the affordability target set by the UN Broadband study where a 1 GB data plan is affordable for more Commission to make 1 GB of monthly data cost less than half the population (based on the 2 percent of than 2 percent of the average monthly income per monthly income threshold).16 Access to electricity capita (although this is not the case for all segments is a further barrier, for network deployment and of the population—see discussion in Chapter 3). for consumers (for example, the ability to charge As a result, the two countries have adoption levels devices). More than half the population in five of the that are higher than the average for Sub-Saharan countries studied (Democratic Republic of Congo, Africa. On the other hand, Democratic Republic of Sierra Leone, Rwanda, Tanzania, and Benin) do not Congo has lower levels of mobile internet adoption have access to electricity.17 and mobile broadband coverage, with more than 40 As a result of these barriers, the majority of the percent of the country’s population not living in an population in all seven countries is unable or unwilling area with a 3G or 4G network. to use mobile internet services.18 Furthermore, The key barriers to adoption are well known. Mobile there are notable digital inclusion gaps within each users that are aware of mobile internet but do not country; for example, those living in rural areas use it are most likely to cite digital skills and literacy, are on average 75 percent less likely to use mobile affordability (including for handsets and data plans), internet than those in urban areas (Figure 11). Figure 10 Figure 10. Connectivity, usage and Connectivity, usage coverage gaps and coverage gaps 100% Coverage Gap 12% 12% 12% 90% 17% 25% 23% Usage Gap 80% Connected 46% 70% 51% 60% 61% 64% 43% 49% 63% 50% 40% 36% 30% 20% 37% 33% 28% 27% 24% 10% 21% 18% 0% Ghana Nigeria Sierra Leone Benin Rwanda Tanzania DRC Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. 15 Source: UNESCO Institute for Statistics. 16 Source: GSMA Mobile Connectivity Index. 17 Source: Sustainable Energy for All and World Bank. 18 Even if we consider populations ages 10 or above, mobile internet adoption ranges from 11 percent in Rwanda to 50 percent in Ghana, although this is likely to overstate adoption within the age group as mobile internet users can be under the age of 10. Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 25 26 Connectivity has Been Delivered by Market Forces Using Geospatial Analysis to Overhaul Connectivity Policies Figure 11. Figure 11 Mobile internet adoption maps Benin DRC Nigeria Ghana Percent 0–5 5–10 10–15 15–20 20–25 25–30 30–35 35–40 40–45 45–50 50–55 55–60 60–65 65–70 Over 70 Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data and Group on Earth Observations. Using Geospatial Analysis to Overhaul Connectivity Policies Connectivity has Been Delivered by Market Forces 27 Sierra Leone Rwanda Tanzania Percent 0–5 5–10 10–15 15–20 20–25 25–30 30–35 35–40 40–45 45–50 50–55 55–60 60–65 65–70 Over 70 Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data and Group on Earth Observations. 28 Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change Using Geospatial Analysis to Overhaul Connectivity Policies 03 Innovation is Key to Unlocking Rural Markets and Boosting the Impact of Policy Change The market alone will not deliver universal connectivity Based on the current trends of coverage and public intervention will therefore be necessary to adoption, the goal of universal connectivity by narrow that gap and bring connectivity to everyone. 2030 will not be met in any of the seven countries However, the impact of these will be enhanced by— considered in the study. This analysis suggests and in some cases contingent on—the innovations adoption of mobile internet and will range from being led by the private sector and supported by 37 percent in Democratic Republic of Congo to governments and international organizations. 54 percent in Nigeria by 2030, with significant rural-urban gaps persisting. Policy reform and Innovative technologies improve the business case in the last mile On the supply side, to address the high costs ‘light towers’, for example Huawei’s RuralStar and associated with expanding network deployment Nokia’s Kuha (GSMA 2019b) or solutions from new in rural areas, the mobile ecosystem is innovating vendors such as Vanu, Inc. and NuRAN Wireless to more nimble technical solutions. These target Inc. While these do not have the same geographic coverage to isolated and dispersed population reach in terms of coverage, they are better suited settlements more cost-efficiently than traditional to covering remote locations with small populations macro-sites. The innovations also reduce the total (Figure 12). The typical cost of ownership over five cost of ownership of a cell site as they are designed years can range from US$50,000 to US$120,000 specifically to work in rural settings and are cheaper for a smaller site (depending on the configuration), to deploy, maintain, and power. Most importantly, compared to US$200,000–500,000 for a macro- they are able to target investment where needed, site.19 One of the reasons for lower costs is because reducing the average investment per person covered. they are powered by renewable energy solutions, particularly solar, rather than more expensive diesel The three main areas where infrastructure costs generators. Further innovation such as open radio can be prohibitive in rural areas are the mobile access network (RAN) standards will continue base station, the backhaul technology that to drive down the costs of last-mile technology. connects mobile sites to the core network, and Additional experience and analysis would be needed energy supply. In recent years, there have been a to ascertain whether this cost reduction will be number of commercially developed innovations large enough to disrupt the fundamental economics for base station solutions that provide lower-cost observed in this study, but this seems unlikely. This 19 Cost data sourced from mobile operators and equipment vendors. The cost of macro-sites varies by country and depends on the commercial model used (for example, whether a tower company operates the passive elements of a site), the technology solution and vendor, power source, the spectrum available, and installation and maintenance costs. Using Geospatial Analysis to Overhaul Connectivity Policies Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change 29 is mainly because the active equipment, such as these new smaller sites are often delivered using base stations or backhaul modems, account for only innovative business models. For example, some a small part of the total cost of ownership of a vendors or network integrators offer revenue- site (around 20 percent for a five-year total cost of sharing or ‘network-as-a-service’ models where they ownership) so even innovations that can drastically finance and operate the infrastructure and lease cut the costs of this active equipment will likely only the added coverage or capacity to mobile operators. affect that share of the costs and will probably not Such models encourage infrastructure sharing and result in coverage benefits much larger to those in enable third-party investment in rural connectivity. the ‘market innovation’ scenarios modelled for this These vendors or integrators, such as Africa Mobile report. Networks and iSAT, have streamlined their solutions Figure 12 to rural areas, driving efficiencies that result in lower Furthermore, while traditional network rollout models Comparison of traditional macro-sites with smaller, lower cost sites network costs. have seen mobile operators providing the necessary capex and opex to build and maintain every site, Figure 12. Comparison of traditional macro-sites with smaller, lower cost sites High performance Radio equipment radio equipment to specifically designed cover wide areas to serve targeted requires high power remote areas High cost large Easy-to-build light towers needing high towers with optimal cost civil engineering height for location on large sites and low cost build on smaller sites Costly transmission links Transmission optimized to to connect to the internet reduce capacity required Expensive energy via diesel Solar energy generators Source: GSMA 30 Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change Using Geospatial Analysis to Overhaul Connectivity Policies For this study, new site innovations were modelled are expected to expand rural coverage across all as part of the expansion of coverage in ‘greenfield’ markets, notably in Democratic Republic of Congo, areas (that is, settlements not covered by any which has a larger coverage gap21 (Rwanda is not mobile technology).20 Figure 13 shows that they included as 2G coverage is almost universal). Figure 13 Additional Figure 13. mobile coverage (any technology) in rural areas expected from recent site Additional mobile innovations coverage for greenfield (any technology) in rural areas expected from recent site investments innovations for greenfield investments 100% 1.7% 1.3% 2.3% 90% 2.5% 3.4% 80% 70% 60% 5.8% 50% 95.9% 90.3% 88.4% 86.2% 40% 79.0% 30% 53.2% 20% 10% 0% Benin Tanzania Nigeria Ghana Sierra Leone DRC Network coverage Market frontier (with innovation) Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Innovation reduces the cost of universal coverage Lower-cost networks will be critical in unlocking and the total cost of achieving near-universal 4G further investments in coverage. To illustrate this, population coverage of 99 percent over a 10-year one can compare the cost of expanding coverage in period would be US$2.6 billion (based on current Democratic Republic of Congo (which has the lowest market costs). By contrast, if only traditional macro- level of coverage across the seven countries) with sites are used, the equivalent cost would be US$4.4 and without new lower-cost sites. When these are billion. This means new lower cost sites will reduce included in the list of potential deployment options the cost of near-universal 4G coverage by around 40 for operators, it is expected that they would account percent. for more than 97 percent of new greenfield sites 20 In ‘brownfield’ areas, upgrading an existing 2G macro-site to 3G or 4G will be more cost effective than deploying a new smaller site. 21 Operators in DRC have already started to partner with network service providers to expand coverage using new and lower-cost sites (for example, Nuran Wireless). Using Geospatial Analysis to Overhaul Connectivity Policies Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change 31 Microwave backhaul will remain the dominant technology for rural connectivity, though more innovation is needed to reduce costs In addition to the site configuration and energy where microwave is not viable. While satellite source, innovation is needed in terms of backhaul. backhaul can provide a solution for some rural In Sub-Saharan Africa, around 85 percent of cell and remote sites (Table 2), they are not currently backhaul uses microwave, compared to 4 percent for expected to deploy 3G or 4G at large scale due to fiber and 5 percent for satellite (with the remainder the high cost (the annual opex of providing satellite using copper).22 In rural areas, the proportion of backhaul for a 3G site is currently around four sites using microwave backhaul is even higher. For times higher than for 2G, while for 4G it is eight example, in rural areas of Ghana and Nigeria, between times higher). The most appropriate technology to 98 percent and 99 percent of mobile sites use use depends on the specific requirements for each microwave backhaul, with most of the remainder individual site, such as the terrain in that precise using fiber. This is not expected to significantly location, the distance to the closest fiber backbone, change in the future considering the high cost of fiber or the capacity required for the site. Every technology deployment. However, in many areas, microwave will has its trade-offs in terms of deployment costs, not be a viable backhaul solution if there is no line of licensing costs, capacity, and range.23 sight—especially in areas that are mountainous or As data traffic in rural areas increase to the observed have rugged terrain. While there are some innovations levels in urban areas, quality of service might become in backhaul technology (for example, the use of an inhibitor of adoption if the backbone network long-term evolution or even laser beams), these have does not provide sufficient capacity.24 Upgrading similar physical constraints to microwave and are not the backbone capacity will require investing in fiber sufficient to extend the reach of links to connect the and high-capacity microwave links. And while the most remote locations. level of coverage is not directly affected by these The next generation of satellite backhaul will investments (and as such not included in this study), therefore play an important role. For this study, we they will be necessary for maintaining a good quality assumed that sites would use satellite backhaul of service and keeping data bundles affordable in the medium term. Table 2. Proportion of new greenfield deployments that use smaller sites and satellite backhaul Percentage of new profitable greenfield sites that Use smaller, lower Use satellite Use satellite backhaul cost configurations backhaul* for 3G or 4G Ghana 84 20 0 Benin 75 35 0 Sierra Leone 80 28 0 Democratic Republic of Congo 92 38 15 Nigeria 28 23 9 Tanzania 100 36 0 GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Note: *Satellite backhaul is needed where sites cannot link to the core network with a microwave link, either directly or through multiple ‘hops’ (that is, connecting to the core network through other cell sites). 22 See, for example, GSMA and ABI. 2018. Mobile Backhaul Options: Spectrum Analysis and Recommendations. These proportions are similar to the backhaul used in other low- and middle-income countries. By contrast, in North America and North East Asia, 70–80 percent of backhaul uses fiber. 23 For a detailed comparison of the trade-offs of backhaul technologies see page 3 the report cited above. 24 For a capacity-based analysis of universal broadband connectivity in Africa, see Oughton, Edward (2022, forthcoming) 32 Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change Using Geospatial Analysis to Overhaul Connectivity Policies Continued price reductions in data plans and devices are fueling adoption While innovation is critical to improving the provision organizations, including KaiOS technologies, UNISOC, of mobile infrastructure, it is equally important to and Google, to bring these devices to several markets increase adoption of mobile and mobile internet in Africa—for example, the MTN Smart Feature services on the demand side. An important area Phone in Nigeria (MTN 2018) and the Orange Sanza is the development of cheaper devices and mobile in Democratic Republic of Congo (Orange 2019). data. Figures 14 and 15 show that over the past five Furthermore, the partnerships often include the years, the cost of accessing 1 GB per month and an leveraging of mobile agents and salespeople to help internet-enabled device as a proportion of monthly new users learn and develop the basic skills needed gross national income (GNI) per capita has fallen to use their phones (GSMA 2021). across almost all the seven markets included in the Similar to the innovations in infrastructure, achieving study, in line with the broader trend in Sub-Saharan universal connectivity in Sub-Saharan Africa will Africa. Ghana and Nigeria have achieved affordability only be possible if these innovations are built on. levels for data that are below the 2 percent of Despite these improvements, even a US$20 phone monthly income threshold set by the UN Broadband represents a significant one-off cost for many of Commission, while device affordability in the two the poorest populations in Sub-Saharan Africa (for countries is also below the global average (around 20 example, the median cost of the cheapest internet- percent) seen in low- and middle-income countries enabled handset represents more than 120 percent worldwide. of monthly income for the poorest 20 percent of The emergence of cheaper smartphones and ‘smart the population) (GSMA 2020c). Other solutions feature phones’ has been particularly important will therefore need to be developed, such as device in reducing device costs in many countries, in financing schemes—these are already in place in some cases at or close to US$20. Operators some markets.25 have partnered with a range of private sector 25 See, for example, Safaricom’s partnership with Google to help consumers access new 4G smartphones at https://www.safaricom.co.ke/about/media-center/ publications/press-releases/release/979. Using Geospatial Analysis to Overhaul Connectivity Policies Innovation is Key to Unlock Rural Markets and Boost the Impact of Policy Change 33 Figure 14 Figure 14. Cost of Cost cheapest monthly of cheapest 1 GBdata monthly 1GB dataplan planas asaa%percentage of monthlyof GDP per capita monthly (2016 GNI per and capita 2020) (2016 and 2020) 50% 45% 42% 40% 35% 30% 25% 20% 20% 15% 10% 10% 10% 9% 6% 5% 3% 3% 3% 3% 3% 4% 3% 2% 1% 1% 0% Benin DRC Ghana Nigeria Rwanda Sierra Leone Tanzania SSA median 2017 2020 Source: GSMA analysis of Tarifica data. Figure 15 Cost of cheapest internet-enabled smartphone or feature phone as a % of monthly Figure 15. GDP per Cost capita (2016 of cheapest and 2020) smartphone or feature phone as a percentage internet-enabled of monthly GNI per capita (2016 and 2020) 100% 95% 90% 80% 63% 60% 60% 49% 44% 44% 39% 40% 31% 28% 26% 23% 21% 20% 17% 13% 11% 0% Benin DRC Ghana Nigeria Rwanda Sierra Leone Tanzania SSA median 2017 2020 Source: GSMA analysis of Tarifica data. 34 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies 04 Enabling Policies Can Increase Network Coverage and Adoption As part of this study, a number of strategies that • Taxation. A tax framework that ensures taxes are could increase coverage and adoption in the seven broad and simple, comparable to other sectors of countries were assessed. In this section, we discuss the economy, can increase coverage and adoption those that could drive the most significant impact by improving investment incentives for operators based on the model simulations. These include the and affordability for consumers. following: The analysis in this section shows the potential • Infrastructure sharing. The sharing of impact of these policy changes on the market infrastructure assets can reduce costs and frontier in some of the seven study countries. In investment risks for operators seeking to expand particular, where they affect the profitability of new coverage in new areas, as well as increase service- site deployments (or upgrades) by changing demand based competition. and/or costs on the supply side, it is possible that operators will be able to expand coverage beyond the • Spectrum policy. Network coverage can be current market frontier. increased, especially in rural areas, if governments and regulators release sufficient spectrum at affordable prices and allow licenses that are technology neutral. Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 35 Infrastructure sharing can enable multiple operators to expand coverage Infrastructure or network sharing at the site level sharing this could increase to 85 percent for 3G can be broadly classified into three categories:26 and almost 88 percent for 4G. Similarly, Figure 17 shows that in Tanzania, which has passive sharing • Passive sharing – the sharing of the physical through tower companies, three networks could be mast and energy supply sustainable for 78 percent of the country’s rural • Active sharing – the sharing of RAN, including population for 3G and 41 percent for 4G. With site, mast, base transceiver station, backhaul, and active sharing, this increases to 83 percent and base station controllers27 63 percent, respectively and, in the case of 3G • Roaming – when a mobile customer uses a especially, is similar to coverage for a single network. network not provided by their operator While the expected coverage provided by network sharing is slightly inferior to the expected coverage In most of the markets included in the study, mobile of a single network,28 network sharing has obvious operators have partnerships with independent advantages by promoting competition at the service tower companies that sometimes own and manage and retail levels, benefitting consumers in terms of the passive elements of their sites. However, more prices and quality. Network sharing allows multiple extensive active sharing that could further reduce service providers to operate in areas where there deployment costs is generally yet to occur. would otherwise only be one. Moreover, in all seven The primary benefit of infrastructure sharing is countries studied, it was found that promoting active that it can enable increased competition and sharing in new greenfield areas would achieve the consumer choice in areas where multiple networks same level of additional coverage as a single network are not commercially sustainable due to a lack of expansion. This again highlights the potential role demand. This is shown in Figures 16 and 17, which of infrastructure sharing as a means of increasing respectively show the level of 3G and 4G coverage service-level competition in areas currently that is financially viable in rural areas in Benin and uncovered, as well as those with only one active Tanzania. Infrastructure sharing also reduces the operator. capital intensity for mobile operators looking to These results show that network sharing is a expand their coverage, freeing capital to roll out viable option to extend coverage and competition infrastructure in less attractive (but still profitable) simultaneously. And because shared networks are locations instead of duplicating infrastructure in co-owned by service-level providers, it avoids the places where there is already another provider. drawbacks of SWNs, which impose a monopoly at Figure 16 shows that in Benin, where minimal the infrastructure level. This can result in a lack network sharing is in place, around 79 percent of of innovation and abuse of monopolistic power on the country’s rural population could have 3G/4G wholesale prices, harming consumers in the long coverage from three networks. However, with active term. 26 For further details on the different types of infrastructure sharing see, for example, Mobile Infrastructure Sharing, GSMA. 27 A deeper level of infrastructure sharing could also involve sharing of spectrum held by each operator (known as Multi-Operator Core Network or MOCN). We focus on Multi-Operator Radio Access Network (MORAN), where radio access networks are shared but each operator uses its own dedicated spectrum. 28 The amount of expected coverage is not the same between a single network (with one operator) and a shared network (with multiple operators) because the sharing of assets between two or more operators imposes some additional costs (for example, in terms of power supply)—though they are much less than having multiple operators deploy separate networks. 36 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies Figure 16 Maximum level of commercially viable 3G/4G rural coverage by 2025 in Benin by Figure 16. of networks and infrastructure sharing type number Maximum level of commercially viable 3G/4G rural coverage by 2025 in Benin by number of networks and infrastructure sharing type 100% Coverage with 1 network 90% Coverage with 2 networks 1.3% 6.2% 2.7% 7.7% Coverage with 3 networks 80% 3.1% 2.1% 70% 60% 50% 85.0% 87.7% 40% 78.4% 79.2% 30% 20% 10% 0% Passive Active Passive Active sharing sharing sharing sharing 3G 4G FigureGSMA Source: 17 analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Maximum level of commercially viable 3G/4G rural coverage by 2025 in Tanzania by Figure number 17. of networks and infrastructure sharing type Maximum level of commercially viable 3G/4G rural coverage by 2025 in Tanzania by number of networks and infrastructure sharing type 100% Coverage with 1 network 90% Coverage with 2 networks 0.8% Coverage with 3 networks 80% 4.7% 0.7% 1.6% 70% 6.8% 60% 16.3% 50% 12.6% 40% 83.0% 78.2% 30% 63.1% 20% 41.0% 10% 0% Passive Active Passive Active sharing sharing sharing sharing 3G 4G Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 37 While active sharing has benefits for operators in Network sharing agreements can have different terms of sharing costs and lowering investment impacts on the degree of competition in the market. risk, as well as offering growth opportunities in new For example, the loss of infrastructure-based areas, operational challenges have thus far meant competition can bring a greater risk of exchange it has not developed at scale in Sub-Saharan Africa of sensitive information at the service level. On (both in terms of RAN sharing and roaming). These the other hand, network sharing deals can improve challenges emerge from the need for coordination competition by allowing smaller mobile players to between mobile operators. Technical coordination is grow faster and cover a broader range of areas in required to ensure the shared network is compatible the country. Considering the high costs of deploying with each operator’s respective legacy networks and in rural and remote areas, it is unlikely that these maintenance tools. In addition, coordinated decision- areas would benefit from infrastructure-based making is required on issues such as the location of competition—the examples of Benin and Tanzania new sites, network maintenance responsibilities, and show that without active sharing, a significant the ownership of the assets. This increased level of proportion of rural populations would only have a coordination involves costs that mobile operators single network (assuming an operator was willing are only willing to incur if the gain of co-investing to deploy at all). In terms of risks to competition, outweighs those costs, which mostly happens in very regulators and competition authorities often dense or remote areas. More sharing is expected to mitigate such concerns by establishing safeguards occur naturally as mobile operators push coverage to and/or monitoring the agreements using ex post more remote areas29 and traffic increases in urban competition laws and frameworks.30 areas drive greater network densification. However, it is important that infrastructure sharing remains voluntary; otherwise, it could negatively affect investment incentives as mobile operators wait for others to invest first. 29 One telling example is the 2×2 RAN sharing agreements in Brazil, where two separate sharing agreements between four operators (Claro-Vivo and Oi-TIM) decided to join forces to co-invest as a way to minimize the investment needed to comply with the ambitious 4G coverage obligations included with their licences. 30 For further discussion, see for example GSMA 2018. 38 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies 3G and 4G coverage can be increased by assigning the sub-1 GHz spectrum and allowing licenses to be technology neutral The spectrum used for 2G, 3G, and 4G mobile bands) are particularly well-suited for covering rural services can be grouped into two broad categories: areas; these have typically been assigned for the coverage bands for frequencies below 1 GHz and use of 4G. Almost all the countries included in this capacity bands for frequencies above 1 GHz. This study have assigned some digital dividend spectrum is based on the bands’ physical properties—lower to deploy 4G (Table 3), though not all operators frequencies suffer less attenuation, while frequencies have access to it. While this can be for commercial above 1 GHz allow operators to carry more capacity. reasons, it can also be due to governments not licensing all available spectrum or setting prices that To expand coverage, having sufficient spectrum in are not market driven. For example, in Ghana, only coverage bands is particularly important, especially one operator participated in the 2015 auction to in rural areas as operators can cover wider areas acquire 800 MHz spectrum, due to a high reserve using fewer sites. In other words, a fixed amount of price.31 Vodafone subsequently purchased spectrum capex will generate higher returns on investment, in this band in 2019 to launch its 4G network, while which means operators can cover more of a other operators are yet to do so. country’s population. The low frequencies of digital dividend spectrum (in the 600, 700, and 800 MHz Table 3. Spectrum bands assigned and used (by technology), 2020 700MHz 800MHz 900MHz 1.8GHz 2.1GHz 2.6GHz Benin 4G 2G/3G 2G/4G 3G 4G Democratic 4G 4G 2G/3G 2G/4G 3G 4G Republic of Congo Ghana 4G 2G 2G 3G 4G Nigeria 4G 4G 2G/3G 2G/4G 3G 4G Rwanda 4G* 4G* 2G/3G 2G 3G 4G* Sierra Leone 4G 2G/3G 2G/4G 3G Tanzania 4G 4G 2G/3G 2G/4G 3G Source: GSMA Intelligence. Note: *In Rwanda, 4G spectrum bands are assigned and used only by the SWN provider. 31 See GSMA and NERA (2017) and GSMA (2020c). Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 39 To illustrate the impact, Figure 18 compares the by using 1,800 MHz spectrum33, but if they also expected level of 4G coverage in rural areas that an had access to 700 or 800 MHz, they could cover operator could commercially achieve in Sierra Leone, another 11 percent. In Ghana, Tanzania, and Nigeria, Ghana, Tanzania, and Nigeria if they deployed in the operators using 1,800 MHz could cover 5 percent, 8 1,800 MHz and 700 or 800 MHz bands.32 In Sierra percent, and 4 percent more of their respective rural Leone, an operator could make a positive financial populations if they had access to 700 or 800 MHz case to cover 43 percent of the rural population spectrum. Figure 18 of 4G coverage in rural areas that is viable for an operator by 2025 when Level18. Figure deploying Level of 4Gon 700/800MHz coverage vs in rural 1800MHz areas spectrum that is bands viable for in Ghana, an operator by Sierra Leone 2025 when Tanzania and Nigeria deploying in 700/800 MHz versus 1,800 MHz spectrum bands 100% 90% Additional coverage with 700/800MHz 4.2% 80% Coverage with 1800MHz 70% 5.0% 7.6% 60% 50% 10.8% 85.7% 40% 30% 63.6% 62.3% 20% 42.9% 10% 0% Sierra Leone Ghana Tanzania Nigeria Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. The analysis assumes that annual spectrum fees do not change. In addition to the assignment of new bands, a example, the decision to make all spectrum bands further enabling spectrum policy is to make licenses technology neutral has yet to be implemented.34 technology neutral and offer operators the flexibility The impact of refarming would be significant: in to introduce and use technologies that best suit their current market conditions, an operator in Ghana customers’ needs. In many Sub-Saharan African would be able to cover around 78 percent of the countries, notably Nigeria, operators have been able rural population with 3G networks using 2,100 MHz to expand 3G coverage significantly in recent years spectrum, but if they used 900 MHz as well this by using refarmed 900 MHz spectrum. However, would increase to more than 81 percent (an increase this has not occurred in all countries. In Ghana, for of almost 450,000 people covered). 32 These are countries where not all operators have sub-1 GHz spectrum to deploy 4G services. 33 Note that in Ghana, the deployment of 1800 MHz spectrum for 4G is hypothetical as operators are not yet able to use the band to deploy 4G networks. 34 As of January 2021. 40 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies Aligning tax policy with best-practice principles can drive significant gains in network coverage and adoption Another important policy reform is the removal almost 25 percent of annual recurring revenues in of sector-specific (excise) taxes, particularly in mobile-specific taxation, including annual license and Sub-Saharan Africa, where the mobile sector is spectrum fees (this compares to around 3 percent in subject to some of the highest overall tax burdens.35 Ghana). Sector-specific taxes are levied on consumers and/ In addition to these, governments can reduce the or providers of mobile services in addition to other cost of network deployment (for operators) and economywide general taxes, such as corporation tax handsets (for consumers) by reducing or removing and value added tax (VAT). This can make mobile import duties on imported equipment —this has services more costly in terms of production and been done in Tanzania, for example. consumption, relative to other goods and services, and distort use and investment decisions. The study models the removal of sector-specific taxes and import duties in all seven countries to In many countries, the application of these taxes assess the impact on mobile broadband coverage is not aligned with the best-practice principles and the adoption of mobile and mobile internet of taxation recommended by international services. Below we show the results for two organizations such as the International Monetary countries with some of the highest tax burdens in Fund (IMF), World Bank, and OECD. These include the region: Democratic Republic of Congo, where the following: there is currently a 10 percent excise duty on the • Taxes should be as broad as possible, to avoid consumption of telecoms services, and Benin, where distorting markets. we model the impact of bringing operator taxes and fees in line with other countries in the region. In both • Taxes should be simple and certain. countries, we also assume the removal of customs • Taxes should not undermine affordability and duties on handsets and network equipment. access to services. In Democratic Republic of Congo, these combined • Taxes should not distort investment. reforms in tax policy could increase both 3G and 4G It is worth noting that sectoral taxation is part rural coverage by more than three percentage points of the broad national taxation policy. A country (equivalent to almost 1.5 million people), as improved tax strategy considers other aspects such as the affordability stimulates demand and therefore the potential for fiscal revenue mobilization in the returns on investment. In Benin, the reduction in short term and medium term through tax policy as operator taxes increases the returns on investment well as tax administration measures. Given fiscal and would drive an increase in 3G/4G rural coverage constraints, countries assess difficult tradeoffs of almost 8 percentage points (equivalent to almost to determine sectoral taxation. This report offers 500,000 people). simulations to understand better the effects of taxes The investment and affordability effects of reducing on coverage and take up of mobile broadband, and sector-specific taxes also increase demand in areas therefore on the digitalization potential of a country. that are already covered, leading to a broader Taxes on the consumption of mobile services can increase in adoption. Figure 19 shows that based on be particularly impactful as one of the primary forecast adoption in 2030, tax policies could increase barriers to adoption is the affordability of handsets mobile adoption and mobile internet adoption and mobile data. Of the seven countries included by 7.4 and 6.4 percentage points, respectively, in in this study, four apply a tax on consumption, Democratic Republic of Congo (equivalent to bringing in addition to VAT, based on a percentage of the more than 7.5 million people online). In Benin, it value of services sold (in Ghana, Rwanda, Tanzania, would increase adoption by 7.1 and 5.5 percentage and Democratic Republic of Congo). On the supply points, respectively (bringing almost 1 million people side, sector-specific taxes on operators can reduce online). investment incentives. In Benin, operators pay 35 See, for example, Rethinking Mobile Taxation to Improve Connectivity, GSMA, 2019. Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 41 Figure 19 Figure 19. of reduction in sector-specific taxes on rural 3G/4G coverage in DRC and Impact Benin Impact by 2025 of reduction in sector-specific taxes on rural 3G/4G coverage in Democratic Republic of Congo and Benin by 2025 100% 90% Coverage with tax reform 7.7% 7.7% 80% Coverage without tax reform 70% 60% 50% 3.7% 40% 81.5% 81.3% 30% 3.1% 45.7% 20% 24.8% 10% 0% 3G rural 4G rural 3G rural 4G rural coverage coverage coverage coverage DRC Benin Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook CIESIN, household survey data, and Group on Earth Observations. 42 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies Figure 20 Impact Figure 20. of reduction in sector-specific taxes on national mobile and mobile internet Impact ofin adoption DRC andin reduction by 2030 taxes on mobile and mobile internet adoption sector-specific Benin in Democratic Republic of Congo and Benin by 2030 100% 90% Adoption with tax reform 80% Adoption without tax reform 70% 60% 7.1% 7.4% 50% 5.5% 40% 6.4% 30% 56.1% 50.3% 20% 44.7% 36.9% 10% 0% Mobile Mobile Mobile Mobile adoption internet adoption internet (2030) adoption (2030) adoption (2030) (2030) DRC Benin Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. While the removal of sector-specific taxes would specific taxes, it is estimated that there will be improve coverage and adoption, the imposition of 700,000 fewer mobile users and 2.5 million fewer new or additional taxes would have the opposite 3G/4G mobile internet users by 2030 (reducing effect. At the end of 2020, a new tax was imposed overall mobile broadband adoption by 2 percentage on mobile consumers in Democratic Republic of points compared to a scenario where taxes stay as Congo, comprising an annual payment of US$1 for they were before the handset tax). Furthermore, the 2G handsets and US$7 for 3G/4G handsets, which reduction in demand is likely to affect the decision will make mobiles less affordable and therefore to expand 3G and 4G network coverage, resulting in reduce future adoption. The proposed charge for 4.5 million fewer people being covered by 4G (or 5 3G/4G handsets is the equivalent of a price increase percent of the population—again, this is compared to of almost 10 percent on the current cost of 1 GB a scenario where taxes stay as they were before the of data. In the absence of reforming other sector- handset tax). Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 43 Policy reforms will drive significant gains in coverage and adoption Table 4 provides a list of operator-led initiatives and the reforms would result in 23 million more people policy reforms that were modelled in each country. covered with mobile broadband networks (more Putting these in place would result in significant than 4 percent of the total projected population gains in 3G and 4G coverage, as well as enable the considered) and bring 37 million people online adoption of mobile and mobile internet services in (almost 7 percent of the total projected population the seven countries. Figure 21 shows how mobile considered). The gains in coverage would also benefit internet adoption and the usage and coverage gaps rural areas that have lower levels of socioeconomic would evolve by 2030 under two scenarios: a baseline development (Figure 22). scenario where adoption and coverage increase However, even with policy reforms in place, none based on prevailing market and regulatory conditions of the countries is expected to achieve universal and a policy reform scenario where the reforms set coverage and mobile internet adoption by 2030. out in Table 4 are applied. In Democratic Republic of Congo, Tanzania, and In almost all countries, policy reforms would drive Rwanda, more than half the population is still significant gains in adoption and mobile broadband expected to be offline in 10 years. coverage. Specifically, over the seven countries, Table 4. Innovation and policy reforms modelled in each country Innovation Infrastructure Spectrum Taxation Import duties sharing Benin Smaller, Active site Reduce annual Reduce Remove lower-cost sharing spectrum fees sector-specific equipment and sites operator taxes handset duties Democratic Smaller, Active site — Reduce Remove Republic of lower-cost sharing sector-specific equipment and Congo sites consumption taxes handset duties Ghana Smaller, Active site Technology Reduce Remove lower-cost sharing neutrality sector-specific equipment and sites consumption taxes handset duties Nigeria Smaller, Active site Sub-1 GHz Reduce Remove lower-cost sharing deployment or sector-specific equipment and sites 4G operator taxes handset duties Rwanda Smaller, Active site — Reduce — lower-cost sharing sector-specific sites consumption taxes Sierra Smaller, Active site Sub-1 GHz Reduce Remove Leone lower-cost sharing deployment or sector-specific equipment and sites 4G operator taxes handset duties Tanzania Smaller, Active site Sub-1 GHz Reduce — lower-cost sharing deployment or sector-specific sites 4G consumption taxes 44 Enabling Policies Can Increase Network Coverage and Adoption Using Geospatial Analysis to Overhaul Connectivity Policies Figure Figure21 21. Connected, usage Connected, and coverage usage and coverage gaps gaps by by 2030 2030 under baseline and under baseline policy reform and policy reform scenarios scenarios 100% 4% 7% 5% 1% 1% 7% 10% 8% 11% 13% 90% 20% 18% 23% 30% 80% 34% 35% 39% 44% 55% 70% 22% 36% 56% 44% 25% 46% 46% 60% 33% 33% 50% 40% 30% 61% 60% 58% 54% 54% 53% 51% 45% 44% 47% 44% 20% 43% 41% 37% 10% 0% Baseline Policy Baseline Policy Baseline Policy Baseline Policy Baseline Policy Baseline Policy Baseline Policy reforms reforms reforms reforms reforms reforms reforms Nigeria Sierra Leone Ghana Benin Rwanda Tanzania DRC Connected Usage Gap Coverage Gap Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. As Rwanda has an SWN for 4G, it is assumed that it will complete deployment for the vast majority of the population by 2030. Using Geospatial Analysis to Overhaul Connectivity Policies Enabling Policies Can Increase Network Coverage and Adoption 45 Figure 22 Figure 22. Location of 4G Location of coverage gains 4G coverage gains by by 2025 2025 due to policy due to policy reforms reforms Benin DRC Nigeria Ghana Sierra Leone Tanzania Change in 4G coverage of gamma scenario relative to alpha_plus p.p 0–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80 80–90 90–100 Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and Center for International Earth Science Information Network (CIESIN), household survey data and Group on Earth Observations. 46 Public Investment is Needed to Achieve universal Connectivity Using Geospatial Analysis to Overhaul Connectivity Policies 05 Public Investment is Needed to Achieve universal Connectivity Near-universal 4G coverage will require additional investment on the supply and demand sides, though there will remain some areas where alternative technologies are needed. While the policy reforms considered in this study the expected losses on unprofitable sites over a 10- would have a significant impact, it would still leave year period. In other words, it is the amount needed around 9 percent of the population in the seven to cover the capex and opex costs that cannot countries (just over 50 million people) without be recovered by expected revenues. It therefore mobile broadband coverage and around 17 percent minimizes the amount of additional investment of the population ages 10 and above (around 65 needed while maximizing the population that million people) without internet access by 2030. benefits from 4G coverage.36 The mechanism for To expand access and usage, other policies that allocation of subsidies is not modelled and further target the barriers not directly addressed in this analysis would be needed to assess alternatives such study, particularly around awareness and digital as ‘pay or play’ or reverse auctions for minimum skills, will therefore be important. However, even subsidies.37 with additional reforms, it is likely that additional The subsidies are calculated according to two investment will be required to fully achieve universal scenarios—one where the policy reforms discussed internet access by 2030 in each of the seven in Section 4 are applied and one where they are countries, especially among the lowest income groups not. The analysis shows that the policy reforms can with little purchasing power. Furthermore, for some save 10–20 percent of the cost required to achieve hard-to-reach areas other broadband technologies near-universal coverage. The results also show that need to be considered. for most countries, 55–65 percent of the total cost One potential policy option depending on the in unprofitable areas requires subsidization, with availability of public funds and the country and the rest covered by operators. In Sierra Leone, the sector situation is supply-side subsidies. Table 5 proportion is higher due to a lack of 4G demand shows the amount of subsidy required to extend and the proportion of sites that require high-cost 4G coverage to the majority of population. This is alternative backhaul solutions. calculated based on the amount needed to cover 36 Other investment models might cover a certain proportion of capex or opex. However, such an approach may overestimate the amount of additional investment needed. For example, if a public or private entity funded 50 percent of capex for all new sites, this might be more than is necessary if a site only requires 20 percent of capex to be funded to be profitable. 37 For a description of alternative ways of providing supply-side subsidies see ITU and World Bank (2020), Digital Regulation Handbook, Chapter 3. Using Geospatial Analysis to Overhaul Connectivity Policies Public Investment is Needed to Achieve universal Connectivity 47 Table 5. Amount of investment needed to provide near-universal 4G coverage 4G coverage 4G coverage Subsidy cost Subsidy cost Subsidy cost Proportion without with subsidy without policy with policy (US$ per of total subsidy* (%) reforms reforms capita) cost that is (%) (US$, millions) (US$, subsidized millions) (%) Nigeria** 95.8 99.5 461 407 1.98 65 Benin 95.4 99.7 37 30 2.47 57 Ghana 86.2 99.1 90 76 2.44 61 Tanzania 81.9 98.8 213 185 3.09 62 Sierra Leone 76.7 96.8 100 83 10.46 70 Democratic Republic of 57.6 94.3 380 296 3.31 61 Congo Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Subsidies are calculated based on the amount needed to cover expected losses on unprofitable sites over a 10-year period. Note: *4G coverage without subsidy assumes that the policy reforms discussed in Chapter 4 are applied. **Analysis for Nigeria assumes that coverage can be deployed in all states. If states with ongoing conflicts are removed from the analysis, it is expected that 4G coverage would reach 92 percent with a subsidy. An alternative, or at least complementary, approach This is the fundamental reason why 4G coverage will to expanding coverage is to focus on investment not be universal without further policy reform and and policy reforms that increase demand for 4G public investment and leapfrogging to 4G is unlikely services. This has the benefit of increasing 4G to occur in rural areas. However, this could change connectivity and making more areas profitable for if further reform or investment can stimulate large operators to deploy 4G networks. Table 6 shows increases in demand. Table 6 shows that if expected that in 2020, penetration of 4G mobile internet 4G penetration was 20 percent in uncovered areas, services was limited across all markets, ranging from operators would extend coverage to more than 0.5 percent in Democratic Republic of Congo to 6 89 percent of the population in all markets. If 4G percent in Nigeria. Over the next five years, this is penetration increased to 40 percent, 4G coverage expected to increase, but current forecasts indicate would almost reach the same levels as would be 4G penetration will not exceed 15 percent in any achieved with a pure infrastructure subsidy (as country. Furthermore, in rural areas, it is expected presented in Table 6). that 4G penetration will be less than 5 percent across all six countries by 2025. 48 Public Investment is Needed to Achieve universal Connectivity Using Geospatial Analysis to Overhaul Connectivity Policies Table 6. Expected 4G coverage with alternative demand scenarios (%) Current and forecast Expected 4G coverage… 4G penetration With no With 20% With 40% 2020 2025 change in penetration in penetration in demand uncovered areas uncovered areas Nigeria* 6 14 95.8 98.2 98.6 Benin 3 11 95.4 98.3 98.9 Ghana 5 15 86.2 95.0 96.7 Tanzania 2 8 81.9 95.4 95.9 Sierra Leone 2 9 76.7 90.5 93.5 Democratic Republic of 0.5 2 57.6 89.1 92.0 Congo Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. Note: *Analysis for Nigeria assumes that coverage can be deployed in all states. The analysis therefore highlights the importance Further research to better identify and evaluate of focusing on demand- as well as supply-side interventions that can affect mobile internet interventions. As 4G penetration remains relatively demand will therefore be important going forward, limited, especially in rural areas, any investment to enable widespread 4G coverage in Sub-Saharan that focuses solely on expanding coverage is Africa. These might include policies to enhance digital unlikely to generate a significant increase in mobile skills and literacy and the availability of relevant internet usage and reduce the internet uptake gap. content, interventions to expand access to electricity, The example of Rwanda illustrates this—in 2020, and/or interventions to improve access to mobile 4G internet penetration was around 2 percent devices. The latter could include replicating and (compared to 22 percent for 2.5/3G mobile internet), extending successful examples of device financing even though investment in the SWN has extended and/or subsidizing handsets directly—although a 4G coverage to more than 98 percent of the subsidy requires careful design to ensure that it is country’s population. targeted and used only by individuals who cannot otherwise access a device. However, there is a risk The modelling carried out in this study suggests of the subsidy crowding out private investment. that even if 4G infrastructure is subsidized to cover The operation of energy service companies (ESCOs) almost all the population in each of the other six could facilitate the provision of electricity to mobile countries, mobile internet adoption would only equipment and unconnected communities. With increase by 1–3 percentage points by 2030, as regard to improving the availability of content, demand would remain limited. Low penetration levels governments can take steps to enable and promote also pose a risk to the viability and sustainability online content, for example, mobile education and of any public investment in expanding coverage, health services. In partnership with the private sector especially if expected revenues are insufficient to and education institutions, governments can support cover opex costs in the long term. programs to increase awareness, digital skills, and literacy. Using Geospatial Analysis to Overhaul Connectivity Policies Public Investment is Needed to Achieve universal Connectivity 49 50 Public Investment is Needed to Achieve universal Connectivity Using Geospatial Analysis to Overhaul Connectivity Policies Why does 4G coverage not reach 100 percent with additional investment? In all the six countries,38 there is still a segment of coverage to 93.5 percent. At a certain point, the population that is unlikely to have coverage. the number of sites needed to expand coverage To understand the reasons for this, and why increases exponentially, as settlements become these can differ by country, Figure 23 shows much smaller and sparse. For example, to expand the number of additional sites needed for each coverage from 98 percent to 99 percent, almost coverage increment in Sierra Leone. This is the 800 sites would be needed. smallest country considered in the study in terms This is also reflected in the cost per covered of population, but there are mountainous areas person. Figure 24 shows that when expanding that can make it difficult to expand coverage. coverage from 85 percent to 86 percent, the To simplify the analysis and interpretation, the cost per covered person is just over US$10. This analysis focuses on expanding mobile coverage in increases to around US$30 when reaching 90 greenfield areas. At the start of 2020, there were percent coverage and then increases exponentially 570 mobile sites in Sierra Leone that covered just once coverage reaches around 96 percent. This is over 85 percent of the country’s population with the level of 4G coverage that is assumed would be at least 2G technology. The next 100 additional reached with additional investment in Table 5. sites would expand coverage to around 92 percent, the subsequent 100 sites would increase while 28 Figure Number of sites needed to achieve near-universal mobile coverage in Sierra Leone Figure 23. Number of sites needed to achieve near-universal mobile coverage in Sierra Leone 100% 99% 2,734 98% 97% 96% 95% Network coverage 94% 93% 92% 91% 90% 89% 88% 87% 86% 85% 570 570 670 770 870 970 1,070 1,170 1,270 1,370 1,470 1,570 1,670 1,770 1,870 1,970 2,070 2,170 2,270 2,370 2,470 2,570 2,670 2,770 Total number of sites Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data, and Group on Earth Observations. 38 Rwanda is not included as it is assumed that the single wholesale 4G provider will expand 4G coverage to almost all of the country’s population by 2030. Using Geospatial Analysis to Overhaul Connectivity Policies Public Investment is Needed to Achieve universal Connectivity 51 Figure 29 Figure 24. Cost per covered person in greenfield sites in Sierra Leone Cost per covered person in greenfield sites in Sierra Leone 1200 1100 1000 900 Cost per covered person ($) 800 700 600 500 400 300 200 100 0 86% 87% 88% 89% 90% 91% 92% 93% 94% 95% 96% 97% 98% 99% 100% Population coverage Source: GSMA analysis of data sourced from mobile operators, GSMA Intelligence, Facebook Connectivity Lab and CIESIN, household survey data and Group on Earth Observations. Cost is calculated based on the 10-year NPV of deploying greenfield sites. While it is technically possible to achieve universal the sites require satellite backhaul, for which coverage with existing technologies, modelling the current costs of deploying 3G and 4G can shows that, at a certain point, the amount of be prohibitive when there is a lack of demand. additional investment needed becomes an order Providing coverage in these areas in a sustainable of magnitude (or sometimes several orders of manner will therefore likely require new innovations magnitude) higher. These are areas that are that are, thus far, yet to be developed or are remote and sparsely populated, with some currently unproven on a large-scale, commercial sites covering no more than a few 100 people basis. One potential solution could be low-earth (sometimes even fewer). orbit satellite constellations which could help significantly reduce the cost of backhaul or Expanding coverage to these locations will even communicate directly from the satellite to therefore be extremely challenging using existing normal existing handsets without the need of any technologies, due to a combination of low terrestrial infrastructure.39 Other solutions could population density and high costs—many of include wide-area cell sites.40 39 For example, Lynk or project Starlink by SpaceX. 40 For example, Altaeros SuperTowers. 52 Conclusions Using Geospatial Analysis to Overhaul Connectivity Policies 06 Conclusions This study focused on how mobile technologies could • Extending mobile broadband to areas with support connectivity in Africa. It had two objectives. no coverage presents a substantial economic The first was to map mobile coverage and adoption challenge. It will require efforts to reduce at a settlement level across seven countries in deployment costs and, more importantly, increase Sub-Saharan Africa that captured the different demand. Both are contingent on continued, socioeconomic and regulatory challenges facing the collaborative action by all stakeholders, building region. The second was to simulate the effects of on private sector innovation, which in recent different policies using granular data on both the years has driven significant cost reductions in location of mobile infrastructure and demand. This rural network deployments, as well as in handset enables more precise calculations and therefore a and data prices. deeper understanding of the impacts that policy • While 3G and 4G coverage are lagging at 74 reforms can have on both coverage and adoption, as percent and 48 percent of the population, well as the additional investment needed to achieve respectively, they could catch up with 2G coverage universal connectivity by 2030. in the coming years if spectrum policy is updated The findings from this study can be used to so that operators have access to sufficient and inform policy making in other countries, while the affordable spectrum in sub-1 GHz bands to analysis can also be leveraged to understand the use the spectrum more efficiently, including impact of policies and investments in different the refarming of existing spectrum so that it geographic locations. Given the extent to which is technology neutral. For example, across four mobile connectivity varies within countries, it is countries where not all operators have access important to continue mapping and track coverage to the sub-1 GHz spectrum for 4G (in Ghana, and adoption at a local level to ensure interventions Nigeria, Sierra Leone, and Tanzania), an operator are targeted where they are needed, to ensure their could increase rural 4G coverage by more than 5 effectiveness. It is also worth noting that addressing percentage points (or 7.5 million people) if they connectivity challenges in Africa require the use of are able to use 700 or 800 MHz spectrum for 4G various technologies that respond to demand needs (assuming existing spectrum fees apply). in a policy environment that allows for innovation on • Infrastructure sharing at the site level would a level playing field. enable coverage to expand while maintaining The simulations in seven countries bring out service-level competition. Active sharing would clear and important messages for governments, allow 2–10 percent of the rural population, the private sector, and the international donor depending on the country, to have mobile community: broadband coverage from more than one operator. Policy makers that are considering SWNs should • Mobile operators are very close to the market also consider that active sharing can deliver frontier for 2G coverage, with at least 87 similar levels of coverage while maintaining a percent of the population covered across the greater degree of service competition (provided seven countries. A limited amount of additional competition safeguards are in place) while 2G coverage can be provided by the private sector avoiding the risks of creating a monopoly at the that is financially viable and sustainable in current infrastructure level, like is the case with SWNs. market conditions. Almost all the uncovered areas are in rural, often remote, locations. Using Geospatial Analysis to Overhaul Connectivity Policies Conclusions 53 • Aligning tax policy with best-practice any of the seven countries, while in rural areas principles and removing distortive sector- it is expected that 4G penetration will be less specific taxes that are solely applied to the than 5 percent in every country by 2025. Lack mobile sector will likely improve investment of demand is the fundamental reason why incentives for operators as well as affordability universal 4G coverage will be challenging for consumers. This includes the removal of excise without further policy reform and public duties on handsets and mobile services that are investment, and why ‘leapfrogging’ to 4G is not applied to other goods and services, as well as unlikely to occur in rural areas. However, this the reduction of taxes levied on mobile operators could change if further reform and/or investment but not on other firms in the economy. This can stimulate large increases in demand. If reform would expand mobile broadband coverage expected 4G penetration was 20 percent in by up to 4 percentage points, depending on the uncovered (mostly rural) areas, operators would country, and would bring more than 30 million extend 4G coverage to more than 89 percent people online by 2030 (increasing mobile internet of the population in all seven markets. If 4G adoption by 6 percentage points). penetration increased to 40 percent, 4G coverage would almost reach the same levels as would • While these policy reforms would drive important be achieved with a pure infrastructure subsidy. gains in connectivity, by 2030 it would still leave Further research to better identify and evaluate 9 percent of the population without mobile interventions that can have an impact on mobile broadband coverage and 17 percent of the internet demand will therefore be important going population ages 10 and above without internet forward, to enable widespread 4G coverage in access. Additional interventions on the supply Sub-Saharan Africa. These could include policies and demand side are still needed to make to enhance digital skills and literacy and the mobile broadband coverage and adoption availability of relevant content, interventions to universal by 2030. In the seven countries, with expand access to electricity and other assets for the policy reforms in place, around US$1.7 billion more productive use of internet, and interventions of additional investment would cover the vast to improve access to mobile devices. majority of populations with 4G networks, of which almost 40 percent could be funded by • Many of the policy findings are likely to apply the private sector, leaving an investment gap in other countries with similar characteristics. of US$1.1 billion. Without the policy reforms For example, the impact of active infrastructure highlighted above, the investment gap would be sharing and spectrum refarming will be US$1.3 billion, meaning policy reforms can save particularly relevant to countries with large around 15 percent of the additional investment coverage gaps and/or large populations that needed to achieve near-universal coverage. live in dispersed and remote areas. For countries that have achieved high coverage but where • An alternative, or at least complementary, a significant usage gap persists, policies that approach to expanding 4G coverage would be increase demand (for example, by improving to focus on additional policy reforms and affordability and access to devices) will be most investment that stimulate demand. Over the impactful. next five years, current forecasts indicate that 4G penetration will not exceed 15 percent in 54 Appendix Using Geospatial Analysis to Overhaul Connectivity Policies Appendix A: Methodology This appendix presents the methodology used to not covered. Therefore, we consider that investing in model the effects of public policy options for mobile middle-mile capacity is generally not a requirement infrastructure deployment in seven Sub-Saharan to increase coverage. However, maintaining a Africa countries. The model is focused on the ‘last good quality of service and keeping data bundles mile’ of infrastructure, that is, the mobile site that affordable will require investing in backbone connects with the end user as well as the backhaul infrastructure to transport larger volumes of data link that connects sites to the core network (see from rural areas as they increase to reach the levels Figure A1). It assesses the incremental profitability of observed in urban areas. expanding networks, which affects network coverage This report focuses on the evolution of coverage and and adoption. While investments in the ‘first mile’ usage, making sure that enough capacity is deployed (for example, international cables) and ‘middle mile’ at the last mile to carry the user traffic. Hence, it (for example, backbone and IXPs) are important in does not consider policy reforms or interventions terms of increasing network capacity, especially in that would increase investment in the core network, urban areas, based on the current and expected first mile, or middle mile, which should be the subject levels of data usage in rural areas across the seven of a separate analysis. countries, mobile operators normally have sufficient network capacity to meet demand in areas that are Figure A1. Last mile First mile Middle mile Edge Backhaul Last mile End user (international (backbone, IXPs, (PoPs, caches) (microwave, fibre, (site) cables) data centres) satellite) Note: PoP = Point of presence There are four modules that underpin the analysis: • A supply module that contains an inventory of • A market and public policy module that estimates existing mobile network infrastructure what level of mobile coverage (for 2G, 3G, and 4G technologies) could be achieved by mobile • A demand module that incorporates operators with and without policy reforms and socioeconomic data to estimate demand for public sector or third-party intervention, based on mobile voice and data within each country expected demand and deployment costs. • An economic and engineering module that combines demand and supply data to estimate the necessary infrastructure and cost of expanding mobile networks in uncovered areas Using Geospatial Analysis to Overhaul Connectivity Policies Appendix 55 Supply To produce the aggregate coverage data for each Irregular Terrain Model (ITM), also known as the country, network infrastructure data were collected Longley Rice Model,41 optimized to deliver accurate from larger MNOs. For each individual relay site, we results in rural and peri-urban areas. collected the following parameters: The ITM uses two sets of input variables. The first • Location in geographical coordinates are the technical parameters of each individual relay site that we collected from operators. The second • Height of the tower hosting the antennas are the characteristics of the transmission medium, • Signal emitting power such as the terrain profile and type of vegetation • Antenna parameters such as the gain, azimuth, in the area.42 The output of the ITM model is a and tilt geocoded image showing the area covered with signal strength above a predefined threshold (see • Frequency band used Figure A2 as an example). Combining the coverage • Type of technology available (2G, 3G, or 4G) provided by each individual site for every operator in the country, we obtain the countrywide coverage • Date of deployment. footprint for each technology (2G, 3G, and 4G). We calculate the coverage of each relay site using a Overlaying these countrywide coverage footprints Radio Propagation Model (RPM). RPMs are empirical with the population, we are able to estimate the mathematical models widely used by operators for coverage status for each individual settlement in the planning the setup of their networks, allowing them country. A settlement is assumed to have coverage to plan the location and characteristics of each relay if it receives at least the medium signal strength site so as to maximize coverage and decrease costs. for the relevant technology. The predefined signal There are several RPMs available that are optimized strength thresholds that we use are presented in for specific settings or technologies. We use an Table A1. Figure A2. Table A1. Area covered with signal strength Signal strength thresholds Relay site Radio Medium signal Strong signal Low (no coverage) technology strength strength Medium Strong 2G -85 dBm -73 dBm 3G -91 dBm -83 dBm 4G -105 dBm -95 dBm Source: GSMA Source: GSMA 41 The ITM model used is based on version 7 of the algorithm released the 26/06/2007 by the National Telecommunications and Information Administration (NTIA) - very similar to version 1.2.2. 42 This is sourced from the Advanced Land Observing Satellite (ALOS) Global Digital Surface Model. 56 Appendix Using Geospatial Analysis to Overhaul Connectivity Policies Demand Demand is estimated based on the probability of To estimate demand in each settlement (including mobile and mobile internet adoption, which we potential demand for settlements currently without calculate at a settlement level based on a function coverage), an econometric model was used to of socioeconomic and demographic characteristics. estimate mobile ownership and mobile internet The following data are used: use. We constructed a model of the probability of adopting mobile phone ownership (simi ) and mobile • Settlements. To overlap network data with internet capability (mii ) conditional upon a vector of population settlements, we use data sourced from independent variables (x). the High-Resolution Settlement Layer database,43 which provide estimates of human population We define a latent independent variable ( yi*) as distribution based on census data and high- resolution satellite imagery. The data identify population agglomerates according to an image y* i (X; β0 ,β) ≡ β0+ βX+εi recognition algorithm. These data give the location and density of the population of a given country. • Household survey data. As part of the study, we This latent variable determines the outcome variable carried out most demand analysis using recent ( yi ) for each individual i in the following way: survey data from Gallup World Poll. Analysis was also checked using other household survey data, including the Ghana Living Standards Survey, Rwanda Integrated Household Living Conditions yi= { 10 if if y* < 0 y* > 0 i i where yi { sim mi i i Survey, Sierra Leone Integrated Household Survey, Tanzania Household Budget Survey, and Nigeria Living Measurement Study. By assuming that the errors follow a logistic • Demographics. We used UN-adjusted distribution (i.i.d.), we used a logit model to estimate unconstrained population data sourced from the conditional probabilities of the two mobile WorldPop44 to calculate the distribution of adoption outcomes. Standard errors were clustered settlement populations in terms of age and at a country ‘region’ level. The independent variables gender. used in the model included gender, age, whether the individual lived in a rural or urban location, income • Night light data. We used Visible Infrared Imaging quintile, employment, education, and electricity Radiometer Suite (VIIRS)45 data to assign a light access (where data were available). Using the radiance value to each settlement. This was used coefficients of the regressions in each country, as a proxy indicator for income and electricity we estimated the probability of demand in each access in each settlement. settlement. 43 https://data.humdata.org/organization/facebook. 44 https://data.humdata.org/organization/worldpop. 45 https://ngdc.noaa.gov/eog/viirs/download_dnb_composites.html. Using Geospatial Analysis to Overhaul Connectivity Policies Appendix 57 As a final step, estimated penetration rates in each The relative differences between urban and rural settlement were then adjusted so that the total ARPU were sourced from survey data gathered number of mobile and mobile internet subscribers in by GSMA Intelligence and, where available, from covered areas matched country-level estimates for operators and national statistical offices. A further unique mobile subscribers and unique mobile internet input was required in terms of the incremental ARPU subscribers. The country-level data were sourced gained when a customer upgrades to a 3G and from GSMA Intelligence and verified with data 4G connection (compared to being on 2G). These from the national regulator where available. This assumptions (ARPU differentials by rural-urban step ensured that the aggregate level of demand geography and technology) were based on analysis is accurate—and that it is distributed across the of mobile spend in the GSMA Intelligence Consumers country based on the relevant demand drivers and in Focus Survey and, where possible, verified with where coverage exists. mobile operators. To estimate revenues, assumptions on ARPU Mobile and mobile internet ARPU were then were required. These were initially sourced from calibrated so that total estimated revenues in GSMA Intelligence, from data reported by mobile covered areas matched the latest annual market operators (this was verified with data published recurring revenues, which were sourced from GSMA by national regulators where available). To reflect Intelligence and mobile operators (or otherwise data spatial variation, we adjusted ARPU according to from the national regulator). whether the settlement is in a rural or urban area. Economic and engineering module To estimate the cost of expanding network coverage, The profitability algorithm searches for the best we combined the supply and demand modules to combination of sites and site configurations that produce a list of new sites and site upgrades (for maximizes the NPV of each site. Site overlaps 3G and 4G network expansion). These site locations (sites whose coverage overlaps with other sites) were optimized according to a profitability algorithm, were considered. This provides an optimal network such that they reflect the sites that operators are deployment to extend coverage to all populations most likely to deploy. The algorithm was designed currently uncovered by any technology. to find the optimal combination of sites that When modelling the deployment or upgrade to 3G or maximizes the NPV of investment. The starting 4G, a similar algorithm was applied to calculate the point is a set of greenfield site simulations, where best combination of site upgrades that maximized we model hypothetical new sites in sub-settlements the NPV of investment among existing (brownfield) that do not have coverage.46 For each of these new site upgrades and new greenfield sites. The algorithm sites, we modelled four different site configurations was run assuming different spectrum bands for (as explained in Chapter 3), with an associated deployment. capex and opex that was sourced from operators and equipment vendors. To estimate the coverage of each site, we applied the propagation model described in the supply module above. 46 Due to the large number of settlements in a country, we limit the number of settlements where we simulate the deployment of new sites. This limit changes for each country but is roughly close to settlements where there are more than 50 uncovered people. Note that even sites below this limit can receive coverage from neighbouring settlements. For every country, we ensure that coverage simulations reach at least 99 percent of the population. 58 Appendix Using Geospatial Analysis to Overhaul Connectivity Policies Market and policy module To develop a model that quantifies the impact on Infrastructure sharing coverage and adoption of different public policies, • No sharing. When modelling the expected the outputs of the economic and engineering module coverage assuming more than one network, this were used to assess the impact of various policies or scenario assumes that additional networks incur changes in the market. The starting point is the list the same capex and network opex, meaning, for of brownfield and greenfield sites that would enable example, that if two networks are modelled in coverage to reach the uncovered populations in each uncovered areas then the amount of capex and country for 2G, 3G, and 4G networks. network opex per site doubles compared to a single network. The model also includes variable For brownfield sites, which already have 2G, the costs, which are estimated based on revenues, analysis considers the costs of upgrading to 3G and capacity costs that depend on the number of or 4G (or both) and determines whether the site users and traffic. is financially viable to generate a positive return, given expected demand and ARPU. If it is, the model • Passive sharing. This assumes that the passive assumes the site will be upgraded and coverage elements of a new site (including the cost of land, will expand. The model selects the most profitable tower, and power) are only incurred once, even if of three potential upgrades—3G, 4G, or 3G+4G multiple networks are assumed. However, an uplift combined (unless the site only needs to be upgraded is applied based on the number of carriers sharing to 4G). the infrastructure. The uplift is sourced from information from operators. In terms of greenfield sites, the model takes the deployment of 2G as the starting point. It • Active sharing. This assumes that the active determines whether a site is financially viable, given elements of a new site (including radio and expected demand and returns for 2G voice and data. backhaul equipment) are only incurred once even It then separately assesses the incremental cost if multiple networks are assumed. Again, an uplift and revenue of deploying 3G and/or 4G at the site. is applied based on the number of carriers sharing The final deployment decision is then based on the the infrastructure. The uplift is sourced from following: information from operators. • If 2G is profitable and 3G/4G is not profitable, the Spectrum site is assumed to be 2G only. To model the impact of deploying different bands, we run the 3G/4G supply and demand analysis for both • If a 2G site with 3G/4G is more profitable than brownfield and greenfield sites based on different a 2G-only site, the site is assumed to have 2G spectrum scenarios. Where lower bands are used, the as well as 3G/4G technology (the latter will be population covered for each site will be higher than determined based on the most profitable of the that when higher frequency bands are used. three potential upgrades, similar to the approach for brownfield sites). Taxation Figure A3 presents how the analysis models tax • If there is no profitable deployment option (that is, policy reforms. Taxes can either apply to consumers both 2G and 2G+3G/4G have a negative NPV), it is or operators and are a function of the tax base assumed that the site will not be built. (for example, recurring revenues, accounting profits, The results of this analysis determine the additional and import value) and the relevant tax rate. When coverage that could be provided by the market modelling the impact of a tax change, adjustments without any further policy reforms or interventions to the tax rates—whereby the sector-specific (as set out in Chapters 2 and 3). The following operator and consumer taxes are either reduced or policies can then be modelled to determine the increased—are applied and the difference between impact on both coverage and adoption. total tax payments in the baseline and scenario are combined with pass-through rates to prices and demand elasticities to adjust for impacts on usage and adoption. Using Geospatial Analysis to Overhaul Connectivity Policies Appendix 59 Figure A3. Modelling the impact of tax reforms Tax or fee change Pass-through Revenue to consumers Revenue to operators Change in price Increased returns Elasticities Change in usage and ARPUs Investment Change in adoption Increased coverage Source: GSMA Pass-through and demand elasticity assumptions three types of tax: consumer taxes, operator taxes, are currently based on previous taxation studies and taxes on equipment.48 The pass-through rates carried out by GSMA and Ernst & Young (EY).47 are based on results from the previous taxation studies carried out by GSMA and EY, which Given the change in tax payments in the baseline estimate the impact on prices through tax shocks in and scenario, operators will make the decision on computable general equilibrium (CGE) models (based how much of this to retain or ‘pass-through’ to on data from the Global Trade Analysis Project). prices. We vary the pass-through for changes in 47 For further details, see https://www.gsma.com/publicpolicy/regulatory-environment/taxation; GSMA and Ernst & Young. 2020. ‘Mobile Taxation Studies: Methodology Documentation. 48 For example, import duties. 60 Appendix Using Geospatial Analysis to Overhaul Connectivity Policies We take an average of the pass-through rates for Once the pass-through is established, we model the the studies that have been carried out on countries impact of changes in prices on consumer demand, in Sub-Saharan Africa49 by the above tax types. We split into usage elasticities based on the price of assume price changes are passed through equally data and voice, ownership elasticities based on the to the prices of each type of service provided. In line price of handsets and services, and 2G to mobile with the previous GSMA/EY studies, pass-through broadband migration elasticities based on the prices rate assumptions are based on and applied to of data and mobile broadband handsets. These ‘effective prices’, which are calculated using recurring elasticities are based on a 2019 literature review as revenues. A change in the effective price reflects part of the previous GSMA/EY taxation studies. This an overall benefit to consumers and does not literature review was limited to studies in the past necessarily imply a reduction in like-for-like prices. 15 years, and we use the average for low-income For example, it could instead reflect an increase in countries.51 the quality of the service received; an increase in the number of included minutes, messages, and/or Table A3. data provided at the same price; or a combination of Elasticity assumptions these outcomes. Tax type Change Change Change in Table A2. in price in price price of Pass-through assumptions of voice of data handsets Tax type Pass-through Usage −0.84 −1.11 — rate (%) Adoption −0.90 — −1.30 Consumer 90 Technology Operator 85 migration Operator - taxes on equipment50 0 (2G to — −0.32 −0.47 mobile broadband) By estimating the change in the price of services (which we calculate by multiplying the changes in tax revenues by the pass-through rates), and subsequently applying the above elasticity assumptions, we calculate new levels of usage and adoption. 49 Sub-Saharan countries included in these studies are DRC, Guinea, Madagascar, Tanzania, and Zambia. 50 For taxes on equipment we assume these influence the decision to buy equipment or to build a new base station, with limited impacts on the final consumer price. 51 For further details, see GSMA and Ernst & Young. 2020. Mobile Taxation Studies: Methodology Documentation. Using Geospatial Analysis to Overhaul Connectivity Policies Appendix 61 Appendix B: Acronyms and Abbreviations ARPU Average Revenue Per User LTE Long-term Evolution (4G) Capex Capital Expenditure MOCN Multi-Operator Core Network CIESIN Center for International Earth Science MORAN Multi-Operator Radio Access Network Information Network MNO Mobile Network Operator DRC Democratic Republic of Congo NPV Net Present Value ESCO Energy Service Companies Opex Operating expenditure EY Ernst & Young RAN Radio Access Network GDP Gross Domestic Product RPM Radio Propagation Model GSMA Global System for Mobile Communications SDG Sustainable Development Goal GNI Gross National Income SIDA Swedish International Development Cooperation Agency HRSL High Resolution Settlement Layer SMS Short Message Service ICT Information and Communications Technology SRAN Single Radio Access Network IMF International Monetary Fund SWN Single Wholesale Network ITM Irregular Terrain Model UN United Nations ITU International Telecommunication VAT Value Added Tax Union WACC Weighted Average Cost of Capital IXP Internet eXchange Point 62 Appendix Using Geospatial Analysis to Overhaul Connectivity Policies References • Bahia, Kalvin; Castells, Pau; Cruz, Genaro; Masaki, Takaaki; Pedros, Xavier; Pfutze, Tobias; Rodriguez-Castelan, Carlos; Winkler, Hernan. 2020. The Welfare Effects of Mobile Broadband Internet: Evidence from Nigeria. Policy Research Working Paper;No. 9230. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/33712 License: CC BY 3.0 IGO. • Bahia, Kalvin; Castells, Pau; Cruz, Genaro; Masaki, Takaaki; Rodriguez- Castelan, Carlos; Sanfelice, Viviane. 2021. Mobile Broadband Internet, Poverty and Labor Outcomes in Tanzania. Policy Research Working Paper;No. 9749. World Bank, Washington, DC. © World Bank. https:// openknowledge.worldbank.org/handle/10986/36172 License: CC BY 3.0 IGO. • Broadband Commission. 2019. Connecting Africa Through Broadband A Strategy for Doubling Connectivity by 2021 and Reaching Universal Access by 2030 • Calderon, Cesar; Cantu, Catalina. 2021. The Impact of Digital Infrastructure on African Development. Policy Research Working Paper;No. 9853. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank. org/handle/10986/36629 License: CC BY 3.0 IGO. • Dutz, Mark; Begazo, Tania; Blimpo, Moussa. Forthcoming. Technological Transformation for Jobs in Africa: How Digital Development Can Support Good Jobs for All. • Facebook Connectivity Lab and Center for International Earth Science Information Network - CIESIN - Columbia University. 2016. High Resolution Settlement Layer (HRSL). Source imagery for HRSL © 2016 DigitalGlobe. • GSMA.2018. Enabling Rural Coverage: Regulatory and Policy Recommendations to Foster Mobile Broadband Coverage in Developing Countries. • GSMA. 2019a. State of Mobile Internet Connectivity Report. • GSMA. 2019b. Closing the Coverage Gap: How Innovation Can Drive Rural Connectivity. • GSMA. 2019c. Rethinking Mobile Taxation to Improve Connectivity. • GSMA. 2020a. Mobile Technology and Economic Growth. • GSMA. 2020b. Effective Spectrum Pricing in Africa: How Successful Awards can Help Drive Mobile Connectivity. Using Geospatial Analysis to Overhaul Connectivity Policies Appendix 63 • GSMA. 2020c. State of Mobile Internet Connectivity Report. • GSMA. 2021. “KaiOS, Enabling Accessible and Affordable Internet for All with the GSMA’s MISTT Module.” • GSMA Intelligence database (2020). • GSMA Intelligence Consumer Survey for Nigeria and Tanzania. • GSMA and ABI. 2018.‘Mobile Backhaul Options: Spectrum Analysis and Recommendations’. • GSMA and Ernst and Young. 2020. Mobile Taxation Studies: Methodology Documentation. • GSMA and NERA. 2017. Effective Spectrum Pricing: Supporting Better Quality and More Affordable Mobile Services • GSMA and World Bank. 2020. The Poverty Reduction Effects of Mobile Broadband in Africa: Evidence from Nigeria. • ITU. 2020. Measuring Digital Development: Facts and Figures 2020. • ITU. 2019. Economic Contribution of Broadband, Digitization and ICT Regulation: Econometric Modelling for Africa. • MTN. 2018. “MTN, KaiOS Technologies, China Mobile and UNISOC Launch Africa’s First Smart Feature Phone.” • Korea Telecom Rwanda Networks, n.d. “Network upgrade and new product launch.” https://www.ktrn.rw/single-post/network-upgrade-and-new- product-launch • Oughton, Edward (2022, forthcoming), “Policy Options for Digital Infrastructure Strategies: A Simulation Model for Broadband Universal Service in Africa”. • Orange. 2019. “Orange launches « Sanza », the Phone which Democratizes Access to Internet in Africa.” • Mobile Infrastructure Sharing. 2012. GSMA. • ITU and World Bank. 2020. Digital Regulation Handbook. • Rotondi, V., Kashyap, R., Pesando, L.M., Spinelli, S. and Billari, F.C., 2020. Leveraging mobile phones to attain sustainable development. Proceedings of the National Academy of Sciences, 117(24), pp.13413-13420.