Policy Research Working Paper 9590 World Development Report 202 1 Background Paper A Demand-Side View of Mobile Internet Adoption in the Global South Rong Chen Development Economics World Development Report 2021 Team March 2021 Policy Research Working Paper 9590 Abstract Mobile technologies show great potential to accelerate inter- likely to adopt mobile internet. Social network effects are net access and usage, especially in developing countries. A found to have a significant positive impact on the usage of better understanding of key drivers and main constraints mobile internet. Those who have more close friends using for mobile internet access is the first prerequisite for gov- an online social network are more likely to adopt mobile ernments to design targeted policy solutions. This study internet. Individuals whose five closest friends are using exploits a household survey that collects information on an online social network (such as Facebook or Twitter) are information and communications technology access and 63.1 percent more likely to adopt it than those without any usage at the household and individual levels in 22 countries close friends using such online social network sites/apps. in the Global South. The study finds that in addition to Across regions, although the factors affecting the adoption infrastructure investment, which has been the main focus of of mobile internet remain largely the same, the magnitudes many developing countries, other demand-side factors are of their impacts vary. In Asia, gender differences are nega- of critical importance. Across the developing world, females, tively associated with mobile internet. In Africa, the impact the elderly, those who live in rural areas, and those who of education level is more salient than in the other two have a relatively low level of income or education are less regions, implying an urgent need to improve digital literacy. This paper is a product of the World Bank’s World Development Report 2021 Team, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/ prwp. The author may be contacted at rchen5@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team A Demand-Side View of Mobile Internet Adoption in the Global South 1 Rong Chen Keywords: Internet adoption, mobile broadband, developing economy, digital gap. JEL codes: O12, I30, L96, O55. 1 This research was undertaken as a Background Paper prepared for Chapter 5 of the World Development Report 2021 Data for Better Lives. The author would like to thank the following colleagues from the World Bank Group for their review and valuable feedback on the paper: Mark Williams (Practice Manager, Digital Development Global Practice), Carlos Rodriguez Castelan (Lead Economist, Poverty and Equity Global Practice) and Takaaki Masaki (Economist, Poverty and Equity Global Practice). The author is also grateful to participants to the InfraXchange seminar: A Demand Side View of Mobile Internet Adoption in the Global South, on November 19th, 2020 for their comments and suggestions. The author appreciates RIA (Research ICT Africa), LIRNEasia (Learning Initiative for Network Economies in Asia), and DIRSI (el Diálogo Regional sobre la Sociedad de la Información / Regional Dialogue on the Information Society) for sharing survey data to support the analyses. 1. Introduction Since its inception in the 1990s, the internet has transformed the ways people interact, businesses conduct commercial transactions, and governments deliver public services. Internet penetration—on both the supply side (infrastructure deployment) and demand side (user adoption)—continues to improve. International internet bandwidth nearly tripled between 2015 and 2019, to 466 terabytes per second. 2 The share of the global population using the internet increased from 8.0 percent in 2001 to 51.4 percent in 2018. 3 However, the penetration of fixed broadband internet is far from satisfactory in low-income countries, where, in 2018, fixed broadband subscriptions averaged 0.74 per 100 people. 4 In this context, mobile technologies show great potential to accelerate internet access and usage. Global mobile cellular subscriptions increased more than 10 times—from 0.74 billion in 2000 to 7.9 billion in 2018. 5 In low- and middle-income countries, mobile cellular subscriptions also reached a relatively high level of 102 per 100 people in 2018. 6 (However, it is worth noting that the unique mobile subscriber penetration rate, as a share of population, was still as low as 45 percent in the Sub-Saharan Africa region in 2018. 7) It is estimated that by 2020, 9 in 10 people will be covered by 3G networks that enable internet access from the palms of their hands (World Bank 2018). The increasing rate of mobile phone ownership in combination with expanding 3G network coverage promises to promote access to and usage of mobile internet. Research findings have provided evidence on the development impacts of mobile internet access and usage. Katz and Callorda (2018) estimate that a 10 percent increase in mobile broadband penetration is associated with a 1.8 percent increase in gross domestic product (GDP) in middle-income countries, and a 2 percent increase in GDP in low-income countries. Mobile broadband coverage is also shown to have large and positive impacts on household consumption levels (Bahia et al. 2020). Mobile internet access and usage are also found to have a positive impact on people’s happiness and well-being (GSMA and Gallup 2018) and women’s empowerment (Bailur and Masiero 2017). It is argued that access to and usage of information and communication technology (ICT) reduce poverty, by fostering access to and exchange of information, and improving the transparency and accessibility of public services—all benefits that would also apply to mobile internet (Cecchini and Scott 2003; Roller and Waverman 2001; Waverman, Meschi, and Fuss 2005). Despite the potential of mobile internet to help achieve development goals, there is a significant digital divide among and within countries. According to the Organisation for Economic Co-operation and Development, the digital divide refers to the “gap between individuals, households, businesses and geographic areas at different socio-economic levels with regard to both their opportunities to access ICTs and to their use of the Internet for a wide variety of activities”. According to the Global System for Mobile Communications (GSMA 2019a), those who are not connected through mobile internet are disproportionately rural, women, or illiterate. For instance, in developing countries, women still remain 10 percent less likely than men to own a mobile phone, and 23 percent less likely than men to use mobile internet services. Similarly, rural populations are 40 percent less likely to use mobile internet than urban populations (GSMA 2019b). 2 TeleGeography. 3 ITU statistics. 4 World Development Indicators. 5 World Development Indicators. 6 World Development Indicators. 7 Statista.com. 2 Some studies consider the factors affecting fixed-internet access and usage and other aspects of the digital divide. For instance, income, installation fees, and age are shown to be significant factors predicting fixed- internet usage (Birba and Diagne 2012; Cerno and Amaral 2006; Katz and Rice 2003). Other socioeconomic variables such as education, gender, locality and household size are also found to be drivers for internet adoption and usage in general. For instance, Goldfarb and Prince (2008) find that low income, less-educated people spend more time online. Similarly, Penard et al. (2012) and Penard et al (2015) find that education and computer literacy increase internet usage. Gilwald et al. (2018) provide evidence on lower internet adoption rate among women and those who live in rural areas. Other infrastructure factors such as increasing the distribution of electricity and improving competition in digital infrastructure are found to help promote the adoption of internet services (Armey and Hosman 2016; Rodriguez-Castelan et al. 2019). However, with regards to mobile internet, existing studies have a few limitations. First, previous research has mainly focused on understanding the determinants of mobile phone ownership in developing countries. Education, employment status, and type of electricity are found to be important factors (Aker and Mbiti 2010; Björkegren 2019; Forenbacher et al. 2019; van Biljon and Kotzé 2007). Second, studies on mobile internet adoption tend to focus on the supply side, emphasizing technological aspects such as the efficient compression of images or data delivery (Kim and Kim 2002). Funk (2005) presents how certain technological products such as the push mail service and micropayment systems promote mobile internet adoption. Third, only specific country or regional data are used in the few studies analyzing the demand- side factors affecting mobile internet adoption. For instance, Srinuan, Srinuan, and Bohlin (2012) use data from Thailand to find that price, the availability of fixed telephony, and individuals’ age and location are strong determinants of mobile internet adoption. Consumers’ perspectives on service applications are also found to affect mobile internet adoption in Taiwan, China (Hsu, Lu, and Hsu 2007). Hasbi and Dubus (2020) provide evidence of the positive impact of being part of an online social community on mobile broadband use in Sub-Saharan Africa. Meanwhile, a cross-country/region analysis that presents commonalities and differences in how specific demand-side factors affect mobile internet adoption across developing countries is missing. This study aims to (1) examine how socioeconomic variables and perception factors affect mobile internet adoption; and (2) compare differences in mobile internet adoption across different countries, regions, and demographic profiles from the demand side. The study adds to the literature in the following ways: (1) it is one of the first studies to explore determinants of mobile internet adoption in developing countries, (2) it uses nationally representative household-level data that follow a consistent methodology across countries allowing for cross-country comparisons, which is a rarity in the literature, and (3) it establishes a significant relationship between a few demand side variables and mobile internet adoption, which present significant policy implications. The study finds that demand-side factors are of critical importance for mobile internet adoption. Across the developing world, females, the elderly, those who live in rural areas, and those who have a relatively low level of income or education are less likely to adopt mobile internet. Social network effects are found to have a significant positive impact on the usage of mobile internet. In addition, across regions, although the factors affecting the adoption of mobile internet remain largely the same, the magnitude of their impacts vary. The paper is organized as follows: section 2 outlines the study’s conceptual framework. Section 3 describes the data used. Section 4 lays out the study’s empirical strategy. Section 5 presents the results of several modeling exercises and explores alternate scenarios. The last section concludes and discusses policy implications. 2. Conceptual Framework 3 Information is critical to social and economic activities that constitute the development process. Efficiency, effectiveness, and equity all matter in leveraging information to promote development. Access and usage are among the first prerequisites to exploiting the potential of ICTs. Unequal levels of access and usage create a digital divide between individuals, households, businesses, and geographic areas at different socioeconomic levels. The world’s significant gap in ICT access and usage, despite broad awareness of their associated benefits, has driven researchers to examine determinants of the digital divide. As shown by Srinuan and Bohlin (2011) and Helbig, Gil-Garcia, and Ferro (2005), the determinants can be grouped into three categories: (1) physical infrastructure on the supply side; (2) socioeconomic factors on the demand side, such as income and education level; and (3) perspective factors on the demand side shaped by the institutional environment, culture, language, and network effects. This is also consistent with the framework proposed by van Dijk (2006), who argues that there is a cumulative and recursive model of digital technology adoption, starting with motivation, followed by physical/material access, subsequently requiring digital skills and complementary services (such as electricity) to achieve usage, leading to strengthened motivation and increased usage over time. A higher level of skills and more customized products could strengthen the initial motivation. Much attention was devoted to the importance of physical infrastructure access following the technological determinism of the early 2000s (Jerome Lim 2002; Lentz and Oden 2001; Moss 2002). However, as the digital divide widened, despite expanding service coverage, many researchers explored other underlying socioeconomic factors affecting demand, such as literacy, education, income, and geography (Bagchi 2005; Gauld, Goldfinch, and Horsburgh 2010; Salajan, Schönwetter, and Cleghorn 2010). At a later stage, other social science disciplines came into play as the research on the digital divide evolved. One strand of the literature argues that psychological attitudes toward ICTs affect their adoption (Srinuan and Bohlin 2011; van Dijk 2006), while a second strand emphasizes the role of social networks in shaping perceptions and driving adoption, such as membership in occupational, religious, or cultural communities (Al‐Jaghoub and Westrup 2009; Andrés et al. 2010). The abovementioned studies focus on assessing the determinants of fixed-internet access and usage, many in developed countries. Nevertheless, the fixed internet and mobile internet are different in a variety of aspects such as the price of access devices, convenience of access, coverage and reliability of signal, connection speed, and so on. This paper will explore whether the three main determining factors behind the adoption of general ICTs apply to mobile internet as well. First, the physical infrastructure needed to access mobile internet at the level of 3G or above differs from the telephone or fixed-line broadband connections needed for fixed internet. Household access to electricity is essential for laptop usage with fixed connections, but people can charge mobile phones at other places outside the home, such as offices, community hotspots, or with neighbors. Second, socioeconomic factors such as income and education level might affect access to and usage of mobile internet differently than they do other types of ICTs. For instance, the cost of a mobile phone device is often less than that of a desktop computer. Many mobile network operators and technology firms (e.g., social media platforms) offer a variety of data promotion packages to attract potential customers, and mobile systems can be more intuitive and easier to learn. Third, in terms of perspective factors, with the recent booming of social media, it is argued that social network effects such as the desire to follow friends or family on social media mobile apps are of particular relevance to the access and usage of mobile internet. Within a framework commonly used in the literature to assess the determinants of general ICT access and usage, this paper examines three sets of determining factors behind mobile internet adoption: (1) physical infrastructure (coverage by a 3G or above mobile network, and electricity access); (2) socioeconomic 4 factors such as gender, income level, location (urban/rural), education level, price of mobile data packages, price of internet-enabled phones, and so on; and (3) perspective factors shaped by the institutional environment, culture, and network effects. 3. Data In 2017–18, ICT policy think tanks in the Global South, including RIA (Research ICT Africa), LIRNEasia (Learning Initiative for Network Economies in Asia), and DIRSI (el Diálogo Regional sobre la Sociedad de la Información / Regional Dialogue on the Information Society), coordinated a global initiative, conducting household surveys to collect information about ICT access and usage at household and individual levels. Ten African countries (Ghana, Kenya, Lesotho, Mozambique, Nigeria, Rwanda, Senegal, South Africa, Tanzania, and Uganda), six Asian 8 countries (India, Sri Lanka, Pakistan, Bangladesh, Nepal, and Cambodia), and six Latin American countries (Argentina, Colombia, Ecuador, Guatemala, Paraguay, and Peru) were covered (see Annex 1 for more survey methodology details). Going beyond other microlevel data sets such as the Living Standards Measurement Study and the Global Findex database, this data set collects rich, in-depth information on ICT access and usage from the demand side. It provides valuable information on mobile phone ownership and usage, internet access and usage, social media activities, digital finance, participation in the gig/sharing economy, as well as the reasons for engaging in those ICT-enabled activities. The surveys are nationally representative, and the data can be disaggregated on the basis of gender, age, location (urban or rural), and income level. Unfortunately, there is no direct question on mobile internet adoption in the surveys. However, such usage can be captured through combining information gained through multiple survey questions regarding whether respondents: (1) have used the internet or not, (2) own a mobile phone, and (3) whether the type of mobile phone they own allows internet access. An individual is counted as using mobile internet if he or she has used the internet and also owns an internet-enabled phone.9 Results show that Latin American countries have the highest rate of mobile internet adoption: 65.6 percent of the total population. The rural and urban divide is significant across all regions (figure 1), especially in Africa, where the adoption rate among urban residents is more than twice that among rural residents. Asian countries in the sample have the widest gender gap: only 8.4 percent of females used mobile internet in 2017. Those who are at the bottom 40 percent of the national income distribution have a lower adoption rate across regions, while the gap is minimal in Latin America. Figure 1. Mobile internet adoption, by subgroup, across regions (% of population) a. Rural vs. urban and male vs. female b. Poor vs. not and educated vs. not 8 The majority of the Asian countries covered in the sample are in South Asia. 9 Internet-enabled phones include smartphones and feature phones. Admittedly, this constructed variable cannot completely rule out cases of individuals owning an internet-enabled phone who in fact accessed the internet through Wi-Fi–enabled fixed broadband. However, data from Telegeography show that household fixed-broadband penetration is low (at 13.5 percent) across countries in the sample; thus, such cases should be minimal. The estimated rate is as low as 3.6 percent among African countries, and 7.7 percent among the Asian countries covered. 5 100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% Asia Latin America Sub-Saharan Average Asia Latin America Sub-Saharan Average Africa Africa Rural Urban Male Female Poor Non-poor Educated Non-educated Note: Individuals who have completed at least primary education are grouped as “educated”; individuals who are in the bottom 40 percent of the national income distribution are classified as “poor.” Individual weights are applied in the calculation. The demographic profiles of mobile internet users and nonusers are notably different (figure 2). Across the countries covered in the sample, 54 percent of the nonusers are female, and 68 percent of them are living in rural areas, while the majority of mobile internet users are male and urban residents. The average age of mobile internet users is 29.6 while that of the nonusers is 37.1, indicating that the elderly lag behind in adopting new technologies. Mobile internet users tend to be more educated and wealthier than nonusers. The average number of years of formal education (12.0 years) of mobile internet users is almost twice that of the nonusers (6.1 years). Furthermore, 92 percent of mobile internet users have at least completed primary school, while more than half of the nonusers have either no formal education or have not completed their primary education. The monthly income of mobile internet users is $303.4, more than double that of nonusers at $123.6. Figure 2. Demographic profiles of mobile internet users and nonusers 100% 40.0 350.0 90% 35.0 39% 300.0 80% 46% 30.0 rural 250.0 70% 65% Male 68% 25.0 Male rural 200.0 60% usd 20.0 50% 150.0 15.0 40% 10.0 100.0 61% 30% 54% 50.0 urban 5.0 20% 35% female 32% 0.0 0.0 female urban 10% mobile internet user non-mobile internet user 0% years of formal educaiton age mobile non-mobile mobile non-mobile internet user internet user internet user internet user monthly income (usd) Note: Individual weights are applied in the calculation. 6 The analysis examines decision factors from two perspectives: owning an internet-enabled phone and using internet through the internet-enabled phone. Having an access device, or internet-enabled phone, is the first prerequisite to using mobile internet. More than half of the total population own a mobile phone in almost all the countries covered except for Mozambique, Rwanda, and Uganda (figure 3). There remains a significant divide between men (77.8 percent) versus women (50.4 percent), and urban (73.6 percent) versus rural residents (58.1 percent). Figure 3. Mobile phone ownership by type across selected countries (% of population) Africa Asia Latin America 100 Mobile phone type ownership (% of total 90 80 70 60 50 40 population) 30 20 10 0 No mobile phone Basic phone Feature phone Smartphone Note: Individual weights are applied in the calculation. With regard to obstacles to owning a mobile phone, different reasons are cited such as affordability (39.3 percent), no mobile coverage (16.3 percent), and no electricity at home (17.7 percent). For those who do not have a mobile phone, affordability is a bottleneck across countries: 68.5 percent of those without a mobile phone in Cambodia cite affordability as such. The obstacle of electricity access is particularly severe in Africa (figure 4), especially in Rwanda (50.9 percent) and Senegal (40.4 percent). A lack of mobile coverage is challenging in Cambodia (32.3 percent), Peru (31.2 percent), and Senegal (31.5 percent). 7 Figure 4. Reasons cited for not having a mobile phone (respondents can select multiple options) 80 Africa Percentage points (% of adults not Asia Latin America 70 having a mobile phone) 60 50 40 30 20 10 0 No mobile coverage No electricity at home Can't afford it Note: Individual weights are applied in the calculation. Around 70 percent of those who own a mobile phone only have a basic or feature phone. 10 Overall, smartphone penetration is higher in Latin American and Asian countries than in African countries. For instance, among those who own a mobile phone, the percentage with a smartphone is above half in all Latin American countries in the sample, and close to half in Asian countries (48.5 percent in Cambodia, 51.8 percent in Nepal, and 47 percent in Sri Lanka). That a smartphone is not necessary and not affordable are the two most cited reasons for not having one (figure 5). The affordability obstacle is especially severe in Mozambique, Rwanda, and Tanzania. Figure 5. Main reason cited for not choosing a smartphone among owners of a basic mobile phone (respondents can choose only one main reason) Africa Asia Latin America 100 Percentage points (% of those who 90 own a basic phone or a feature 80 70 60 50 40 phone) 30 20 10 0 No necessity Can't afford Too complicated Other Note: Individual weights are applied in the calculation. 10 Annex 2 contains detailed country-level information on smartphone ownership and smartphone uptake rates. 8 Besides owning an internet-enabled phone, individuals face challenges in accessing and using the internet. For those who do not use the internet at all, the most cited reasons relate to digital literacy, including “do not know what internet is” (58.9 percent) and “do not know how to use the internet” (10.1 percent) (figure 6). No access to a device (computer or mobile phone) is another internet usage constraint (11.5 percent). Other factors such as lack of content in the local language or data privacy concerns do not appear to be key obstacles. Among those who do actually use the internet, the extent of their usage is constrained primarily by lack of time, high data costs, and internet speed. Figure 6. Main reason for not using the internet across the population as a whole (in percentage points) Africa Asia Latin America Percentage points (% of those who do not 100.00 90.00 80.00 70.00 60.00 use Internet) 50.00 40.00 30.00 20.00 10.00 0.00 Digital literacy Affordability Relevance Other Note: The digital literacy category includes “do not know what internet is” and “do not know how to use internet”; affordability includes “no access device” and “too expensive”; the relevance category includes “no interest/not useful” and “no relevant content in local language.” Individual weights are applied in the calculation. People also engage in various mobile internet activities. Survey results show that social networking apps are the most often used app type among people who have an internet-enabled phone. More than a quarter of internet-enabled phone owners use social networking apps daily. Average monthly expenditure on mobile phone data, as a percentage of total income, is 0.9 percent in all the countries covered. In a few countries, such as Sri Lanka, South Africa, and Ghana, monthly mobile data expenditure is more than 2 percent of monthly income—an affordability threshold set by the United Nations. 11 Meanwhile, male, urban residents and young people tend to spend more, which may be associated with their income and level of digital awareness. Those who use the internet adopt different methods to save data charges, such as accessing the internet in a free Wi-Fi area (37.2 percent) or at home or at work (41.9 percent), and taking advantage of special data promotions (49.6 percent). Using special data promotions is the most popular method to save on data charges across countries. This may be related to not only the various promotion packages offered by mobile network operators, but also zero-rating plans offered by content providers such as Facebook. 11 The United Nation’s “1 for 2” threshold of internet affordability is defined as 1 gigabyte (GB) for no more than 2 percent of average monthly income. https://a4ai.org/affordable-internet-is-1-for-2. 9 4. Empirical Strategy To further unravel the factors affecting mobile internet adoption, following the approach of Forenbacher et al. (2019) and Hasbi and Dubus (2020), this analysis tries to estimate the probability of using mobile internet through the following model: Mobile Interneti = α+βk xik+…+ βk xik (1) k = 0,1,…,m; i = 0,1,…,n where Mobile Interneti is a binary variable that equals 1 if individual i uses mobile internet, and 0 if not. Given that the dependent variable is a binary variable, a logit model is adopted for the estimation to predict the probability of an individual adopting mobile internet. By logarithmically transforming the outcome variable, it allows the examination of a nonlinear association in a linear way. The coefficient βk measures, ceteris paribus, the effect of a one-unit change in xik on the dependent variable. Following the conceptual framework, which categorizes factors affecting the digital divide into three groups, variables in the model include: (1) infrastructure factors (3G network coverage and household access to electricity); (2) socioeconomic factors such as gender, income level, education level, and rural/urban location; and (3) social network effects (table 1). Those factors also reflect what the descriptive statistics show to be the main obstacles to owning an internet-enabled phone or using internet, as outlined in the data section above. Infrastructure factors include whether an individual’s location is covered by a 3G mobile network signal, and whether the individual’s household has access to electricity. Whether an address is covered by a 3G mobile signal is identified by mapping the household survey data (which feature GPS location data at the individual level) with the global mobile network coverage data from the Collins Bartholomew data set. 12 Socioeconomic factors include gender, age, education level, income level, marital status, and urban/rural location. Demographic information at the individual level is taken directly from the household survey. Lastly, the number of close friends using an online social network like Facebook or Twitter is used as a proxy for social network effects. Though three waves of data were collected in years 2008, 2012, and 2017, the survey samples were different in each wave, so no panel is available. Given the rapid pace of change in the ICT sector, analysis is confined to the most recently available data for the year 2017, based on cross-sectional estimation. Table 1. Data description Variable Definition, Year Source Dependent variable Accessing the internet through a smartphone Mobile internet After Access data, 2017 or feature phone Infrastructure variable The individual’s household is covered by a After Access data, 2017; Collins 3G network coverage 3G mobile network signal Bartholomew data, 2017 12 Not all countries have geolocation data at the individual/household level in the ICT household survey. For countries that do not have geolocation data at the individual/household level, a household is considered as not having a 3G mobile network coverage if anyone from the same survey enumerator area identified a lack of mobile coverage as an obstacle to accessing the internet. 10 The individual’s household has access to Electricity access After Access data, 2017 electricity Socioeconomic variables Female Respondent is female, 2017 Respondent’s age group, 2017 (1 stands for young people, 15-30 years old; 2 stands for Age group (1-3) middle-age people, 30-60 years old; 3 stands for elderly people, >60 years old) Respondent’s education level, 2017 (1 stands for no formal education; 2 stands for less than 6 years of formal education; 3 stands for more After Access data, 2017 Education (1–5) than 6 but less than 12 years of formal education; 4 stands for more than 12 but less than 16 years of formal education; 5 stands for more than 16 years of formal education) Log value of monthly Log value of the individual’s monthly income income in USD Married Individual is married Urban Individual resides in an urban area Social network effect The number (up to 5) of the respondents’ closest friends who use an online social Close friend (0–5) network like Facebook or Twitter (0 refers to After Access Data, 2017 no close friend and 5 refers to all five closest friends) Table 2. Data summary Mean SD Min Max Mobile internet 0.338 0.473 0 1 Household 3G network coverage 0.909 0.288 0 1 Household having access to electricity 0.843 0.364 0 1 Female 0.563 0.496 0 1 Age group (1-3) 1.695 0.632 1 3 Log value of monthly income 4.709 1.411 -6.274 11.353 Married 0.558 0.497 0 1 Urban residence 0.541 0.510 0 2 Education level (1-5) 2.740 1.035 1 5 Number of closest friends using online social 2.072 2.211 0 5 networks (e.g. Facebook/Twitter (0–5) 5. Results The results section first presents the results from the basic model (1) specification applied to the entire sample and conducts the checks for robustness. The impact of several alternative modeling strategies is then explored, including disaggregation of the sample by geographic region, and the impact of alternative ways of modeling the influence of income on uptake decisions. 5.1 Factors affecting mobile internet adoption in the Global South Estimations show that factors affecting the adoption of mobile internet are largely consistent with those that affect other types of digital divides (table 3). Among infrastructure factors, having access to electricity at 11 the household level is found to significantly improve the likelihood of mobile internet adoption. Socioeconomic factors such as gender, age, income, and education level all have a significant impact on the usage of mobile internet. Females, the elderly, those who live in rural areas, and those who have lower levels of income or education are less likely to adopt mobile internet. Social network effects have a significant positive impact on the usage of mobile internet.13 Individuals whose five closest friends are using an online social network (e.g., Facebook, Twitter) are 63.1 percent more likely to use the mobile internet than those who do not have any close friends using such online social network sites/apps. Table 3. Logit regression results on mobile internet adoption Model Logit Mobile internet adoption Marginal Coefficient effects Infrastructure factors Household 3G network coverage 0.188** 0.039** (0.085) (0.017) Household having access to 0.759*** 0.143*** electricity (0.110) (0.017) Social-economic factors Female -0.142*** -0.031*** (0.044) (0.010) Age group (middle age, 30-60) -0.695*** -0.150*** (0.048) (0.010) Age group (old people, >60) -1.985*** -0.282*** (0.109) (0.009) Log value of monthly income 0.195*** 0.042*** (0.018) (0.004) Married -0.138*** -0.030*** (0.049) (0.011) Urban 0.378*** 0.081*** (0.046) (0.010) Years of formal education (less 0.676*** 0.155*** than 6) (0.124) (0.029) Years of formal education (6-12) 1.603*** 0.347*** (0.118) (0.024) Years of formal education (12-16) 2.388*** 0.535*** (0.125) (0.022) Years of formal education (>16) 2.551*** 0.552*** (0.139) (0.020) Network effects 1 close friend uses an online social 1.461*** 0.349*** network 13 It is also worth noting that the adjusted R2 decreases from 0.448 to 0.346 if the variable “how many close friends are using an online social network” is removed from the model. 12 (0.087) (0.020) 2 close friends use an online social 1.335*** 0.319*** network (0.076) (0.018) 3 close friends use an online social 1.791*** 0.420*** network (0.076) (0.016) 4 close friends use an online social 2.115*** 0.480*** network (0.096) (0.017) 5 close friends use an online social 3.049*** 0.631*** network (0.064) (0.010) _cons -4.287*** (0.225) Country fixed effects Yes Yes Number of observations 19,979 19,979 R2 Adjusted R2 0.448 0.448 Note: *** p<0.01, ** p<0.05, * p<0.1. Marginal effects at mean are presented. To evaluate the validity of model (1), a Hosmer–Lemeshow goodness-of-fit test is conducted after the estimation. Given the large number of observations, the test is conducted with 100 groups. A p-value of 0.7387 of the Pearson chi-square from the Hosmer and Lemeshow’s goodness-of-fit test indicates that the model fits the data well. Furthermore, to detect if any potential observations have a significant impact on the model, model (1) is retested by excluding observations with Pearson residual value more than 2, deviance residual value more than 2, and leverage value more than 3 times the average leverage. The sign and significance of the coefficient of different factors affecting mobile internet adoption are still retained after excluding those observations. 5.2 Differences in factors affecting mobile internet adoption across regions Results are slightly different if model (1) is estimated for each region separately (table 4). Among the infrastructure factors, 3G network coverage’s impact is significant in Asia, and the impact of household access to electricity is particularly high in Africa and Latin America. In terms of socioeconomic factors, gender significantly affects the adoption of mobile internet in Asian countries, showing potential gender inequality in digital access and usage in the region. In Asia, married people tend not to use mobile internet as much as others. Residence in a rural or urban location has a significant impact in African and Latin American countries. Income level’s impact is significant in Africa and Asia. The effects of education level are particularly salient in Africa, implying the region’s needs to improve digital literacy. 13 Table 4. Logit regression results for mobile internet adoption by region Model Logit Marginal effects Latin All region Africa Asia America Infrastructure factors Household 3G network coverage 0.039** 0.012 0.053 0.037** (0.017) (0.021) (0.046) (0.018) Household having access to 0.143*** 0.106*** 0.229 0.070** electricity (0.017) (0.016) (0.149) (0.028) Social-economic factors Female -0.031*** -0.020* -0.002 -0.042*** (0.010) (0.012) (0.016) (0.012) Age group (middle age, 30-60) -0.150*** -0.112*** -0.174*** -0.099*** (0.010) (0.014) (0.018) (0.014) Age group (old people, >60) -0.282*** -0.153*** -0.538*** -0.148*** (0.009) (0.016) (0.025) (0.013) Log value of monthly income 0.042*** 0.037*** 0.015*** 0.048*** (0.004) (0.005) (0.006) (0.006) Married -0.030*** 0.000 -0.034* -0.062*** (0.011) (0.013) (0.018) (0.017) Urban 0.081*** 0.069*** 0.084*** 0.049*** (0.010) (0.013) (0.019) (0.011) Years of formal education (less 0.155*** 0.093** 0.119** 0.142*** than 6) (0.029) (0.037) (0.057) (0.039) Years of formal education (6-12) 0.347*** 0.269*** 0.321*** 0.267*** (0.024) (0.029) (0.060) (0.034) Years of formal education (12-16) 0.535*** 0.491*** 0.344*** 0.511*** (0.022) (0.035) (0.034) (0.047) Years of formal education (>16) 0.552*** 0.583*** 0.296*** 0.466*** (0.020) (0.035) (0.019) (0.052) Network effects 1 close friend uses an online social 0.349*** 0.391*** 0.156*** 0.257*** network (0.020) (0.030) (0.028) (0.034) 2 close friends use an online social 0.319*** 0.298*** 0.167*** 0.272*** network (0.018) (0.027) (0.024) (0.029) 3 close friends use an online social 0.420*** 0.427*** 0.202*** 0.384*** network (0.016) (0.026) (0.020) (0.030) 4 close friends use an online social 0.480*** 0.492*** 0.227*** 0.480*** network (0.017) (0.032) (0.018) (0.036) 5 close friends use an online social 0.631*** 0.666*** 0.555*** 0.553*** network 14 (0.010) (0.015) (0.027) (0.020) Number of observations 19,979 8,068 5,469 6,442 R2 Adjusted R2 0.448 0.454 0.366 0.394 Note: *** p<0.01, ** p<0.05, * p<0.1. Marginal effects at mean are presented. 5.3 Alternative way of modeling the influence of income Model (1) includes the income variable, which is an important explanatory factor for ICT adoption in general according to the literature. Individuals’ income level could affect their decisions to purchase an Internet-enabled phone, and afford mobile data packages to access internet, which jointly have an impact on the decision of mobile internet adoption. To further unpack the impact of income on affordability, the analysis tries an alternative specification with inclusion of a few affordability dummies while other co- variates remain the same. To understand how the prices of mobile data packages affect the adoption of mobile internet, the analysis considers a country’s average cost per 1 gigabyte (GB) of mobile data, and constructs a dummy variable that equals 1 if the country’s average cost per 1 GB of mobile data is more than 2 percent of the individual’s monthly income. Moreover, since the affordability of an access device is often cited as an obstacle to internet usage, the analysis uses data about a country’s average cost of an internet-enabled phone from International Data Corporation (IDC) and constructs a dummy variable that equals 1 if the average phone cost is more than the individual’s monthly income. Due to collinearity concerns, the inclusion of these affordability measures precludes the incorporation of the income variable directly. Results confirm the important role of affordability issues in driving mobile internet adoption. The high cost of mobile data packages and expensive mobile phones negatively affect the adoption of mobile internet. If the country’s average cost per 1 GB of mobile data is more than 2 percent of an individual’s monthly income, that individual is 5 percent less likely to adopt mobile internet. If the country’s average cost of an internet- enabled phone is more than an individual’s monthly income, that individual is 4 percent less likely to adopt mobile internet. Meanwhile, the sign and level of significance of other variables in model (1) remain largely similar (see tables 5 and 3). For instance, under both the original and new model specifications, if the individual’s residence has access to electricity, he or she is 14 percent more likely to use mobile internet; and urban residents are 8 percent more likely to use mobile internet. With regard to education level, the impacts of completing primary and secondary education remain largely the same, while the impact of completing tertiary education is at a higher level in the new model specification. Individuals who have completed tertiary education are 61.2 percent more likely than those with no formal education to use mobile internet, while the impact level is 55.2 percent in model (1). Finally, the overall explanatory power of the regression is slightly lower when income is used (0.45) than when the two affordability dummies are used (0.47). Table 5. Logit regression results with different treatment of the income variable Model Logit Mobile internet usage Marginal Coefficient effects Infrastructure factors Household 3G network coverage 0.180* 0.037* (0.108) (0.022) Household having access to 0.770*** 0.142*** electricity 15 (0.121) (0.018) Social-economic factors Female -0.235*** -0.050*** (0.049) (0.010) Age group (middle age, 30-60) -0.709*** -0.150*** (0.054) (0.011) Age group (old people, >60) -1.914*** -0.268*** (0.118) (0.010) Married -0.130** -0.028** (0.055) (0.012) Urban 0.362*** 0.076*** (0.052) (0.011) Years of formal education (less 0.801*** 0.185*** than 6) (0.164) (0.039) Years of formal education (6-12) 1.743*** 0.365*** (0.154) (0.029) Years of formal education (12-16) 2.541*** 0.561*** (0.161) (0.028) Years of formal education (>16) 3.016*** 0.612*** (0.182) (0.021) High cost of internet-enabled -0.189*** -0.040*** phone (0.071) (0.015) High cost of mobile data package -0.226*** -0.048*** (0.074) (0.016) Network effects 1 close friend uses an online social 1.577*** 0.374*** network (0.097) (0.021) 2 close friends use an online social 1.416*** 0.337*** network (0.086) (0.020) 3 close friends use an online social 1.797*** 0.421*** network (0.086) (0.018) 4 close friends use an online social 2.205*** 0.497*** network (0.110) (0.019) 5 close friends use an online social 3.073*** 0.631*** network (0.072) (0.011) _cons -3.119*** (0.250) Number of observations 16,591 16,591 R2 Adjusted R2 0.466 0.466 16 Note: *** p<0.01, ** p<0.05, * p<0.1. Marginal effects at mean are presented. 6. Conclusion As digital technologies penetrate various aspects of social and economic life, having access to affordable and reliable internet becomes essential for individuals to stay connected with their social networks, efficiently engage in online economic activities, and better receive public services. The United Nations Sustainable Development Goals set the target (9.c) 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.” A better understanding of key drivers and main constraints for internet access is the first prerequisite for governments to design targeted policy solutions. Results from the study show that besides infrastructure investment, which has been the main focus of many developing countries, other demand-side factors are of critical importance. Across the developing world, females, the elderly, those who live in rural areas, and those who have a relatively low level of income or education are less likely to adopt mobile internet. Therefore, policy measures targeted at reducing gender inequality and the urban/rural divide could support wider adoption of mobile internet. Enhancing education levels to increase people’s awareness of the benefits of being digitally connected could also positively affect mobile internet adoption. Among infrastructure factors, having access to electricity at the household level is significantly associated with mobile internet adoption, particularly owning an internet-enabled phone. 3G network coverage remains to be a significant factor, especially in Asia. Moreover, social network effects are found to have a significant positive impact on the usage of mobile internet. Those who have more close friends using an online social network are more likely to adopt mobile internet. Individuals whose five closest friends are using an online social network (e.g., Facebook, Twitter) are 63.1 percent more likely to adopt it than those without any close friends using such online social network sites/apps. Collaborating with different firms across social media platforms to leverage these positive social network effects is one route to reduce the digital divide in mobile internet connections. Lastly, income is a strong driver of mobile internet adoption in its own right. Moreover, the high cost of smartphones and mobile data packages negatively affects the adoption of mobile internet. It is estimated that if a country’s average cost for 1 GB of mobile data is more than 2 percent of an individual’s monthly income, that individual is 5 percent less likely to adopt mobile internet. If the country’s average cost of an internet-enabled phone is more than the individual’s monthly income, that individual is 4 percent less likely to adopt mobile internet. This implies the necessity to create a competitive market with multiple players offering data packages at affordable prices, and to take measures to reduce the retail cost of smartphones. Across regions, although the factors affecting the adoption of mobile internet remain largely the same, the magnitude of their impacts vary. In Asia, gender differences are negatively associated with mobile internet in the region. In Africa, the impact of education level is more salient than the other two regions, implying an urgent need to improve digital literacy. 17 References Aker, J. C., and I. M. Mbiti. 2010. “Mobile Phones and Economic Development in Africa.” Journal of Economic Perspectives 24 (3): 207–32. https://doi.org/10.1257/jep.24.3.207. Al‐Jaghoub, S., and C. Westrup. 2009. “Reassessing Social Inclusion and Digital Divides.” Journal of Information, Communication and Ethics in Society 7 (2/3): 146–58. https://doi.org/10.1108/14779960910955864. Andrés, L., D. Cuberes, M. Diouf, and T. Serebrisky. 2010. “The Diffusion of the Internet: A Cross- Country Analysis.” Telecommunications Policy 34 (5–6): 323–40. https://doi.org/10.1016/j.telpol.2010.01.003. Armey, L.E. and L. Hosman. 2016. “The centrality of electricity to ICT use in low-income countries,” Telecommunications Policy, 40(7): 617-627. Bagchi, K. 2005. “Factors Contributing to Global Digital Divide: Some Empirical Results.” Journal of Global Information Technology Management 8 (3): 47–65. https://doi.org/10.1080/1097198X.2005.10856402. Bahia, K., P. Castells, G. Cruz, T. Masaki, X. Pedros, T. Pfutze, C. Rodriguez Castelan, and H.J. Winkler. 2020. “The Welfare Effects of Mobile Broadband Internet: Evidence from Nigeria.” Policy Research Working Paper 9230. Washington, D.C.: World Bank Group. Bailur, S., and S. Masiero. 2017. “Women’s Income Generation through Mobile Internet: A Study of Focus Group Data from Ghana, Kenya, and Uganda.” Gender, Technology and Development 21 (1–2): 77–98. https://doi.org/10.1080/09718524.2017.1385312. Birba, O., and A. Diagne. 2012. “Determinants of Adoption of Internet in Africa: Case of 17 Sub-Saharan Countries.” Structural Change and Economic Dynamics 23 (4): 463–42. https://doi.org/10.1016/j.strueco.2012.06.003. Björkegren, D. 2019. “The Adoption of Network Goods: Evidence from the Spread of Mobile Phones in Rwanda.” The Review of Economic Studies 86 (3): 1033–60. https://doi.org/10.1093/restud/rdy024. Cecchini, S., and C. Scott. 2003. “Can Information and Communications Technology Applications Contribute to Poverty Reduction? Lessons from Rural India.” Information Technology for Development 10 (2): 73–84. https://doi.org/10.1002/itdj.1590100203. Cerno, L., and T. P. Amaral. 2006. “Demand for Internet Access and Use in Spain.” In Governance of Communication Networks: Connecting Societies and Markets with IT, Contributions to Economics, edited by Brigitte Preissl and Jürgen Müller, 333–53. Heidelberg: Physica-Verlag HD. https://doi.org/10.1007/3-7908-1746-5_18. Forenbacher, I., S. Husnjak, I. Cvitić, and I. Jovović. 2019. “Determinants of Mobile Phone Ownership in Nigeria.” Telecommunications Policy 43 (7): 101812. https://doi.org/10.1016/j.telpol.2019.03.001. Funk, J. L. 2005. “The Future of the Mobile Phone Internet: An Analysis of Technological Trajectories and Lead Users in the Japanese Market.” Technology in Society 27 (1): 69–83. https://doi.org/10.1016/j.techsoc.2004.10.001. Gauld, R., S. Goldfinch, and S. Horsburgh. 2010. “Do They Want It? Do They Use It? The ‘Demand- Side’ of E-Government in Australia and New Zealand.” Government Information Quarterly 27 (2): 177–86. https://doi.org/10.1016/j.giq.2009.12.002. 18 Gillwald, A., F. Odufuwa and O. Mothobi. 2018. “The state of ICT in Nigeria. Policy paper series (5): After access state of ICT in Nigeria.” Retrieved from: http://extensia-ltd.com/wp- content/uploads/2018/11/After-Access-Nigeria-State-of-ICT-2017.pdf Goldfarb A. and J. Prince. 2008. "Internet adoption and usage patterns are different: Implications for the digital divide,” Information Economics and Policy 20: 2–15. GSMA (GSM Association). 2019a. The State of Mobile Internet Connectivity Report 2019: Mobile for Development. London: GSMA. GSMA. 2019b. The Mobile Gender Gap Report 2019: Mobile for Development. London: GSMA. GSMA and Gallup. 2018. The Impact of Mobile on People’s Happiness and Well-Being. London: GSMA. Hasbi, M., and A. Dubus. 2020. “Determinants of Mobile Broadband Use in Developing Economies: Evidence from Sub-Saharan Africa.” Telecommunications Policy 44 (5): 101944. https://doi.org/10.1016/j.telpol.2020.101944. Helbig, N., J. R. Gil-Garcia, and E. Ferro. 2005. “Understanding the Complexity of Electronic Government: Implications from the Digital Divide Literature.” Government Information Quarterly 26 (1): 89–97. https://doi.org/10.1016/j.giq.2008.05.004. Hsu, C.-L., H.-P. Lu, and H.-H. Hsu. 2007. “Adoption of the Mobile Internet: An Empirical Study of Multimedia Message Service (MMS).” Omega 35 (6): 715–26. Jerome Lim, J. 2002. “East Asia in the Information Economy: Opportunities and Challenges.” info 4 (5): 56–63. https://doi.org/10.1108/14636690210453226. Katz, J. E., and R. E. Rice. 2003. “Social Consequences of Internet Use: Access, Involvement and Interaction.” info 5 (4): 46–46. https://doi.org/10.1108/14636690310495274. Katz, R., and F. Callorda. 2018. The Economic Contribution of Broadband, Digitization and ICT Regulation. Geneva: International Telecommunication Union. Kim, H.-Y., and J.-W. Kim. 2002. “An Empirical Research on Important Factors of Mobile Internet Usage.” Asia Pacific Journal of Information Systems 12 (3): 89–113. Lentz, R. G., and M. D. Oden. 2001. “Digital Divide or Digital Opportunity in the Mississippi Delta Region of the US.” Telecommunications Policy 25 (5): 291–313. Moss, J. 2002. “Power and the Digital Divide.” Ethics and Information Technology 4 (2): 159–65. Penard, T., N. Poussing, B. Mukoko, G. Bertrand and T. Piaptie. 2015. “Internet adoption and usage patterns in Africa: Evidence from Cameroon,” Technology in Society 42: 71-80. Penard, T., N. Poussing, G. Zomo Yebe and P. Nsi Ella. 2012. “Comparing the Determinants of Internet and Cell Phone Use in Africa: Evidence from Gabon,” Communications & Strategies, 86, 2nd Quarter, 65-83. Rodriguez-Castelan, C., A. Araar, E. Malasquez Carbonel, S. Daniel Olivieri, and T. Vishwanath, 2019. “Distributional Effects of Competition: A Simulation Approach.” Policy Research Working Paper 9230. Washington, D.C.: World Bank Group. Roller, L.-H., and L. Waverman. 2001. “Telecommunications Infrastructure and Economic Development: A Simultaneous Approach.” The American Economic Review 91 (4): 909–23. Salajan, F. D., D. J. Schönwetter, and B. M. Cleghorn. 2010. “Student and Faculty Inter-Generational Digital Divide: Fact or Fiction?” Computers & Education 55 (3): 1393–403. Srinuan, C., and E. Bohlin. 2011. “Understanding the Digital Divide: A Literature Survey and Ways Forward (No. 52191).” Paper presented at 22nd European Regional Conference of the 19 International Telecommunications Society (ITS), “Innovative ICT Applications—Emerging Regulatory, Economic and Policy Issues,” Budapest, Hungary, September 18–21, 2011. Srinuan, C., P. Srinuan, and E. Bohlin. 2012. “An Analysis of Mobile Internet Access in Thailand: Implications for Bridging the Digital Divide.” Telematics and Informatics 29 (3): 254–62. https://doi.org/10.1016/j.tele.2011.10.003. van Biljon, J., and P. Kotzé. 2007. “Modelling the Factors that Influence Mobile Phone Adoption.” In Proceedings of the 2007 Annual Research Conference of the South African Institute of Computer Scientists and Information Technologists on IT Research in Developing Countries, SAICSIT ’07, 152–61. New York: Association for Computing Machinery. https://doi.org/10.1145/1292491.1292509. Van de Ven, Wynand P. M. M. and van Praag, Bernard. 1981. “The demand for deductibles in private health insurance: A probit model with sample selection.” Journal of Econometrics, 17, issue 2, p. 229-252. van Dijk, J. A. G. M. 2006. “Digital Divide Research, Achievements and Shortcomings.” Poetics 34 (4– 5): 221–35. https://doi.org/10.1016/j.poetic.2006.05.004. Waverman, L., M. Meschi, and M. Fuss. 2005. “The Impact of Telecoms on Economic Growth in Developing Countries.” The Vodafone Policy Paper Series 2: 10–24. World Bank. 2018. Information and Communication for Development 2018: Data-Driven Development (English). Washington, DC: World Bank (accessed April 26, 2020). http://documents.worldbank.org/curated/en/987471542742554246/Information-and- Communication-for-Development-2018-Data-Driven-Development. 20 Annexes Annex 1. Survey methodology The three think tanks, including RIA (Research ICT Africa), LIRNEasia (Learning Initiative for Network Economies in Asia), and DIRSI (el Diálogo Regional sobre la Sociedad de la Información / Regional Dialogue on the Information Society), conducted the survey separately in 2017/2018. They adopted the same questionnaire and sampling methodology. The random sampling is based on a Census sample. A Census divides a country into Enumerator Areas (EAs) which roughly have a household density of 200. The desired level of accuracy for the survey was set to a confidence level of 95% and a margin of error of 5%, which yields a minimum sample size per tabulation group of 385. Weights at household level and individual level are calculated based on the inverse selection probabilities and I gross up the data to the national level when applied. Annex 2. Mobile Phone Ownership (%) Years of Years of Years of formal formal formal No formal education education education Country Male Female Rural Urban Elderly Young Non-poor Poor education (less than 6) (6–12) (12–16) Total Argentina 84 87 94 86 81 92 85 86 51 59 85 93 86 Bangladesh 87 58 72 78 71 75 78 64 62 70 80 97 74 Cambodia 78 62 64 81 64 71 76 56 41 63 78 90 68 Colombia 89 90 84 92 90 89 92 87 61 84 90 95 90 Ecuador 84 83 78 86 84 82 82 85 44 73 83 84 83 Ghana 81 67 62 84 72 75 84 58 52 70 80 90 74 Guatemala 90 82 85 87 83 88 89 81 75 80 88 96 86 India 79 43 55 71 59 62 67 47 37 57 67 86 61 Kenya 92 83 85 93 88 86 93 79 58 76 84 98 87 Lesotho 80 78 72 87 74 83 82 74 49 64 82 92 79 Mozambique 50 32 33 55 43 39 57 16 22 33 60 93 40 Nepal 80 65 65 76 64 78 70 82 n.a. 75 85 96 72 Nigeria 70 57 54 78 67 61 74 48 25 56 77 93 63 Pakistan 68 43 56 59 64 53 74 33 39 59 74 87 57 Paraguay 89 85 81 90 84 91 92 80 66 78 91 98 87 Peru 86 80 80 83 81 83 88 76 n.a. 66 80 90 82 Rwanda 60 37 45 61 51 46 63 28 23 47 62 92 48 Senegal 81 74 72 84 76 79 82 71 67 86 82 96 78 21 South Africa 83 85 80 86 85 83 85 82 62 74 84 97 84 Sri Lanka 86 72 77 84 76 82 85 67 32 57 77 94 78 Tanzania 64 53 51 74 63 55 75 33 26 41 62 93 59 Uganda 58 40 44 64 54 46 66 24 23 41 50 89 49 Total 78 50 58 73 64 64 71 50 39 59 71 89 64 Annex 3. Smartphone (%) Years of Years of Years of formal formal formal No formal education education education Country Male Female Rural Urban Elderly Young Non-poor Poor education (less than 6) (6–12) (12–16) Total Argentina_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Argentina_smartphone ownership 62 66 74 64 52 79 60 70 13 23 62 77 64 Argentina_smartphone uptake 62 66 74 64 52 79 60 70 13 23 62 77 64 Bangladesh_3g 97 96 97 96 97 96 96 98 94 98 97 100 97 Bangladesh_smartphone ownership 22 12 16 21 10 23 19 14 4 10 25 47 17 Bangladesh_smartphone uptake 23 12 16 21 10 23 19 14 5 10 25 47 18 Cambodia_3g 80 82 81 82 78 84 81 81 86 80 81 76 81 Cambodia_smartphone ownership 40 28 28 48 19 46 39 24 9 19 48 73 33 Cambodia_smartphone uptake 42 30 29 49 20 47 40 25 9 20 50 75 34 Colombia_3g 99 100 100 100 100 100 100 100 97 99 100 100 100 Colombia_smartphone ownership 56 57 50 60 42 69 64 48 3 25 60 81 57 Colombia_smartphone uptake 56 57 50 60 43 69 64 48 3 25 60 81 57 Ecuador_3g 94 95 87 99 94 96 95 96 100 93 94 96 95 Ecuador_smartphone ownership 65 59 57 64 47 71 61 62 0 31 62 75 61 Ecuador_smartphone uptake 65 59 55 64 47 71 61 62 0 30 61 74 61 Ghana_3g 73 83 54 98 79 78 85 69 70 81 77 91 78 Ghana_smartphone ownership 29 22 16 33 16 32 32 16 4 9 27 60 25 Ghana_smartphone uptake 33 26 19 34 20 36 35 17 5 7 32 60 29 Guatemala_3g 97 98 96 99 98 97 98 96 99 97 98 97 97 22 Guatemala_smartphone ownership 56 39 45 49 28 62 47 47 17 30 55 72 47 Guatemala_smartphone uptake 56 39 45 49 28 61 47 47 17 29 55 72 47 India_3g 93 95 95 91 94 94 94 94 95 96 92 94 94 India_smartphone ownership 25 10 11 28 9 24 20 11 1 8 19 53 17 India_smartphone uptake 25 10 11 28 9 24 20 11 1 8 19 54 17 Kenya_3g 76 69 65 92 68 74 77 64 37 69 68 82 72 Kenya_smartphone ownership 29 20 14 51 15 30 33 11 1 3 12 44 24 Kenya_smartphone uptake 33 23 16 50 17 33 35 15 2 4 16 44 28 Lesotho_3g 74 73 77 63 76 71 72 75 74 73 73 72 73 Lesotho_smartphone ownership n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Lesotho_smartphone uptake n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Mozambique_3g 93 94 94 93 93 94 94 94 93 94 95 86 94 Mozambique_smartphone ownership 9 5 2 17 4 8 10 2 0 1 16 49 7 Mozambique_smartphone uptake 9 4 2 16 4 8 10 2 0 1 15 50 7 Nepal_3g 98 95 96 97 96 97 96 98 95 97 97 97 Nepal_smartphone ownership 44 31 30 41 21 50 37 40 23 53 79 37 Nepal_smartphone uptake 45 31 30 42 22 50 37 41 24 54 80 38 Nigeria_3g 72 73 58 94 72 72 73 71 46 72 83 84 72 Nigeria_smartphone ownership 18 11 9 23 12 16 16 12 1 1 17 36 15 Nigeria_smartphone uptake 22 14 12 23 14 20 20 14 1 1 19 35 18 Pakistan_3g 93 92 92 94 93 92 94 91 93 97 88 91 93 Pakistan_smartphone ownership 14 11 9 19 15 11 17 7 3 10 21 50 13 Pakistan_smartphone uptake 14 10 8 20 15 11 17 7 3 11 21 53 13 Paraguay_3g 88 88 75 96 87 89 90 85 77 84 90 94 88 Paraguay_smartphone ownership 51 53 37 61 36 77 59 44 11 23 69 87 52 Paraguay_smartphone uptake 54 54 37 62 38 78 60 46 15 24 69 88 54 Peru_3g 99 99 97 100 99 100 99 99 98 99 99 99 Peru_smartphone ownership 52 48 30 55 36 61 54 44 12 48 61 49 Peru_smartphone uptake 52 48 29 55 36 61 54 44 11 48 61 49 23 Rwanda_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Rwanda_smartphone ownership 5 3 1 16 3 5 7 1 1 1 4 53 4 Rwanda_smartphone uptake 5 3 1 16 3 5 7 1 1 1 4 53 4 Senegal_3g 80 83 77 87 81 82 85 77 81 81 80 83 82 Senegal_smartphone ownership 29 17 17 30 15 31 27 18 8 18 29 64 23 Senegal_smartphone uptake 31 20 20 32 17 33 29 21 9 21 31 69 26 South Africa_3g 89 89 86 90 89 89 90 87 88 81 89 94 89 South Africa_smartphone ownership 50 44 33 54 38 56 46 47 13 12 48 78 47 South Africa_smartphone uptake 51 46 33 56 39 57 49 47 14 10 50 78 48 Sri Lanka_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Sri Lanka_smartphone ownership 45 30 35 45 23 57 42 28 3 6 33 62 37 Sri Lanka_smartphone uptake 45 31 34 47 22 58 42 30 4 7 32 62 37 Tanzania_3g 98 97 96 100 97 97 98 96 87 96 99 100 97 Tanzania_smartphone ownership 15 11 5 29 8 16 19 4 0 1 10 71 13 Tanzania_smartphone uptake 16 11 5 29 8 17 19 4 0 2 10 71 13 Uganda_3g 90 94 92 95 94 91 92 92 94 92 92 91 92 Uganda_smartphone ownership 9 7 3 23 5 10 12 2 0 1 6 36 8 Uganda_smartphone uptake 9 7 3 24 4 11 12 2 0 1 6 36 8 Total_3g 91 92 91 93 92 91 92 91 89 94 92 93 92 Total_smartphone ownership 25 15 13 32 13 25 23 14 2 8 23 54 20 Total_smartphone uptake 25 15 13 32 13 25 23 15 2 9 23 55 20 Note: Smartphone ownership reflects the ownership rate of the entire population. The uptake rate refers to the ownership rate for the subpopulation covered by 3G. Annex 4. Internet Usage Ratio Internet Usage by Country and Subgroup 24 Years of Years of Years of formal formal formal No formal education education education Country Total Male Female Rural Urban Elderly Young education (less than 6) (6–12) (12–16) Argentina 79 80 78 79 79 66 96 9 29 79 93 Bangladesh 13 18 7 11 19 6 18 2 4 20 47 Cambodia 40 47 34 34 57 21 56 6 21 57 85 Colombia 77 77 78 72 80 58 94 8 38 88 98 Ecuador 80 85 77 76 82 61 93 19 40 84 94 Ghana 26 31 21 15 35 13 36 1 9 29 63 Guatemala 62 70 55 58 66 36 81 17 33 78 92 India 19 26 11 14 27 8 28 1 8 22 56 Kenya 26 31 21 16 53 13 34 1 3 15 46 Lesotho 32 36 31 18 54 13 52 2 3 31 83 Mozambique 23 26 20 9 44 16 26 3 5 33 60 Nepal 45 43 50 34 51 23 60 0 20 51 83 Nigeria 29 37 20 20 41 19 35 0 4 37 69 Pakistan 17 21 12 16 18 10 20 2 24 24 55 Paraguay 57 55 58 40 67 39 84 11 26 77 90 Peru 71 77 68 46 78 50 90 0 17 70 88 Rwanda 39 46 29 28 59 39 39 75 14 43 69 Senegal 30 33 26 21 39 15 43 7 25 41 83 South Africa 71 74 68 61 75 62 77 31 37 69 90 Sri Lanka 37 45 30 35 45 20 62 3 6 31 67 Tanzania 31 32 29 13 55 22 34 2 4 22 81 Uganda 45 44 46 39 54 30 50 0 17 40 64 Annex 5. Mobile Internet Adoption and Uptake Rates by Country and Subgroup No formal Years of Years of Years of Country Male Female Rural Urban Elderly Young Non-poor Poor education formal formal formal Total 25 education education education (less than 6) (6–12) (12–16) Argentina_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Argentina_mobile internet 73 73 78 73 59 92 68 81 9 22 73 88 73 Argentina_mobile internet_uptake 73 73 78 73 59 92 68 81 9 22 73 88 73 Bangladesh_3g 97 96 97 96 97 96 96 98 94 98 97 100 97 Bangladesh_mobile internet 17 7 11 18 6 17 14 9 2 4 19 45 13 Bangladesh_mobile internet_uptake 18 7 11 18 6 18 14 9 2 4 19 45 13 Cambodia_3g 80 82 81 82 78 84 81 81 86 80 81 76 81 Cambodia_mobile internet 39 25 25 46 15 46 36 22 5 16 47 75 31 Cambodia_mobile internet_uptake 41 27 27 47 15 47 37 24 5 17 49 79 32 Colombia_3g 99 100 100 100 100 100 100 100 97 99 100 100 100 Colombia_mobile internet 66 68 61 71 49 86 74 60 8 28 75 92 68 Colombia_mobile internet_uptake 67 68 61 71 49 86 74 60 9 29 75 92 68 Ecuador_3g 94 95 87 99 94 96 95 96 100 93 94 96 95 Ecuador_mobile internet 78 71 69 76 54 89 73 75 0 33 77 91 74 Ecuador_mobile internet_uptake 79 71 69 76 54 89 73 75 0 34 76 91 74 Ghana_3g 73 83 54 98 79 78 85 69 70 81 77 91 78 Ghana_mobile internet 27 19 13 32 12 32 27 17 1 9 25 58 23 Ghana_mobile internet_uptake 32 22 14 32 14 36 30 19 1 8 28 60 27 Guatemala_3g 97 98 96 99 98 97 98 96 99 97 98 97 97 Guatemala_mobile internet 64 47 52 58 30 74 54 56 14 29 70 85 55 Guatemala_mobile internet_uptake 64 47 52 58 31 74 54 56 15 29 70 84 55 India_3g 93 95 95 91 94 94 94 94 95 96 92 94 94 India_mobile internet 22 8 10 24 7 22 18 9 1 5 17 51 15 26 India_mobile internet_uptake 23 8 10 25 7 22 18 9 1 6 16 53 15 Kenya_3g 76 69 65 92 68 74 77 64 37 69 68 82 72 Kenya_mobile internet 30 19 15 50 12 32 32 13 1 3 13 45 24 Kenya_mobile internet_uptake 35 23 18 50 15 37 36 17 2 5 16 48 29 Lesotho_3g 74 73 77 63 76 71 72 75 74 73 73 72 73 Lesotho_mobile internet n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Lesotho_mobile internet_uptake n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. Mozambique_3g 93 94 94 93 93 94 94 94 93 94 95 86 94 Mozambique_mobile internet 20 16 6 36 12 19 22 8 2 3 26 55 18 Mozambique_mobile internet_uptake 20 15 6 35 12 19 22 7 2 3 25 57 17 Nepal_3g 98 95 96 97 96 97 96 98 95 97 97 97 Nepal_mobile internet 39 27 25 37 16 45 32 35 14 48 77 32 Nepal_mobile internet_uptake 40 27 25 38 17 46 33 35 15 49 79 33 Nigeria_3g 72 73 58 94 72 72 73 71 46 72 83 84 72 Nigeria_mobile internet 33 16 15 37 16 29 27 20 0 3 31 62 24 Nigeria_mobile internet_uptake 41 20 22 38 21 36 33 26 0 4 33 65 30 Pakistan_3g 93 92 92 94 93 92 94 91 93 97 88 91 93 Pakistan_mobile internet 15 7 10 14 10 13 16 5 1 13 17 52 12 Pakistan_mobile internet_uptake 17 8 11 15 10 14 17 6 1 14 19 57 12 Paraguay_3g 88 88 75 96 87 89 90 85 77 84 90 94 88 Paraguay_mobile internet 53 56 38 64 37 81 62 45 11 24 73 89 55 Paraguay_mobile internet_uptake 55 58 38 65 39 83 63 47 15 25 74 91 57 Peru_3g 99 99 97 100 99 100 99 99 98 99 99 99 Peru_mobile internet 65 60 35 70 42 81 68 53 11 61 77 62 Peru_mobile internet_uptake 66 60 36 70 42 81 69 54 12 61 78 62 Rwanda_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Rwanda_mobile internet 31 24 16 50 32 26 35 11 18 8 28 65 28 27 Rwanda_mobile internet_uptake 31 24 16 50 32 26 35 11 18 8 28 65 28 Senegal_3g 80 83 77 87 81 82 85 77 81 81 80 83 82 Senegal_mobile internet 30 24 19 37 14 40 29 25 7 22 35 82 27 Senegal_mobile internet_uptake 32 27 21 38 16 42 31 26 7 26 36 85 29 South Africa_3g 89 89 86 90 89 89 90 87 88 81 89 94 89 South Africa_mobile internet 68 60 51 69 56 69 65 62 20 33 60 84 64 South Africa_mobile internet_uptake 69 61 50 70 57 70 66 62 21 26 61 85 65 Sri Lanka_3g 100 100 100 100 100 100 100 100 100 100 100 100 100 Sri Lanka_mobile internet 41 24 29 39 17 55 36 23 2 4 26 59 31 Sri Lanka_mobile internet_uptake 41 24 29 39 15 56 35 24 4 6 24 60 31 Tanzania_3g 98 97 96 100 97 97 98 96 87 96 99 100 97 Tanzania_mobile internet 29 27 12 50 20 31 37 8 2 4 20 77 28 Tanzania_mobile internet_uptake 29 27 12 50 20 31 37 8 2 4 20 77 28 Uganda_3g 90 94 92 95 94 91 92 92 94 92 92 91 92 Uganda_mobile internet 30 38 22 47 24 37 40 15 0 4 25 56 33 Uganda_mobile internet_uptake 30 39 22 48 22 39 41 16 0 5 27 55 34 Total_3g 91 92 91 93 92 91 92 91 89 94 92 93 92 Total_mobile internet 26 14 12 32 11 26 22 15 1 7 23 57 20 Total_mobile internet_uptake 26 14 13 32 12 27 23 15 1 7 23 59 20 Note: Mobile internet reflects the adoption rate of the entire population. The uptake rate refers to the ownership rate for the subpopulation covered by 3G. 28