ABSTRACT

Although mobile phone subscription rates in Nigeria have grown significantly since mobile service inception in the early 2000s, over half of the population remains unconnected. As the focus of governments and mobile carriers shifts to driving Internet adoption, mobile phone serving as a facilitating condition for the use of traditional and over-the-top (OTT) services is likely to have multiple implications. Using conjoint analysis, this article presents insight into the impact of access to OTT services on users' demand interrelationship and willingness to pay for mobile services. Critical determinants of user preferences for mobile services in Nigeria are identified and the policy implications for regulators and mobile carriers are considered.

Although mobile phone adoption rates are increasing in Sub-Saharan Africa (SSA), over 500 million people representing over 50% of the population were unconnected in 2018.1 Internet adoption rates in the region are significantly lower, with less than a third of the population using the Internet.2 The Nigerian government aims to drive up Internet adoption and provide 4G/5G network coverage to 90% of the population by 2025.3 In Nigeria and other SSA countries, it is expected that broadening Internet access could create millions of job opportunities. Given that Internet access in SSA is predominantly via mobile broadband, efforts in the region to improve Internet adoption could have a significant impact on mobile phone adoption and use.

Multiple services/applications are now available on a single mobile device with applications supporting traditional mobile services such as voice and text sitting next to over-the-top (OTT) communications applications—“apps”—such as WhatsApp, Viber, and Skype. These apps enable low-cost services such as Voice over Internet Protocol (VoIP4) calling and instant messaging, which can substitute for traditional mobile services. This phenomenon of mobile phones serving as a facilitating condition for traditional and OTT services is likely to have multiple implications for mobile carriers and users.

Efforts to increase Internet adoption and use could have an impact on the use of traditional mobile services. Existing research examines this potential impact primarily by looking at demand interrelationships: how the volume of use of traditional mobile services changes as Internet use increases.5 Some of these studies show a complementary relationship,6 whereas others show a substitution relationship7 between mobile Internet use and the use of text and cellular calls. However, most of the existing literature focuses on developed mobile markets in western countries.

Competition from OTT services could impact on mobile carriers' revenues and many carriers have argued that their voice and text revenues are threatened by OTTs. Globally, between 2012 and 2018, telecom operators claim to have lost $32 billion and $49 billion in mobile voice and SMS revenues, respectively, due to OTT adoption and use.8 In developing countries where the mobile market is still growing, revenue loss could have critical implications by reducing much-needed investment in infrastructure. To defend their revenues against OTTs, African mobile operators are adopting strategies such as advocating for regulating or banning the use of OTTs; developing their own OTTs in direct competition to foreign-owned companies; bundling voice, text, and data into packages that offer OTT-like services; and embracing OTTs to gain mobile market share.9

In the form of data bundles, subsidized access, or free access only to certain websites or OTT apps, zero-rating Internet access plans have been introduced in developing countries since the 2010s. Mobile carriers have used these plans to attract new customers but little empirical research has examined the impact of OTT access on consumer preferences in emerging markets. An increase in Internet use has been shown to affect the use of traditional mobile services among existing users in developed mobile markets, but how does the offering of mobile Internet service affect consumer preferences in emerging markets? In these markets, where cost is a barrier to mobile adoption,10 low-cost OTT services may drive a substitution effect11 and also incentivize the continued use of mobile phones.

Zero-rating tariffs for services are being debated in countries around the world. Proponents argue that by providing a low-cost alternative to access the Internet, albeit with restricted content, more people will use the Internet. However, net neutrality advocates argue that if a mobile carrier offers its customers the possibility of accessing an app or website under a zero-rating plan, this discriminates against all other providers of websites and apps.12 Other techno-economic factors, such as network reliability and network effects, may also affect adoption rates. For example, users may be more likely to substitute cellular calls with VoIP if the network bandwidth is sufficient and if their family or friends use VoIP.

Relatively little attention has been given to the preferences of consumers in the face of a rapidly changing mobile industry. In developed countries where most studies on mobile adoption and use are focused, the mobile market is saturated and the impact of changes in service quality or access on the subscription rates is generally found to be minimal. In contrast, the mobile market in emerging markets is snowballing, and the impact of changes in service quality and access could be significant with practical relevance for mobile users, carriers, and regulators. The analysis of historical data to investigate the relationships between the demand for text and for OTT services is of limited help in understanding of how users are likely to respond to future trends in the market. For example, a good-quality VoIP call requires between 90 and 156 kbps and this is supported by 3G network speeds. However, Internet speeds may fluctuate due to environment and network factors and users may not use VoIP if the service is unreliable. An analysis using 3G historical data would provide a limited insight since 4G and 5G services are becoming prevalent in developed countries. Therefore, it is important to examine the possible effect of user preferences for mobile services provided with 4G/5G networks in emerging markets such as those in SSA.

This study examines the impact of selected techno-economic factors on user preferences for mobile services in Nigeria and, specifically, the impact of access to OTT services on user preferences for mobile services.

  1. How does access to OTT impact user preferences for traditional mobile phone services? In particular, how does access to OTT affect the likelihood of consumer mobile subscription and mobile service demand?

  2. How does affordability affect user preferences for adopting mobile phone services?

Analysis of stated preferences of a sample of mobile Internet users in Nigeria in a choice-based conjoint (CBC) experiment is used to answer these questions. The study extends research on how access to OTT affects mobile phone use, the demand interrelationship between Internet services and traditional mobile services, and the impact of zero-rating services on consumer preferences.

Nigeria plays a vital role in the broader SSA mobile market. It is an ideal country to study since between 2019 and 2025, there are projected to be approximately 170 million additional subscribers in the SSA mobile market and 19% of these are expected to come from Nigeria as compared to Ethiopia (11%), Democratic Republic of Congo (9%), Tanzania (6%), and Kenya (5%).13 Mobile, in this study, refers to a device used to call and/or text over a cellular network and the mobile services in this study are limited to voice (cellular calls), text (text messaging over a cellular network), OTT (using VoIP and instant messaging apps such as WhatsApp and Skype), and general mobile Internet use.

In the section “Literature Review,” we summarize the relevant literature. The section “Methods” outlines our methodology and describe our sample. Our results are presented in the section “Results” with a discussion of the findings and of the policy implications in the section “Discussion,” followed by the “Conclusion.”

Literature Review

Net Neutrality and Zero-Rating

Net neutrality refers to the principle that Internet service providers (ISPs) should enable equal access to content and apps without preference or bias and without charging content providers for sending information to consumers. These principles are referred to as the “nondiscrimination rule” and the “zero-price rule,” respectively.14 The “nondiscrimination rule” includes not allowing ISPs to impose a nonneutral data cap, for example, zero-rating of data for some content.

Many developed countries have implemented net neutrality rules, but the question as to whether or not zero-rating infringes on net neutrality is widely debated. Net neutrality advocates argue that if a mobile carrier offers its customers the possibility of accessing an app or website under a zero-rating plan, this discriminates against all other websites and app providers.15 They also argue that a new Internet user is likely to be manipulated and to assume that the Internet is entirely made up of the websites and apps that they can access.16 Proponents of zero-rating argue in contrast that by providing a low-cost alternative to access the Internet, albeit with restricted access to content, more people will use the Internet, especially in developing countries where data costs relative to income are high.17 They also suggest that zero-rating offers a gateway to the Internet and helps to bridge the digital divide by enabling restricted access to online content for millions of previously unconnected people.

In the form of data bundles, subsidized access to certain websites or OTT apps, or free access to popular websites, mobile carriers have used zero-rating plans to attract new customers. Still, not all developing countries have allowed zero-rating pricing plans. In India, strict net neutrality laws exist banning the throttling, blocking, and zero-rating of Internet data, with some exceptions. In 2016, the Indian government banned Facebook's Free Basics project that allowed free access to selected websites in developing countries, citing the unfair advantage it gave to wealthy American companies over local startups.18 This view has not been mirrored in other developing countries and as of February 2021, Free Basics was reportedly available in sixty-five countries, including thirty African countries.

In Africa where the data costs are high and the mobile market is dominated by a handful of players within each nation, zero-rating has been embraced as a means of driving mobile service adoption and enhancing competition. Indeed, researchers argue that net neutrality rules are less likely to be required in developing countries and that strict enforcement could hinder competition in the market.19 Although African countries have seen significant growth in OTT adoption and use with zero-rating plans, empirical studies on the impact of OTT are scarce and the voice of consumers remains an underrepresented in policy discussions.

The Impact of OTT on the Telecommunications Industry

OTT services have disrupted the global telecommunications industry. As consumers increasingly rely on OTT for communication, mobile carriers have seen an increasing shift in their revenue streams from predominantly voice and text to data. The majority of the studies examine the relationship between OTT and traditional mobile services (voice and text) with divergent results. For example, a study of Italian mobile users found that the use of VoIP negatively correlated with voice calling and texting, indicating a substitution effect.20 However, this work was based on self-reports, which may not be consistent with the usage patterns of early or future adopters of mobile services. The results of this study were not validated by research using handset data from 183 early smartphone adopters in Finland between 2008 and 2010. In this case there was no clear evidence of a substitution effect between mobile Internet services, such as VoIP and instant messaging, and traditional mobile operator-provided communication services.21

In studies between 2010 and 2017, Gerpott used consumer billing records of a major telecommunication company in Germany and an Arabian Gulf Corporation Council (GCC) country. The work in 2010, in Germany, suggested that mobile Internet use had a significantly negative effect on text traffic and no effect on voice traffic,22 although he also found that the effect on mobile carrier revenue was minimal. Gerpott's later work between 2014 and 2016, also in Germany, suggested that mobile Internet and voice are complementary and associated with a significant increase in text use,23 and this finding was confirmed in studies using data from a GCC country,24 Norway,25 and Thailand.26 In addition, results from a qualitative study in Germany revealed that 50% of OTT users substituted OTT for traditional mobile services, whereas the other 50% used it as a complement.27

In 2011, the Netherlands and South Korea were found to have experienced a significant shift from SMS to OTT.28 A carrier in the Netherlands saw its SMS revenues experience a 33% increase in the third quarter of 2010, falling to an 11% decrease in the second quarter of 2011. Over the same period, WhatsApp's penetration in the country (measured as the number of smartphones with the app installed) grew from 5% to 85%. In South Korea, a similar trend occurred as SMS traffic declined by over 50% while downloads of Kakao Talk, an OTT app, increased.

In SSA, research has focused more on identifying the strategies that mobile carriers can adopt in response to a changing telecommunications industry. Defensive strategies such as regulating or throttling OTT use and strategies such as embracing OTTs, developing their own OTTs, or introducing new business models offering zero-rating plans have been recommended.29 In this region, there is also some evidence that beyond protecting the revenues of mobile carriers, zero-rating plans may lead to better outcomes for consumers in terms of estimated consumer surplus.30

The Determinants of Technology Adoption

Global Determinants of Technology Adoption

Technology adoption and use are studied in multiple disciplines, including psychology, sociology, and information systems. The majority of these studies are focusing on the developed world, although some studies have explored the digital divide at the country level and difference in the Internet and mobile phone adoption rates in developed and developing countries. These studies suggest a substantial overlap in the key determinants of mobile phone and Internet adoption, such as per capita income, urbanization, and the spread of telecommunications infrastructure at the country level. Developed countries with higher per capita income and more extensive network coverage have higher mobile phone adoption rates with positive correlations between a well-developed fixed telephone infrastructure, higher population density, higher average educational levels, and higher mobile phone penetration rates.31

Although the literature on country-level determinants in African countries is limited, some findings are worth highlighting. Results from a study using data from 1995 to 2000 in forty SSA countries showed that gross domestic product (GDP) per capita and per capita investment in telecommunications infrastructure are important factors impacting on Internet diffusion in the region.32 Besides the spread of Internet infrastructure, high urbanization also has been found to have a positive effect on Internet adoption in SSA33 as does performance on Human Development Index (HDI) metrics such as literacy, health, and income.34

Individual Determinants of Technology Adoption

Theoretical models developed to explain differences in attitude and behavior toward technology adoption at the individual level include the technology acceptance model, the social cognitive theory, and the unified theory of acceptance and use of technology (UTAUT). These theories propose constructs that explain individual intentions to use a technology, which, in turn, is expected to lead to actual use. The UTAUT explains up to 70% of the variance in intent to use a technology and proposes four constructs,35 three determining intent to use (performance expectancy,36 effort expectancy,37 and social influence38), and the fourth, facilitating conditions39 joined with behavioral intention, determining usage behavior.40 The relationships between these constructs and technology adoption are affected by age, gender, experience, and voluntariness of use. The UTAUT model has been used to understand ICT acceptance in SSA. For example, a study in Angola showed that performance expectancy and facilitating conditions were significant drivers of individual-level ICT adoption.41 Another study in Ghana showed that effort expectancy had a positive impact on ICT adoption.42 Some studies have directly compared the adoption of mobile phones and/or the Internet across different sociodemographic groups in SSA finding that some factors drive mobile phone or Internet adoptions, not both. For example, evidence from a household survey from Gabon showed that income influences mobile phone adoption, whereas level of education and computer skills impact Internet use.43

Given that the Internet is predominantly accessed via mobile phones in SSA, some determinants of mobile and Internet adoption have been found to have comparable effects, whereas others have the opposite impact on mobile adoption compared to Internet adoption. For example, the impact of gender on technology adoption in Africa remains consistent across mobile phone and Internet adoption, with males being significantly more likely to own a phone or use the Internet.44 In contrast, although age has been found to have a positive correlation with mobile adoption, it has been found to have a negative correlation with Internet use.45 Furthermore, older Internet users have been shown to be less likely to use the Internet for leisure activities.46

The affordability of computers and Internet access are key issues that affect the adoption and use of mobile phones and the Internet in SSA. In the early 2000s, the rapid growth in mobile subscriptions was surprising considering the high poverty rates and the relatively high cost of phones and services. For example, in Kenya the cheapest phone was the equivalent of half of a monthly income.47 Research in 2015 showed that affordability was a crucial barrier to Internet adoption in the SSA region.48 Empirical studies of consumer willingness to pay (WTP) for digital technology in SSA are also relatively limited and it cannot be assumed that because this is a low-income region there is a low WTP for mobile services, or that if WTP is low, it cannot be overcome by other factors. It has been shown that in low-income populations in the United States, there is a WTP for broadband,49 and this study examines whether this holds for an emerging market such that the WTP for mobile Internet services may be higher than for traditional services of text and even voice calling.

Determinants of OTT Adoption

Research on the factors that influence the adoption of OTT is limited, but evidence from the Netherlands, France, and Spain has shown that users are willing to adopt converged OTT services (VoIP, instant messaging, and photo sharing) when they are satisfied with reliability, security, and the interoperability of OTT and other services on their device.50 Using an extension of the technology acceptance model to analyze consumer behaviors in Korea, Shin found that call and service quality, mobility, and coverage significantly affected VoIP's adoption.51 An analysis of OTT adoption in the Netherlands and South Korea indicated that core drivers were the technology readiness of the market, that is, 3G or better networks and high smartphone penetration; the cost incentive of OTTs, that is, users have an opportunity to arbitrage data and text costs; network effects, that is, an individual's social group uses OTT; and the strength of the OTT alternative, that is, there is a dominant OTT player in the market.52

Only a handful of studies have examined factors that drive OTT adoption and use in SSA. In Kenya, for example, the user-friendly features and low cost of OTTs were critical factors that appeared to prompt consumers to shift away from voice and text services.53 A Nigerian study suggested that OTT messaging apps were the preferred means of communications,54 but that, due to the reliability of voice calls, most users defaulted to cellular voice to make urgent calls.

Technical skills have been found to be a significant predictor of the type and frequency of Internet use.55 However, despite OTT adoption being a type of Internet use, the impact of technical skills on OTT adoption has not been studied extensively. Given that the basic features of OTTs (VoIP and instant messaging) are similar to cellular voice and text messaging, in this study we do not expect that advanced digital skills are required to use OTTs and assume that the skills needed to operate a feature phone are similar to those required to use OTTs.

The studies in developed countries on the interrelationship between demand for OTT and telecommunications services emphasize that time and budget limit customer use of different mobile services with the implication that the use of one service could reduce the use of other services.56 In our study, we use a choice-based experiment to investigate user preferences for voice, text, and data (OTT) under time and budget constraints and we also examine how user preferences change with changes in OTT reliability using the proxy of network bandwidth.

An Overview of the Nigerian Mobile Market

With an estimated population of 190 million in July 2019, Nigeria is the most populous country, the second-largest economy, and the largest mobile market in SSA. Mobile telecommunications dominate the telecommunications market with access to fixed telecommunications consistently remaining at less than 0.2%.57 As shown in Figure 1, since its inception in 2001, the Nigerian mobile market has grown to over 170 million subscriptions in 2018; a connection rate of 89%.58 However, the unique subscriber rate is only 45% given that, on average, each subscriber has two subscriber identity modules (SIMs). Although multiple SIM ownership is common with a global average of 1.6 SIM cards per subscriber, in developing countries, each subscriber has an average of two SIMs.59. In SSA, many consumers use multiple SIMs often from different mobile carriers in order to make use of the best network coverage and call quality, and to maximize pricing arbitrage.60 Significant variations in penetration exist within Nigeria with mobile subscription rates ranging from 40% to 170% across thirty-six states.61 The adoption of mobile Internet similarly varies, with subscription rates ranging from 20% to 110% across Nigerian states.62 The Nigerian telecommunications providers consists of four major mobile carriers—MTN (40%), Airtel (27%), Globacom (27%), and 9mobile (6%).63

FIGURE 1

Mobile Subscription Growth in Nigeria (2000–2018).

Data Source: The International Telecommunication Union (ITU).

FIGURE 1

Mobile Subscription Growth in Nigeria (2000–2018).

Data Source: The International Telecommunication Union (ITU).

Close modal

Competition, the introduction of 3G services in 2007, and regulatory reforms are a few of the key drivers of rapid growth in the mobile industry since the early 2000s.64 Major mobile carriers offered 4G Long-Term Evolution (LTE) in 2016 when their market share was threatened by independent mobile ISPs (such as Smile and Spectranet) offering 4G LTE from 2014. Despite this rapid growth, 3G and 4G LTE networks65 remain restricted to large urban areas, with 93% of the population covered by a 2G signal, 78% covered by a 3G signal, and 45% covered by 4G.66 It is likely that addressing other consumer barriers to Internet adoption, for example, by increasing access to 3G and 4G LTE networks and reducing the cost of smartphones could enable more Nigerians to access Internet apps and services.

As mobile markets are moving to 4G and 5G networks that facilitate Internet use and are optimized for the use of OTT apps, the Nigerian government is aiming to provide 4G/5G network coverage to 90% of the population by 2025.67 This transition could promote a decline in the use of voice and text and growth in the use of OTT platforms, a shift seen in other emerging markets such as Brazil and China where approximately 60% and 70% of the population, respectively, now use OTT more than text services.68

Methods

We designed a CBC experiment to elicit user preferences for mobile services. This approach assumes that consumers have predetermined preferences independent of choices available in the experiment or in the current market.69 By asking respondents to choose a preferred alternative, which is similar to their selection process in the market, CBC methods have closely matched market demand estimates.70 CBC methods have been used in analyzing consumer preferences for communication technologies as a function of technology security,71 privacy, and ease of use72 of different mobile devices such as smartphones versus feature phones.73 Although CBC is an appropriate method for our study, we recognize that CBC results depend on available alternatives and this limits the generalizability of our results. However, this method does provide a basis for developing further insight as we cover in the “Discussion” section.

Experiment Design

Using the CBC method, we designed a survey using Sawtooth Software.74 Respondents completed the survey online using local hosts on tablets or using paper and pencil. The survey consisted of four parts with fifty-one questions. First, the participants were introduced to the context of the study by providing background information on mobile and OTT use. The second part, which was the CBC experiment, consisted of fourteen choice questions. Each choice task was made up of three alternatives; two alternatives had mobile plans and the third alternative was “none”—which meant the respondent indicate a preference for going without mobile service rather than choosing any mobile plan.75 The “none” alternative indicated the participant would be unable to make calls or send messages but could receive calls and text messages. The “none” alternative was included to model the current market in Nigeria where there is no access fee for mobile users. The last part of the survey consisted of demographic and follow-up questions.

A sample of the choice interface presented to respondents is shown in Figure 2. Participants were asked to choose one option out of three choices: two mobile plans or no mobile service. They were asked to assume that their current mobile plan ended the next day and that the plans offered in the choice interface were the only options available for the next month. Each mobile plan was characterized by a combination of four attributes (see Table 1). The levels of the attributes were selected to mirror the current market, but to understand future trends in the mobile market, some levels not offered in Nigeria were also included. For example, 4G LTE, unlimited voice, unlimited text, and bundling of services were not widely available in Nigeria when this study was conducted.

FIGURE 2

Example Choice Screen for a Respondent. Each Task Was Displayed on a Separate Screen/Page.

FIGURE 2

Example Choice Screen for a Respondent. Each Task Was Displayed on a Separate Screen/Page.

Close modal

Measures of the Attributes of the Mobile Plans

TABLE 1
Measures of the Attributes of the Mobile Plans
AttributeDescriptionValue
XjPRICE
 
Monthly price for the bundle of mobile services $6.3/$12.7/$25.376 
XjINT
 
Bundle of mobile Internet attributes; speed, data limit, and application/web access Described in Table 2  
XjCCALLS
 
Minutes of cellular calls made in a month 0/300/500/unlimited 
XjTEXTS
 
Number of text messages sent in a month 0/300/500/unlimited 
AttributeDescriptionValue
XjPRICE
 
Monthly price for the bundle of mobile services $6.3/$12.7/$25.376 
XjINT
 
Bundle of mobile Internet attributes; speed, data limit, and application/web access Described in Table 2  
XjCCALLS
 
Minutes of cellular calls made in a month 0/300/500/unlimited 
XjTEXTS
 
Number of text messages sent in a month 0/300/500/unlimited 

Description of the Mobile Internet Attribute and its Levels

TABLE 2
Description of the Mobile Internet Attribute and its Levels
Mobile Internet ValueSubattributes
Volume of DataNetwork SpeedAvailable Access
None 0 GB None None 
4G-OTT 3.5 GB 4G IM and VoIP 
2G-Limited 3.5 GB 2G Everything 
2G-Unlimited Unlimited 2G Everything 
3G-OTT 5 GB 3G IM and VoIP 
3G-Unlimited Unlimited 3G Everything 
2G-High-Limit 5 GB 2G Everything 
Mobile Internet ValueSubattributes
Volume of DataNetwork SpeedAvailable Access
None 0 GB None None 
4G-OTT 3.5 GB 4G IM and VoIP 
2G-Limited 3.5 GB 2G Everything 
2G-Unlimited Unlimited 2G Everything 
3G-OTT 5 GB 3G IM and VoIP 
3G-Unlimited Unlimited 3G Everything 
2G-High-Limit 5 GB 2G Everything 

The first attribute is the mobile plan's economic cost to the user, the monthly price. To determine the monthly price levels to include in the experiment, we estimated the cost of the chosen attribute levels using the current market price and self-reported monthly expense estimates for mobile Internet, voice calling, and text from the survey pretest.

The second attribute is the mobile Internet, which was described using the volume of data, Internet network speed, and Internet service access referring to what websites or apps could be accessed under a plan. Various levels representing the characteristics in different markets were included in the experiment (see Table 2).

Network characteristics in mobile markets across the world were included such as 3G and 4G networks. Attributes predominantly limited to emerging markets were included, for example, low-cost Internet plans allowing limited access only to OTT and social networks. OTT access was presented to respondents as “Internet access: IM & VoIP,” which is Internet access that is restricted to instant messaging and VoIP. To reduce respondent error, the choice bundles were selected to ensure unrealistic attribute combinations, for example, a 4G LTE speed and no Internet access could not exist within a bundle.

The third attribute, voice, describes minutes of outgoing calls in the mobile plan. Finally, the fourth attribute, text, describes the number of SMS sent via a cellular connection in a month. The levels of voice and text were informed by price levels to ensure realistic combinations.

Checks were undertaken to determine whether respondents were paying attention to the choice tasks in the experiment. Before completing the choice tasks, we provided information on OTT apps and then asked follow-up questions on the information to see whether the respondent understood. We also included two fixed choice tasks with a dominant option in the conjoint choices to test whether the respondents were randomly selecting their choices.77

Sample

Since this study aimed to investigate the impact of access to OTT on user preferences for mobile services, we sampled active mobile users in Nigeria, bearing in mind that in SSA, mobile service is widely adopted across education, age, and income levels, but young mobile users with at least a secondary school education are more likely to be Internet users.78

Respondents were recruited between June and August 2016 from the States of Lagos79 and Delta80 in Nigeria using a site-based sampling method (N = 390). Participants were recruited to represent users in urban and rural regions and with diverse socioeconomic and employment backgrounds. Recruitment took place in higher educational institutions and public places near shopping malls and markets. Participation was limited to users who could read and write in the English Language, the official language in Nigeria.

Table 3 presents the self-reported demographic characteristics of the sample by urban and rural respondents and as compared to the Nigerian population. Although an attempt was made to recruit a representative sample, there are some deviations from national demographics and that reflects demographic characteristics of most Internet users in Nigeria, young and educated. Approximately 20% of our sample had completed high school or a lower educational level, compared to 86% of the Nigerian population. Furthermore, there are disparities in the age distribution of the sample as compared to the national distribution. Over 50% of the Nigerian population is below 20 years old, but the young population is underrepresented in our study since respondents younger than 18 years of age were excluded. In our analysis, we weighted the data to better represent the Nigerian population's age and education distribution. Our sample includes rural and urban respondents with rural respondents being older, having lower educational qualifications, shorter duration of Internet use, and lower monthly spending on mobile services as compared to urban respondents.

Distribution of the Individual Demographic Characteristics of the Sample, Split into Urban and Rural Respondents in Comparison to the National Statistics. The Sample is Weighted to the Nigerian Population on Age and Education

TABLE 3
Distribution of the Individual Demographic Characteristics of the Sample, Split into Urban and Rural Respondents in Comparison to the National Statistics. The Sample is Weighted to the Nigerian Population on Age and Education
Rural
(55%)
Urban
(45%)
Sample
(n = 388)
Nigeria
(n = 200 Million)
Weighted Mean
Gender      
°Female 44% 51% 47% 49%  
°Male 51% 49% 50% 51%  
°Did not respond 5% 0% 3%   
Age (years)      
°18–24 13% 13% 13% 14% 0.27 
°25–34 40% 57% 48% 14% 0.27 
°35–44 19% 23% 21% 10% 0.20 
°45–54 14% 5% 9% 6% 0.18 
°>54 9% 2% 6% 7% 0.14 
Did not respond 5% 0% 3%   
Highest Education      
°High school or below 29% 13% 21% 86% 0.86 
°Bachelor's or above 66% 87% 76% 14% 0.14 
°Did not respond 5% 0% 3%   
Monthly Income      
°Below $160 37% 28% 33%   
°$160–$320 32% 36% 34%   
°$321–$630 14% 20% 17%   
°Above $630 10% 15% 12%   
°Did not respond 7% 1% 4%   
Duration of Internet Use      
°Less than 1 year 17% 8% 13%   
°1–3 years 20% 14% 17%   
°4–6 years 28% 26% 27%   
°Over 6 years 29% 51% 39%   
°Did not respond 6% 1% 4%   
Monthly ARPU $13 $19 $16   
Rural
(55%)
Urban
(45%)
Sample
(n = 388)
Nigeria
(n = 200 Million)
Weighted Mean
Gender      
°Female 44% 51% 47% 49%  
°Male 51% 49% 50% 51%  
°Did not respond 5% 0% 3%   
Age (years)      
°18–24 13% 13% 13% 14% 0.27 
°25–34 40% 57% 48% 14% 0.27 
°35–44 19% 23% 21% 10% 0.20 
°45–54 14% 5% 9% 6% 0.18 
°>54 9% 2% 6% 7% 0.14 
Did not respond 5% 0% 3%   
Highest Education      
°High school or below 29% 13% 21% 86% 0.86 
°Bachelor's or above 66% 87% 76% 14% 0.14 
°Did not respond 5% 0% 3%   
Monthly Income      
°Below $160 37% 28% 33%   
°$160–$320 32% 36% 34%   
°$321–$630 14% 20% 17%   
°Above $630 10% 15% 12%   
°Did not respond 7% 1% 4%   
Duration of Internet Use      
°Less than 1 year 17% 8% 13%   
°1–3 years 20% 14% 17%   
°4–6 years 28% 26% 27%   
°Over 6 years 29% 51% 39%   
°Did not respond 6% 1% 4%   
Monthly ARPU $13 $19 $16   

Source: Nigerian Bureau of Statistics, 201981 and the National Population Commission, 2018.82

Preference Modeling and Analysis

The responses to the survey were analyzed using a random utility model in which, for each individual n, with J alternatives, the utility, Unj, derived from an alternative j is derived as a sum from two components: (1) an observable part, Vnj, which is a function of the attributes of the CBC questions, Xnj, and the individual-specific attributes, Snj, and (2) an unobserved part, εnj, which is assumed to be random and independent and identically distributed. A model that is linear in parameters and is assumed to have the basic form:
Unj=β0+β1XjPRICE+β2XjINT+β3XjCCALLS+β4XjTEXT+εnj,
(1)

where each β represents the coefficient for each attribute X.

We used three random utility models: the multinomial logit (MNL), mixed logit (ML), and nested logit (NL) models. The MNL model assumes that all alternatives have equal weight, and participants do not discriminate between them. To model how people can be assumed to make choices, we estimate the NL model where participants would first decide whether they want a mobile plan. If they want a mobile plan, then they make a selection from two available bundles.83 The nested logit consists of two nests: one the “none” alternative and the other the two mobile plans. The ML model relaxes some of the assumptions of the NL model and was estimated using 1000 draws to allow for individual heterogeneity across preferences for the services.

Directly comparing the logit coefficients provided little insight into user preferences for different mobile service bundles and the tradeoffs informing their choices. To strengthen the basis for interpreting our results, we present them as conditional probabilities that respondents will select a mobile bundle plan with a specified combination of attributes. The conditional choice probability84 of a mobile service plan was computed for an NL model using:

P(nm)=PnP(m|n),
(2)

where there are n nests and each nest contains m alternatives.

The probability of a nest n being chosen is given as
P(n)=exp(λVn)/n'=1Nexp(λVn')
and the probability of an alternative m being chosen from within the nest n is given as
P(m|n)=exp(βTXnmλ)/exp(Vn)
. λ is the dissimilarity coefficient, β is the coefficient of the regressor X, and Vn is the inclusive value for nest n and is given as
Vn=log(m=1Nnexp(βTXnm))
.

Estimating WTP

Different methods have been used to estimate WTP in various industries such as health,85 energy,86 food,87 and telecommunications.88 Studies estimated WTP, for example, by directly asking users how much they would pay89; by applying iterative bidding direct contingent valuation90; or by using a discrete choice experiment.91 In this study, WTP for mobile services is estimated by using a discrete choice experiment, which has the advantage that WTP can be estimated for services that do not exist in the market. We acknowledge that respondents may experience a bias and overstate their WTP, given that there is no commitment to pay for the service. The WTP for each mobile service was computed using the coefficients of the logit model—the ratio of the coefficient of that service to the coefficient of the price:
WTP=βSERVICE÷βPRICE
(3)

Results

The estimation results for the MNL, ML, and NL models are reported in Table 4. The NL and ML models address the effects of linear attributes by capturing the none option in a different nest. We compared the performance of the models using the Hausman–McFadden test and the Akaike information criterion (AIC). The Hausman test for independence of irrelevant alternatives (IIA) indicates that IIA does not hold and that the NL model differs significantly from the MNL model. The ML model has the lowest AIC, indicating that is the best fit. However, for ease of interpretation, we use the NL model for estimating the main effects and the ML model to examine differences across demographic groups.

Estimation Results of the Multinomial, Nested Logit, and Mixed Logit Models. The Baseline Mobile Service Plan is the “None” Option Which Sets the Values of All Attributes to 0

TABLE 4
Estimation Results of the Multinomial, Nested Logit, and Mixed Logit Models. The Baseline Mobile Service Plan is the “None” Option Which Sets the Values of All Attributes to 0
AttributeAttribute LevelMNL ModelNL ModelML Model (Mean)ML Model (Standard Deviation)
Price  −0.071 (0.003)*** −0.022 (0.001)*** −0.36 (0.02)***  
Internet 2G-High_Limit 0.143 (0.072)** 0.216 (0.017)*** 1.23 (0.11)*** 0. 30 (0. 17) 
 2G-Limited 0.196 (0.074)*** 0.207 (0.016)*** 1.04 (0.11)*** 0.52 (0.18)** 
 2G-Unlimited 0.304 (0.072)*** 0.254 (0.017)*** 1.44 (0. 12)*** 0.65 (0.17)*** 
 3G-OTT 0.240 (0.072)*** 0.257 (0.017)*** 1.35 (0.12)*** 0.78 (0.18)*** 
 3G-Unlimited 0.764 (0.072)*** 0.452 (0.021)*** 2.04 (0.13)*** 1.16 (0.16)*** 
 4G-OTT 0.371 (0.068)*** 0.237 (0.016)*** 1.31 (0.11)*** 0.56 (0.19)** 
Calls 300 0.499 (0.057)*** 0.247 (0.015)*** 1.12 (0.09)*** 1.12 (0.11)*** 
 500 0.571 (0.059)*** 0.269 (0.016)*** 1.26 (0.10)*** 1.46 (0.12)*** 
 Unlimited 0.599 (0.061)*** 0.299 (0.017)*** 1.29 (0.10)*** 1.40 (0.13)*** 
Texts 300 −0.063 (0.054) 0.085 (0.012)*** 0.46 (0.09)*** 1.09 (0.10)*** 
 500 −0.010 (0.061) 0.087 (0.013)*** 0.43 (0.10)*** 1.07 (0.14)*** 
 Unlimited 0.201 (0.057)*** 0.171 (0.013)*** 0.87 (0.09)*** 1.05 (0.11) 
iv.mobile  0.316 (0.016)*** 
Log- likelihood  −5,478 −5,421 −3797 
AIC  10983 10873 7647  
Respondents  390 390 390 
Observations  5398 5398 5398 
AttributeAttribute LevelMNL ModelNL ModelML Model (Mean)ML Model (Standard Deviation)
Price  −0.071 (0.003)*** −0.022 (0.001)*** −0.36 (0.02)***  
Internet 2G-High_Limit 0.143 (0.072)** 0.216 (0.017)*** 1.23 (0.11)*** 0. 30 (0. 17) 
 2G-Limited 0.196 (0.074)*** 0.207 (0.016)*** 1.04 (0.11)*** 0.52 (0.18)** 
 2G-Unlimited 0.304 (0.072)*** 0.254 (0.017)*** 1.44 (0. 12)*** 0.65 (0.17)*** 
 3G-OTT 0.240 (0.072)*** 0.257 (0.017)*** 1.35 (0.12)*** 0.78 (0.18)*** 
 3G-Unlimited 0.764 (0.072)*** 0.452 (0.021)*** 2.04 (0.13)*** 1.16 (0.16)*** 
 4G-OTT 0.371 (0.068)*** 0.237 (0.016)*** 1.31 (0.11)*** 0.56 (0.19)** 
Calls 300 0.499 (0.057)*** 0.247 (0.015)*** 1.12 (0.09)*** 1.12 (0.11)*** 
 500 0.571 (0.059)*** 0.269 (0.016)*** 1.26 (0.10)*** 1.46 (0.12)*** 
 Unlimited 0.599 (0.061)*** 0.299 (0.017)*** 1.29 (0.10)*** 1.40 (0.13)*** 
Texts 300 −0.063 (0.054) 0.085 (0.012)*** 0.46 (0.09)*** 1.09 (0.10)*** 
 500 −0.010 (0.061) 0.087 (0.013)*** 0.43 (0.10)*** 1.07 (0.14)*** 
 Unlimited 0.201 (0.057)*** 0.171 (0.013)*** 0.87 (0.09)*** 1.05 (0.11) 
iv.mobile  0.316 (0.016)*** 
Log- likelihood  −5,478 −5,421 −3797 
AIC  10983 10873 7647  
Respondents  390 390 390 
Observations  5398 5398 5398 

Note: p < 0.1

*

p < 0.05

**

p < 0.01

***

p < 0.001.

How Does Access to OTT Impact Users' Preferences for Traditional Services?

Our primary aim was to investigate the impact of access to OTT services on people's preferences for mobile services. Our results show that users in our sample were more likely to select a plan that provides only OTT service than a plan that provides text. Figure 3 compares the conditional choice probabilities of a traditional service and going without mobile service across two scenarios: (1) the “no OTT” scenario—two options of the same traditional mobile service and a none option and (2) the “OTT” scenario—a traditional mobile service, 4G-OTT service, and a none option. In the “no OTT” scenario, both traditional mobile services are identical; hence, their choice probabilities are equal. “None” means no service and “Trad” represents the traditional services.

FIGURE 3

Comparing the Choice Probabilities of Traditional Services and No Service Across Scenarios with 4G-OTT as an Alternative and When It Is Not. Price Is $10, and the Levels of the Traditional Services Are in the Upper x Axis.

FIGURE 3

Comparing the Choice Probabilities of Traditional Services and No Service Across Scenarios with 4G-OTT as an Alternative and When It Is Not. Price Is $10, and the Levels of the Traditional Services Are in the Upper x Axis.

Close modal

When the cost of the mobile plan was set at $10, introducing 4G-OTT had the greatest impact on the choice probabilities of text services. When 4G-OTT was introduced, the choice probability of 300 texts was reduced by 60% (from 52% to 21%). However, for voice, or a combination of voice and text, introducing 4G-OTT was associated with a smaller reduction in the choice probability of the traditional service. The smallest effect occurred when the traditional service was the bundle of unlimited voice and text: the choice probability reduced by 33% and respondents were twice as likely to choose the plan offering the bundle of unlimited voice and text, 40%, 95% confidence interval [CI] [39%, 40%] than the 4G-OTT service, 19%, 95% CI [19%, 20%]. There is no evidence that introducing 4G-OTT affected respondents' preferences for subscribing to a mobile service: there were no significant changes in the choice probability of no service.

The 95% CIs for the coefficient estimate for OTT for each user were computed to understand how respondent preferences differed relative to each other indicating there was no significant difference.

The impact of the other Internet plans on respondent preferences was also examined. As seen with 4G-OTT, other Internet plans reduced the choice probability of text services by over 50%, indicating that our respondents had a stronger preference for the Internet plans than text services. When the 3G-unlimited plan (3G, unlimited GB, and access to all websites) was introduced, the choice probabilities of voice and the bundle of 300 voice minutes and 300 texts were reduced by over 50%. However, our respondents were ambivalent about 3G-unlimited and the bundle of unlimited voice and unlimited texts. The 3G-OTT plan (3G, 5GB, and access to only IM & VoIP) and the 2G-unlimited plan (2G, unlimited GB, and access to all websites) had a similar effect as the 4G-OTT plan on respondent preferences for traditional mobile services.

How Does Affordability Affect Users' Preferences?

The price was increased to $30 to examine the effect of affordability on user preferences. For our respondents, at $30, the choice probability of no service increased, ranging between 50% (two plans with unlimited voice and text were the alternatives) and 60% (when two plans with 300 texts were the alternatives) for all services and service levels. Respondents had a significantly higher choice probability for no service than for the other available alternatives in both scenarios.

WTP for different mobile services is plotted in Figure 4. The results show that the WTPs for 3G-unlimited was significantly higher than WTP for voice calling, text services, and all other Internet plans. The WTP for voice is higher than the WTP for 4G-OTT, but the difference was not significant. The individual estimates for WTP have greater variation for the Internet services as compared to traditional services and the higher volumes of voice and text yielded a higher WTP than lower volumes, all other things being equal.

FIGURE 4

Boxplot Showing the Willingness to Pay ($/Month) for Different Mobile Services. The x Axis Shows the Different Attribute Levels of the Mobile Services and the WTP Is Shown on the y Axis.

FIGURE 4

Boxplot Showing the Willingness to Pay ($/Month) for Different Mobile Services. The x Axis Shows the Different Attribute Levels of the Mobile Services and the WTP Is Shown on the y Axis.

Close modal

How Do Preferences and WTP Vary Across Demographics?

Interactions among the attributes of the mobile plan and demographics (gender, age, education level, income, and location) are shown in Supplemental Table 1 in the Appendix. Age, income, and location were critical moderators of user preferences and WTP for mobile services in this study.

Older and low-income respondents and those with at most high school education were more price sensitive. The preferences for Internet services varied significantly across demographic subgroups. Older and rural respondents had a lower preference for Internet services compared to young and urban respondents and those with a higher income had a greater preference for Internet plans offering full access to all websites as compared to respondents with lower incomes. There is also evidence of an interaction between gender and OTT: women had a lower preference for OTT plans as compared to men in our sample. Respondents with at least a bachelor's degree had greater preference for 3G/4G Internet plans, suggesting an interaction between educational level and broadband speeds.

Our examination of the interaction between demographics and preferences for traditional services yielded a significant result. Older and low-income respondents had a lower preference for texts compared to younger and high-income respondents and older respondents and those living in rural areas had a greater preference for voice as compared to younger respondents and those living in urban areas.

Discussion

Our results suggest that access to OTT had a significant impact on our respondents' preferences for mobile services. This impact was greatest when a text plan was the available alternative. However, there was a considerable variation in respondent preferences and this needs to be taken into consideration in interpreting implications of our results for access by Nigerian users to mobile services and OTT as well as the likely impact of access to OTT on the mobile operators.

As indicated earlier, this research focused on two research questions: (1) How does access to OTT impact users' preferences for traditional services? and (2) How does affordability affect users' preferences?

In response to the first question, our results suggest that our respondents' preferences were shifting from mobile voice/text to Internet communications. This result based on a limited sample in Nigeria is consistent with findings that access to the mobile Internet significantly reduces the number of texts sent and received in a month.92 The results of our study suggest that access to an Internet plan is likely to change user preference and that the effect is likely to be significantly higher for the Internet plans that allow access to all websites and apps on a 3G network. In our study, there was no evidence to suggest that access to OTT affects the probability of a consumer not subscribing to any mobile service. This finding differs from results in the literature indicating that zero-rating enables competition and drives up subscription rates.93

Our results also indicated that the business model (pay-as-you-go versus all-you-can-eat) was a determinant of the impact of OTT on our respondents' preferences, a finding that is consistent with previous work.94 We found that with access to OTT, the choice probability of 300 texts (pay-as-you-go) was reduced by 60%, whereas the choice probability of the bundle of unlimited minutes and texts (all-you-can-eat) was reduced by less than 33%. This suggests that the users in our sample did not have a strong preference for traditional services over OTT. A possible explanation is that the perceived benefit (cost savings) from the use of OTT in a pay-as-you-go market operates as a driver for the choice of mobile plan. This finding suggests support for advocates of zero-rating: it may expand Internet usage to consumers who would otherwise be unable to afford it.

The results of our study confirm that mobile users are price sensitive and may be willing to spend more on Internet services than on traditional mobile services. In our analysis, increasing the price from $10 to $30 significantly increased the choice probability of no service. This result offers insight into findings from previous studies that have found that economic cost is one of the critical barriers to Internet use in Nigeria and other SSA countries.95 Our results suggest that this economic cost as a barrier could be addressed by strategically leveraging zero-rating plans and offering mobile plans at different price brackets.

We also found that our respondents' preferences and WTP were impacted by demographic factors such that although age, education, and location had a significant effect on their preferences for OTT services, there was no evidence of an interaction with income. If we take age and education as a proxy for digital skills (which were not captured in this study), we might assume that the younger population of digital natives who have grown up with mobile phones, computers, and the Internet and those with a high educational level are more likely to have the digital skills to use OTT services. The sample distribution in this study was composed of older, rural respondents with lower educational levels as compared to the urban respondents and this sample distribution characteristic may explain the rural respondents' lower preferences for OTT.

Implications for the Regulator

Increasing Internet access and adoption is a key policy objective in Nigeria. To meet this objective, the Nigerian government has embraced zero-rating and enabled mobile carriers to offer special Internet plans with low-cost access limited to OTT apps. The results of this study offer new insight into the likely impact of zero-rating on user preferences. In this study, there is evidence that mobile users are price sensitive indicated by the finding of a considerably higher probability of a respondent subscribing to a mobile plan when the price is $10 compared to when it is $30. Evidence of price sensitivity suggests that policies aimed at reducing the cost of Internet services for users such as granting tax breaks to mobile carriers, promoting competition, or imposing a price ceiling could help to reduce the economic cost barrier.

In our study, there was no evidence that introducing OTT is likely to result in an increase in mobile phone adoption. However, the significant difference in the estimated WTP for full access to the Internet as compared to that for an OTT plan in our study suggests that a low-cost OTT plan could provide a gateway for existing mobile phone users who cannot afford a more expensive full-access plan. Furthermore, our results suggest that our respondents had a similar WTP for OTT services compared to voice. Given that mobile, and thus traditional services including voice, is widely adopted and used by the Nigerian population, our evidence suggests that a significant subset of the population might be able to afford OTT plans. In many developing countries, popularly zero-rated social networks sites, such as Facebook, have been shown to be significant drivers of Internet uptake.96 If policies that facilitate OTT access have this effect, contrary to recent policies in other developing mobile markets, the Nigerian policymakers should dismiss calls to ban OTT; ensure that any regulation of OTT service does not discourage its adoption or use; and adopt policies that aim to facilitate the conditions needed for OTT adoption (mobile and Internet access), for example, by driving investment in broadband technologies in underserved areas.

Implications for the Mobile Carriers

Mobile carriers have engaged in lobbying to encourage policymakers to regulate OTTs in a bid to avoid the erosion of their markets and loss of revenues in the global mobile industry. The results of this study are consistent with those who have argued for encouraging the mobile carriers to introduce progressive strategies in response to OTT adoption.97 Our results indicate that the OTT services can present a threat to mobile carrier text revenues, not so much to voice revenues. Based on our study's evidence of a shift in preferences toward Internet and OTT services in the Nigerian market, embracing OTT and prioritizing the provision of Internet access over text offerings is likely to benefit the carriers. In Nigeria, where Internet access is predominantly via mobile broadband, the mobile carriers are likely to be able to substitute lost revenues from traditional mobile services with new revenues generated by providing Internet access to new customers. Our analysis suggests that users are likely to be willing to pay comparable prices for OTT and voice and this suggests that zero-rating strategies could be used effectively to drive up Internet adoption among the low-income population. Insofar as our study suggests that supporting the use of OTT could drive up carrier revenues, given that people have a higher WTP for OTT as compared to text, investing in the broadband infrastructures in underserved areas could facilitate Internet access and adoption and lead to revenue growth.

As the mobile and broadband markets in SSA are developing, regulators argue that net neutrality rules, particularly that preferential pricing should not hinder the free flow of information across the Internet to encourage growth, are not a helpful debate since the evidence is accumulating that zero-rating can help stimulate growth.98 If a customer wants to buy a prioritized service, then he or she should be able to do so provided this does not affect the service offered to other consumers. In this context, by embracing the OTTs the mobile carriers can use them to help increase their customer base. For example, by offering competitive Internet plans with access limited to social networks and specific OTT communication apps such as WhatsApp and Facebook Messenger, a carrier may be able to drive adoption among the low-income subset of the population.

Mobile carriers also would benefit from offering new mobile plans using a bundled all-you-can-eat business model. By setting attractive prices for plans, in view of the price-sensitivity of mobile users as indicated by this study, and providing close to unlimited voice and text, bundled with a limited gigabytes of Internet data, these carriers are likely to limit the negative impact of OTT services on their revenues.

Conclusion

A CBC experiment has been used to examine how access to OTT is likely to impact on Nigerian user preferences for mobile services and to assess whether access to OTT should be encouraged as a driver of Internet use and mobile market revenue growth in the country.

With half of the Nigerian population unconnected it is essential to understand the factors that are likely to reduce barriers to expanding mobile service adoption and use in the context of an emerging market. In SSA countries, the focus of governments, mobile carriers, and regulators is shifting to driving Internet adoption and use, but the preferences of service users have received relatively little attention. Our study of the relationship between the demand for OTTs and voice/text mobile offerings provides evidence that a sample of Nigerian survey respondents was significantly more likely to choose OTT services than they were to choose text when presented with options. We also found that the mobile carrier business model and the price of a mobile plan were the most critical determinants of the impact of access to OTTs on user preferences for mobile services. We showed that an increase in the cost of a mobile plan from $10 to $30 increased the probability of our respondents not subscribing to a mobile plan by up to 20%. We have also demonstrated that respondents in our sample were willing to pay a similar price for OTT, full Internet access (on a 2G network), and voice service and that, compared to the price that they were willing to pay for OTT, respondents were willing to pay significantly more for full Internet access (on a 3G network) and significantly less for text service. User demographics such as age, education, income, and gender, as well as the location, have been shown to be moderating factors on the impact of OTT access on user preferences and their WTP for mobile services.

As is the case with any empirical study of this kind, there are limitations. The limitations of this study include the fact that respondents were recruited using a site-based sampling method that resulted in a sample that did not fully represent the population of mobile service users in Nigeria although an effort was made to correct for this bias by weighting the sample. This enabled us to strengthen the basis for interpreting the results, but it is acknowledged that our results will benefit from future comparisons based on a more representative sample using the same method and care should be taken in generalizing the results.

In this study although we explored the impact of access to OTT services, network bandwidth, and costs of mobile services, we did not control for the impact of network effects on the respondents' choices. In future work it will be interesting to investigate the relationship between respondent choice probabilities and the percentage of their contacts (family, friends, and business associates) who use OTT services.

A further limitation of this study is that respondents' preferences were measured as conditional probabilities using a conjoint experiment. The results were therefore dependent on the alternatives presented to them and it is likely that in a different market where service users would be presented with different alternatives, the results may differ. Nevertheless, this study has allowed for the identification of distinctive preferences and it offers a strong basis for further research, for example, by providing additional alternative choice clusters. Furthermore, although estimating user preferences using a conjoint experiment may be better than directly asking respondents for their preferences, our survey may be subject to cognitive bias. For instance, the choices of the respondents may have been affected by their prior experiences with mobile carriers or mobile services. For example, if a participant used VoIP in the past and it was not reliable due to slow connection speed, then they could decide not to choose it regardless of if it is over a 4G network. Additionally, we modeled mobile services as a bundle, whereas in Nigeria, prepaid, pay-as-you-go plans are the norm and this may have introduced cognitive bias if the respondents interpreted conjoint choices differently than we assumed. Finally, although we found that demographics such as age, gender, location, and education had some effects, other studies have shown that demographics are weakly correlated with choice preferences.99 This suggests that our results should be compared in future work with an analysis using historical customer records to examine their consistency using different methods.

Our research provides a basis for several policy recommendations including the need for policymakers in Nigeria, and SSA generally, to support measures that are likely to encourage investment in the broadband infrastructure and increase the affordability of mobile services and Internet access. This research provides a basis for future studies that test the utility of studying the techno-economic determinants of the adoption and use of mobile services using a CBC experiment, which removed the limitation of the ability of pay in the experiment, and a nested logit analytical approach. This enabled us to more closely model the way users make buying decisions. In future work, the model can be used with larger samples from Nigeria and other developing countries to develop a better understanding of the impact of mobile Internet adoption on the use of traditional mobile services in emerging markets.

Acknowledgments

The authors acknowledge Brian Sergi and Cristobal de la Maza for their helpful comments. This work was funded by the Petroleum Technology Development Fund Nigeria and the Department of Engineering and Public Policy at Carnegie Mellon University. These sponsors had no involvement in the study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

This study was first presented at TPRC48: The Research Conference on Communications, Information, and Internet Policy, February 17–19, 2021.

Appendix Supplemental Information—Sample Survey Instrument

Interactions between Stated Preferences for Attributes and Gender, Age, Education, Income, and Location.

SUPPLEMENTAL TABLE 1
Interactions between Stated Preferences for Attributes and Gender, Age, Education, Income, and Location.

The Baseline Mobile Service Plan Is the “None” Option, Which Sets the Values of All Attributes to Zero

AttributeAttribute LevelGenderAgeEducationIncomeLocation
Coefficient
(95% Confidence Interval [CI])
Coefficient
(95% CI)
Coefficient
(95% CI)
Coefficient
(95% CI)
Coefficient
(95% CI)
Price   −0.01
(−0.03,
  0.01) 
 −0.28***
(−0.30,
 −0.26) 
 0.16**
(0.14,
 0.19) 
 0.26***
(0.24,
 0.28) 
 −0.04
(−0.06,
 −0.02) 
Internet 2G-High_Limit  −0.51*
(−0.53,
 −0.50) 
 −0.85***
(−0.86,
 −0.83) 
 0.13
(0.12,
 0.14) 
 −0.35
(−0.36,
 −0.34) 
 −0.67**
(−0.69,
 −0.66) 
 2G-Limited  0.07
(0.05,
 0.09) 
 −1.49***
(−1.51,
 −1.47) 
 0.30
(0.28,
 0.32) 
 1.17***
(1.15,
 1.19) 
 −0.47.
(−0.50,
 −0.45) 
 2G-Unlimited  0.39
(0.36,
 0.43) 
 −0.64*
(−0.68,
 −0.61) 
 0.12
(0.09,
 0.16) 
 1.06**
(1.03,
 1.10) 
 −0.38
(−0.42,
 −0.35) 
 3G-OTT  −0.83**
(−0.90,
 −0.76) 
 −1.13***
(−1.19,
 −1.06) 
 0.63.
(0.57,
 0.70) 
 −0.06
(−0.13,
  0.01) 
 −0.62*
(−0.69,
 −0.56) 
 3G-Unlimited  0.41
(0.31,
 0.51) 
 −1.31***
(−1.41,
 −1.22) 
 0.37
(0.27,
 0.47) 
 2.05***
(1.95,
 2.15) 
 −1.24***
(−1.34,
 −1.14) 
 4G-OTT  −0.41.
(−0.44,
 −0.38) 
 −1.13***
(−1.16,
 −1.09) 
 0.92**
(0.89,
 0.96) 
 −0.12
(−0.15,
 −0.09) 
 −0.95***
(−0.98,
 −0.91) 
Calls 300  0.13
(0.09,
 0.16) 
 0.62**
(0.59,
 0.65) 
 −0.31
(−0.35,
 −0.28) 
 −0.53*
(−0.56,
 −0.50) 
 0.11
(0.08,
 0.14) 
 500   0.02
(−0.02,
  0.06) 
 0.29
(0.25,
 0.33) 
 −0.21
(−0.25,
 −0.17) 
 0.12
(0.08,
 0.16) 
 0.51*
(0.47,
 0.56) 
 Unlimited  −0.41.
(−0.47,
 −0.35) 
 1.13***
(1.07,
 1.19) 
  0.01
(−0.06,
  0.07) 
 −0.61.
(−0.68,
 −0.55) 
 0.64**
(0.58,
 0.70) 
Texts 300  −0.46*
(−0.49,
 −0.42) 
 −0.64**
(−0.68,
 −0.60) 
 −0.29
(−0.33,
 −0.26) 
 0.90***
(0.81,
 0.93) 
 −0.04
(−0.08,
 −0.00) 
 500  −0.19
(−0.24,
 −0.15) 
 −0.70**
(−0.74,
 −0.65) 
 −0.19
(−0.23,
 −0.15) 
 0.17
(0.13,
 0.21) 
  0.02
(−0.02,
  0.07) 
 Unlimited  −0.07
(−0.11,
 −0.03) 
 −0.40.
(−0.44,
 −0.36) 
 −0.42
(−0.46,
 −0.38) 
 1.14***
(1.10,
 1.18) 
 −0.07
(−0.11,
 −0.03) 
AttributeAttribute LevelGenderAgeEducationIncomeLocation
Coefficient
(95% Confidence Interval [CI])
Coefficient
(95% CI)
Coefficient
(95% CI)
Coefficient
(95% CI)
Coefficient
(95% CI)
Price   −0.01
(−0.03,
  0.01) 
 −0.28***
(−0.30,
 −0.26) 
 0.16**
(0.14,
 0.19) 
 0.26***
(0.24,
 0.28) 
 −0.04
(−0.06,
 −0.02) 
Internet 2G-High_Limit  −0.51*
(−0.53,
 −0.50) 
 −0.85***
(−0.86,
 −0.83) 
 0.13
(0.12,
 0.14) 
 −0.35
(−0.36,
 −0.34) 
 −0.67**
(−0.69,
 −0.66) 
 2G-Limited  0.07
(0.05,
 0.09) 
 −1.49***
(−1.51,
 −1.47) 
 0.30
(0.28,
 0.32) 
 1.17***
(1.15,
 1.19) 
 −0.47.
(−0.50,
 −0.45) 
 2G-Unlimited  0.39
(0.36,
 0.43) 
 −0.64*
(−0.68,
 −0.61) 
 0.12
(0.09,
 0.16) 
 1.06**
(1.03,
 1.10) 
 −0.38
(−0.42,
 −0.35) 
 3G-OTT  −0.83**
(−0.90,
 −0.76) 
 −1.13***
(−1.19,
 −1.06) 
 0.63.
(0.57,
 0.70) 
 −0.06
(−0.13,
  0.01) 
 −0.62*
(−0.69,
 −0.56) 
 3G-Unlimited  0.41
(0.31,
 0.51) 
 −1.31***
(−1.41,
 −1.22) 
 0.37
(0.27,
 0.47) 
 2.05***
(1.95,
 2.15) 
 −1.24***
(−1.34,
 −1.14) 
 4G-OTT  −0.41.
(−0.44,
 −0.38) 
 −1.13***
(−1.16,
 −1.09) 
 0.92**
(0.89,
 0.96) 
 −0.12
(−0.15,
 −0.09) 
 −0.95***
(−0.98,
 −0.91) 
Calls 300  0.13
(0.09,
 0.16) 
 0.62**
(0.59,
 0.65) 
 −0.31
(−0.35,
 −0.28) 
 −0.53*
(−0.56,
 −0.50) 
 0.11
(0.08,
 0.14) 
 500   0.02
(−0.02,
  0.06) 
 0.29
(0.25,
 0.33) 
 −0.21
(−0.25,
 −0.17) 
 0.12
(0.08,
 0.16) 
 0.51*
(0.47,
 0.56) 
 Unlimited  −0.41.
(−0.47,
 −0.35) 
 1.13***
(1.07,
 1.19) 
  0.01
(−0.06,
  0.07) 
 −0.61.
(−0.68,
 −0.55) 
 0.64**
(0.58,
 0.70) 
Texts 300  −0.46*
(−0.49,
 −0.42) 
 −0.64**
(−0.68,
 −0.60) 
 −0.29
(−0.33,
 −0.26) 
 0.90***
(0.81,
 0.93) 
 −0.04
(−0.08,
 −0.00) 
 500  −0.19
(−0.24,
 −0.15) 
 −0.70**
(−0.74,
 −0.65) 
 −0.19
(−0.23,
 −0.15) 
 0.17
(0.13,
 0.21) 
  0.02
(−0.02,
  0.07) 
 Unlimited  −0.07
(−0.11,
 −0.03) 
 −0.40.
(−0.44,
 −0.36) 
 −0.42
(−0.46,
 −0.38) 
 1.14***
(1.10,
 1.18) 
 −0.07
(−0.11,
 −0.03) 

Note:p < 0.1

*

p < 0.05

**

p < 0.01

***

p < 0.001.

Data are dummy coded conditional logit model coefficients and 95% CI for the interaction terms from models containing main effects of attribute levels and moderators, and their interactions. Coefficients represent the estimated difference in attribute level coefficients from the conjoint experiment between levels of the binary moderator. Males, young adults (below 35 years old), low education, low income, and urban were used as the reference categories. Monthly income: low (below $160 and $160–$320) and high ($321–$630 and above $630).

FOOTNOTES

1.

GSMA, “Mobile Economy. Sub-Saharan Africa,” 7.

2.

International Telecommunication Union, “ITU-D Dataset.”

3.

Nigerian Communications Commission, “Nigerian National Broadband Plan 2020–2025,” 38.

4.

In this paper, VoIP refers to voice calls over the Internet using an OTT application.

5.

Gerpott and Meinert; Cecere and Corrocher, “Usage of VoIP”; Cecere and Corrocher, “Intensity of VoIP”; Gerpott, May, and Nas.

6.

Gerpott, Thomas, and Weichert, 503; Wellmann, 12; Arnold, Schneider, and Hildebrandt.

7.

Gerpott and Meinert, 73; Sawe.

8.

Ogidiaka and Ogwueleka, 438.

9.

Stork, Esselaar, and Chair; Gillwald, 43.

10.

Tayo, Thompson, and Thompson, 4; Aker and Mbiti.

11.

Sawe, 378.

12.

Galpaya.

13.

GSMA, “Mobile Economy. Sub-Saharan Africa,” 7.

14.

Robb and Hawthorne, 2.

15.

Galpaya.

16.

Robb and Hawthorne.

17.

Gillwald.

18.

Robertson.

19.

Robb and Hawthorne, 342.

20.

Cecere and Corrocher, “Usage of VoIP Services,” 574.

21.

Karikoski and Luukkainen, 311.

22.

Gerpott, “Mobile Internet Use Intensity,” 441.

23.

Gerpott, Thomas, and Weichert, 503; Gerpott, “SMS Use Intensity,” 818; Gerpott and Meinert, 70.

24.

Gerpott, May, and Nas, 9.

25.

Wellmann, 12.

26.

Jirakasem and Mitomo, 5.

27.

Arnold, Schneider, and Hildebrandt.

28.

Chavin, Ginwala, and Spear.

29.

Stork, Esselaar, and Chair, 614; Sawe, 380.

30.

Berglind, 2452.

31.

Abu and Tsuji, 28–30; Zhang, 444–46.

32.

Oyelaran-Oyeyinka and Lal, 521, 522.

33.

Birba and Diagne, 466.

34.

Fuchs and Horak, 105.

35.

Venkatesh et al., 467.

36.

Performance Expectancy is the degree to which an individual believes that using the technology will help to improve target outcome (e.g., using a mobile phone to connect with potential customers will lead to more sales).

37.

Effort Expectancy is the degree of ease associated with the use of the technology.

38.

Social Influence is the degree to which an individual perceives that people important to him/her believe he/she should use the new technology.

39.

Facilitating Conditions are the degree to which an individual believes that external constraints exists to support use of the technology (e.g. availability of the required organizational or technical infrastructure).

40.

Venkatesh et al., 467.

41.

Goncalves, Oliveira, and Cruz-Jesus, 283.

42.

Attuquayefio and Addo, 82.

43.

Penard, Poussing, and Yebe, 76–77.

44.

Penard et al., 77; Ani, Chika, and Atseye, 360.

45.

Penard, Poussing, and Yebe, 76; Birba and Diagne, 468.

46.

Penard et al., 72.

47.

Aker and Mbiti, 211.

48.

Tayo, Thompson, and Thompson, 2; Oyelaran-Oyeyinka and Adeya, 78; GSMA, “Consumer Barriers,” 14.

49.

News Bites US - NASDAQ; Carare et al.

50.

Nikou, Bouwman, and Reuver, 30.

51.

Shin, 320.

52.

Chavin, Ginwala, and Spear, 1–2.

53.

Sawe, 379.

54.

Ogidiaka and Ogwueleka.

55.

Ogbo et al.

56.

Gerpott, Thomas, and Weichert, 492.

57.

International Telecommunication Union, “Fixed_broadband”; International Telecommunication Union, “Fixed Telephone Statistics.”

58.

International Telecommunication Union, “Mobile_cellular.”

59.

GSMA, “The Mobile Economy.”

60.

GSMA, “The Mobile Economy. GSMA Intelligence”; GSMA, “The Mobile Economy: Sub-Saharan Africa 2017.”

61.

National Bureau of Statistics.

62.

Ibid.

63.

Nigerian Communications Commission, “Industry Statistics.”

64.

GSMA, “Digital Inclusion,” 12.

65.

2G offers up to 0.3 Mbps and is based on GSM, 3G offers between 0.1 Mbps to 42 Mbps, whereas 4G offers between 12 Mbps to 450 Mbps and is based on LTE technology.

66.

GSMA, “The State of Mobile Internet Connectivity,” 39.

67.

Nigerian Communications Commission, “Nigerian National Broadband Plan 2020 - 2025,” 38.

68.

GSMA, “Mobile Economy,” 16.

69.

McFadden, 345.

70.

Hensher, Louviere, and Swait, 18; Toubia, Hauser, and Simester, 116.

71.

Nikou, Bouwman, and Reuver, 30.

72.

Pu and Grossklags, 9.

73.

Ueda.

74.

Sawtooth Software is the most widely used program to design, collect and analyze choice-based survey data (www.sawtoothsoftware.com).

75.

Advantage of the none option is that it makes the choice decision more realistic and therefore leads to better predictions of preferences. A disadvantage is that participants may select none to avoid difficult choices and this affects the validity of preference predictions. Another disadvantage is that the respondent may consider “none” to mean that they can choose some other option in the market, that is, an option offered by MTN and if they prefer the MTN option they might select “none.” In our experiment, the none option was selected in 8% of the choice tasks, and at least once by 55% of the respondents.

76.

The price was displayed to the participants in Nigerian Naira. The Nigerian official exchange rate of 316 Naira/$ is used. However, it is significantly different from the volatile black-market rate which reached a maximum of ~480 Naira/$ between January to November 2016.

77.

Analyzing the data from the attention checks show that 66% of all participants passed all the attention checks, and 99% passed the attention check in the conjoint choice tasks. The results of the group that passed the attention checks were, on average, similar to those of the entire sample.

78.

Birba and Diagne, 468; Penard et al., 77; Penard, Poussing, and Yebe, 76; Ani, Chika, and Atseye, 361.

79.

A mega city and also the economic hub of the country with a population density of between 6,400 (based on the 2006 population census) to 15,000 people/square mile (based on 2016 population estimate by the National Population Commission of Nigeria).

80.

Recruitment took place in Warri, a medium sized metropolitan city in the underdeveloped south of Nigeria with an estimated population density of 200 people/square mile based on the 2006 population census, and also in a cluster of remote villages in southern Nigeria with an estimated population density of 60 people/square mile.

81.

National Bureau of Statistics.

82.

National Population Commission, “Demographic and Health Survey.”

83.

Haaijer, Kamakura, and Wedel, 96.

84.

Following the technique derived in the literature, see Mcfadden, 73; Haaijer, Kamakura, and Wedel, 98. Conditional choice probability is the probability that a particular service is chosen given certain alternatives.

85.

Rasche et al.; Wilson et al.; Hansen et al.

86.

Taale and Kyeremeh.

87.

Meenakshi et al.

88.

Rasche et al.; Carare et al.

89.

Rasche et al.; Carare et al., 22; Taale and Kyeremeh, 282.

90.

Wilson et al.; Hansen et al., 2.

91.

Meenakshi et al., 65; Daly, Hess, and Train, 22; Armstrong, Garrido, and Ortuzar, 144.

92.

Gerpott, “Mobile Internet Use,” 440.

93.

Robb and Hawthorne.

94.

Gerpott, Thomas, and Weichert, 502.

95.

Tayo, Thompson, and Thompson, 2; Oyelaran-Oyeyinka and Adeya, 78.

96.

Futter and Gillwald, 6.

97.

Stork, Esselaar, and Chair, 614.

98.

Internet Service Providers' Association.

99.

Orme and Howell, 2.

BIBLIOGRAPHY

Abu, Sheikh Taher, and Masatsugu Tsuji.
“The Determinants of the Global Mobile Telephone Deployment: An Empirical Analysis.”
Informatica Economica
14
, no.
3
(
2010
):
21
33
. Accessed August 20, 2016, http://revistaie.ase.ro/content/55/1002%20-%20Sheikh%20Taher%20Abu.pdf.
Aker, Jenny C., and Isaac M. Mbiti.
“Mobile Phones and Economic Development in Africa.”
Journal of Economic Perspectives
24
, no.
3
(
2010
):
207
32
. doi:.
Ani, Okon Edet, Uchendu Chika, and Emmanuel U. Atseye.
“Bridging the Digital Divide in Nigeria: A Study of Internet Use in Calabar Metropolis, Nigeria.”
Library Management
28
, no.
6/7
(
2007
):
355
65
. doi:.
Armstrong, Paula, Rodrigo Garrido, and Juan De Dios Ortúzar.
“Confidence Intervals to Bound the Value of Time.”
Transportation Research Part E: Logistics and Transportation Review
37
, no.
2–3
(
2001
):
143
61
. doi:.
Arnold, Rene C. G., Anna Schneider, and Christian Hildebrandt.
“All Communications Services Are Not Created Equal – Substitution of OTT Communications Services for ECS from a Consumer Perspective.”
TPRC44
(
2016
)
1
22
. Accessed February 5, 2017, http://ssrn.com/abstract=2756395.
Attuquayefio, Samuel Niiboi, and Hillar Addo.
“Using the UTAUT Model to Analyze Students’ ICT Adoption.”
International Journal of Education and Development Using Information and Communication Technology (IJEDICT)
10
, no.
3
(
2014
):
75
86
. Accessed February 25, 2019, https://files.eric.ed.gov/fulltext/EJ1059042.pdf.
Berglind, Oscar Saenz de Miera.
“The Effect of Zero-Rating on Mobile Broadband Demand: An Empirical Approach and Potential Implications.”
International Journal of Communication
10
(
2016
):
2442
59
. Accessed February 20, 2021, https://ijoc.org/index.php/ijoc/article/viewFile/4651/1659.
Birba, Ousmane, and Abdoulaye Diagne.
“Determinants of Adoption of Internet in Africa: Case of 17 Sub-Saharan Countries.”
Structural Change and Economic Dynamics
23
(
2012
):
463
72
. doi:.
Carare, Octavian, Chris McGovern, Raquel Noriega, and Jay Schwarz.
“The Willingness to Pay for Broadband of Non-Adopters in the U.S.: Estimates from a Multi-State Survey.”
Information Economics and Policy
30
(
2015
):
19
35
. doi:.
Cecere, Grazia, and Nicoletta Corrocher.
“The Intensity of VoIP Usage in Great Britain: Users’ Characteristics and Firms’ Strategies.”
Telecommunications Policy
35
(
2011
):
522
31
. doi:.
Cecere, Grazia, and Nicoletta Corrocher.
“The Usage of VoIP Services and Other Communication Services: An Empirical Analysis of Italian Consumers.”
Technological Forecasting & Social Change
79
(
2012
):
570
78
. doi:.
Chavin, James, Aadil Ginwala, and Max Spear.
“The Future of Mobile Messaging: Over-the-Top Competitors Threaten SMS,”
2012
. Accessed February 14, 2021, http://telecomsmckinsey.com.
Daly, Andrew, Stephane Hess, and Kenneth Train.
“Assuring Finite Moments for Willingness to Pay in Random Coefficient Models.”
Transportation
39
, no.
1
(
2012
):
19
31
. doi:.
Fuchs, Christian, and Eva Horak.
“Africa and the Digital Divide.”
Telematics and Informatics
25
(
2008
):
99
116
. doi:.
Futter, Ariel, and Alison Gillwald.
“Zero-Rated Internet Services: What Is to Be Done?”
Broadband 4 Africa
,
2015
,
1
10
. Accessed November 27, 2020, https://www.researchictafrica.net/docs/Facebook%20zerorating%20Final_Web.pdf.
Galpaya, Helani.
“Zero-Rating in Emerging Economies.”
2017
. Accessed February 13, 2021, https://www.cigionline.org/static/documents/documents/GCIG%20no.47_2.pdf.
Gerpott, Torsten J.
“‘SMS Use Intensity Changes in the Age of Ubiquitous Mobile Internet Access’ A Two-Level Investigation of Residential Mobile Communications Customers in Germany.”
Telematics and Informatics
32
(
2015
):
809
22
. doi:.
Gerpott, Torsten J.
“Impacts of Mobile Internet Use Intensity on the Demand for SMS and Voice Services of Mobile Network Operators: An Empirical Multi-Method Study of German Mobile Internet Customers.”
Telecommunications Policy
34
, no.
8
(
2010
):
430
43
.
Gerpott, Torsten J., Sebastian May, and Gokhan Nas.
“The Impact of Mobile Internet on Mobile Voice Usage: A Two-Level Analysis of Mobile Communications Customers in a GCC Country.”
Information & Management
54
(
2017
):
958
970
. doi:.
Gerpott, Torsten J., and Phil Meinert.
“The Impact of Mobile Internet Usage on Mobile Voice Calling Behavior: A Two-Level Analysis of Residential Mobile Communications Customers in Germany.”
Telecommunications Policy
40
(
2016
):
62
76
. doi:.
Gerpott, Torsten J., Sandra Thomas, and Michael Weichert.
“Usage of Established and Novel Mobile Communication Services: Substitutional, Independent or Complementary? An Empirical Study of Residential Mobile Network Operator Customers in Germany.”
Information Systems Frontiers
16
, no.
3
(
2014
):
491
507
. doi:.
Gillwald, Alison.
“Beyond Access: Addressing Digital Inequality in Africa.”
Pop Brands
,
2017
. doi:. Accessed February 20, 2020, https://media.africaportal.org/documents/GCIG_no.48_0.pdf
Goncalves, Gloria, Tiago Oliveira, and Frederico Cruz-Jesus.
“Understanding Individual-Level Digital Divide: Evidence of an African Country.”
Computers in Human Behavior
87
(
October
2018
):
276
91
. doi:.
GSM Association
.
“Digital Inclusion and the Role of Mobile in Nigeria.” GSMA Intelligence
,
2015
. Accessed September 5, 2016, https://www.gsma.com/publicpolicy/wp-content/uploads/ 2016/09/GSMA2015_Report_DigitalInclusionAndTheRoleOfMobileInNigeria.pdf.
GSM Association
.
“The Mobile Economy.”
2016
. Accessed August 23, 2016, https://www.gsmaintelligence.com/research/?file=97928efe09cdba2864cdcf1ad1a2f58c&download.
GSM Association
.
“Consumer Barriers to Mobile Internet Adoption in Africa.”
GSMA Intelligence
,
2016
. Accessed August 9, 2017, https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2016/07/Consumer-Barriers-to-mobile-internet-adoption-in-Africa.pdf.
GSM Association
.
“The Mobile Economy: Sub-Saharan Africa 2017.”
GSMA Intelligence
,
2017
. Accessed August 7, 2017, https://www.gsmaintelligence.com/research/?file=7bf3592e6d 750144e58d9dcfac6adfab&download.
GSM Association
.
“The Mobile Economy.”
GSMA Intelligence
,
2018
. Accessed September 9, 2018, https://data.gsmaintelligence.com/api-web/v2/research-file-download?id=28999769&file=The%20Mobile%20Economy%202018.pdf.
GSM Association
.
“The Mobile Economy. Sub-Saharan Africa.”
GSMA Intelligence
,
2019
. Accessed September 2, 2019, https://www.gsmaintelligence.com/research/2019/02/the-mobile- economy-2019/731/.
GSM Association
.
“The State of Mobile Internet Connectivity.”
2020
. Accessed February 14, 2021, https://www.gsma.com/r/wp-content/uploads/2020/09/GSMA-State-of-Mobile-Internet-Connectivity-Report-2020.pdf.
Haaijer, Rinus, Wagner Kamakura, and Michel Wedel.
“The ‘no-Choice’ Alternative in Conjoint Choice Experiments.”
International Journal of Market Research
43
, no.
1
(
2001
):
93
106
. doi:.
Hansen, Kristian Schultz, Debora Pedrazzoli, Anthony Mbonye, Sian Clarke, Bonnie Cundill, Pascal Magnussen, and Shunmay Yeung.
“Willingness-to-Pay for a Rapid Malaria Diagnostic Test and Artemisinin-Based Combination Therapy from Private Drug Shops in Mukono District, Uganda.”
Health Policy and Planning
28
(
2013
):
185
96
. doi:.
Hensher, David, Jordan Louviere, and Joffre Swait.
Stated Choice Methods: Analysis and Applications
.
Cambridge, UK
:
Cambridge University Press
,
2000
. doi:.
International Telecommunication Union
.
“Fixed_broadband_2000-2019.”
ITU-D Dataset
,
2020
.
International Telecommunication Union
.
“Fixed Telephone Statistics 2000–2019.”
ITU-D Dataset
,
2020
.
International Telecommunication Union
.
“ITU_Key_2005-2019_ICT_data.”
ITU-D Dataset
,
2020
.
International Telecommunication Union
.
“Mobile_cellular_2000-2018.”
ITU-D Dataset
,
2020
.
Internet Service Providers’ Association
.
“‘Net Neutrality’ A Non-Issue in South Africa for the Present.”
2014
. Accessed November 27, 2020, https://ispa.org.za/press_releases/net-neutrality-a-non-issue-in-south-africa-for-the-present-says-ispa/.
Jirakasem, Monarat, and Hitoshi Mitomo.
“Substitution between Mobile Telephony and OTT Communication Services: The Case of Thailand.”
Paper presented at The 22nd Biennial Conference of the International Telecommunications Society: Beyond the Boundaries: Challenges for Business, Policy and Society
,
Seoul, Korea
,
2018
. Accessed January 20, 2019, http://hdl.handle.net/10419/190338.
Karikoski, Juuso, and Sakari Luukkainen.
“Substitution in Smartphone Communication Services.”
Paper presented at 2011 15th International Conference on Intelligence in Next Generation Networks, IEEE
,
Berlin, Germany
,
2011
,
313
18
. doi:.
Mcfadden, Daniel.
“Modeling the Choice of Residential Location.”
Transportation Research Record
673
, no.
2
(
1978
):
72
77
. Accessed December 21, 2016, http://onlinepubs.trb.org/Onlinepubs/trr/1978/673/673-012.pdf.
McFadden, Daniel L.
“Economic Choices.”
The American Economic Review
91
, no.
3
(
2001
):
351
78
. doi:.
Meenakshi, J. V., A. Banerji, Victor Manyong, Keith Tomlins, Nitya Mittal, and Priscilla Hamukwala.
“Using a Discrete Choice Experiment to Elicit the Demand for a Nutritious Food: Willingness-to-Pay for Orange Maize in Rural Zambia.”
Journal of Health Economics
31
(
2012
):
62
71
. doi:.
National Bureau of Statistics
.
“Nigerian Telecommunications Sector Q1 2016 (Summary Report): State Disaggregated Data.”
2016
. Accessed December 8, 2016, http://www.nigerianstat.gov.ng/download/396.
National Population Commission
.
“Censuses.”
2006
. Accessed January 2, 2017, http://www.population.gov.ng/index.php/censuses.
National Population Commission
.
“Demographic and Health Survey.”
2018
. Accessed May 20, 2020, https://dhsprogram.com/pubs/pdf/FR359/FR359.pdf.
News Bites US - NASDAQ
.
“Comcast Extends Internet Essentials, Its High-Speed Internet Adoption Program, To Low-Income Senior Citizens in San Francisco.”
2015
. Accessed March 26, 2020, http://search.proquest.com/docview/1705294712/.
Nigerian Communications Commission
.
“Industry Statistics.”
NCC Website
,
2021
. Accessed February 14, 2021, https://www.ncc.gov.ng/stakeholder/statistics-reports/industry-overview#view-graphs-tables.
Nigerian Communications Commission
.
“Nigerian National Broadband Plan 2020 - 2025.”
2019
. Accessed February 15, 2021, https://www.ncc.gov.ng/documents/880-nigerian-national-broadband-plan-2020-2025/file.
Nikou, Shahrokh, Harry Bouwman, and Mark De Reuver.
“The Potential of Converged Mobile Telecommunication Services: A Conjoint Analysis.”
Info
14
, no.
5
(
2012
):
21
35
. doi:.
Nothias, Toussaint.
“Access Granted: Facebook’s Free Basics in Africa.”
Media, Culture and Society
42
, no.
3
(
2020
):
329
48
. doi:.
Ogbo, Erezi, Tim Brown, Jon Gant, and Douglas Sicker.
“When Being Connected is not Enough: An Analysis of the Second and Third Levels of the Digital Divide in a Developing Country.”
Journal of Information Policy
(
Forthcoming
).
Ogidiaka, Emoghene, and Francisca Nonyelum Ogwueleka.
“Over-The-Top Services (OTT) on Telecommunication Operators in Nigeria: Exploring Consumers’ Behaviour.”
International Journal of Information Technology
12
, no.
2
(
June
10
,
2020
):
437
46
. doi:.
Orme, Bryan K., and John Howell.
“Application of Covariates within Sawtooth Software’s CBC/HB Program: Theory and Practical Example.”
Sawtooth Software Research Paper Series
98382
, no.
360
(
2009
):
1
20
. Accessed June 8, 2019, https://sawtoothsoftware.com/resources/technical-papers/application-of-covariates-within-sawtooth-softwares-cbc-hb-program-theory-and-practical-example.
Oyelaran-Oyeyinka, Banji, and Catherine Nyaki Adeya.
“Internet Access in Africa: Empirical Evidence from Kenya and Nigeria.”
Telematics and Informatics
21
(
2004
):
67
81
. doi:.
Oyelaran-Oyeyinka, Banji, and Kaushalesh Lal.
“Internet Diffusion in Sub-Saharan Africa: A Cross-Country Analysis.”
Telecommunications Policy
29
(
2005
):
507
27
. doi:.
Penard, Thierry, Nicolas Poussing, Blaise Mukoko, Georges Bertrand, and Tamokwe Piaptie.
“Internet Adoption and Usage Patterns in Africa: Evidence from Cameroon.”
Technology in Society
42
(
2015
):
71
80
. doi:.
Penard, Thierry, Nicolas Poussing, and Gabriel Zomo Yebe.
“Comparing the Determinants of Internet and Cell Phone Use in Africa: Evidence from Gabon.”
Communications and Strategies
1
, no.
86
(
2012
):
65
83
. Accessed July 2, 2017, https://ssrn.com/abstract=2270509.
Pu, Yu, and Jens Grossklags.
“Using Conjoint Analysis to Investigate the Value of Interdependent Privacy in Social App Adoption Scenarios.”
Paper presented at International Conference on Information Systems: Exploring the Information Frontier (ICIS 2015)
,
Fort Worth, Texas
,
2015
. Accessed September 4, 2016, https://aisel.aisnet.org/icis2015/proceedings/SecurityIS/12/.
Rasche, Peter, Alexander Mertens, Christopher Brandl, Shan Liu, David Weber, Philipp Lichte, Priv Doz, Matthias Knobe, and Priv Doz.
“Satisfying Product Features of a Fall Prevention Smartphone App and Potential Users’ Willingness to Pay: Web-Based Survey Among Older Adults.”
JMIR Mhealth Uhealth
6
, no.
3
(
2018
):
e75
. doi:.
Robb, Genna, and Ryan Hawthorne.
“Net Neutrality and Market Power: The Case of South Africa.”
Telecommunications Policy
43
, no.
9
(
2019
):
101814
. doi:.
Robertson, Adi.
“India Just Approved New Net Neutrality Rules That Ban ‘any Form’ of Data Discriminations.”
2018
. Accessed February 14, 2021, https://www.theverge.com/2018/7/11/17562108/india-department-of-telecommunications-trai-net-neutrality-proposal-approval.
Sawe, Timothy Kipkorir.
“Emergence of OTT Communication Services and Sustenance of Revenue among Kenya Telcos.”
International Journal of Innovative Science, Engineering & Technology
3
, no.
8
(
2016
):
377
81
. Accessed February 23, 2019, http://ijiset.com/vol3/v3s8/IJISET_V3_I8_52.pdf.
Shin, Dong Hee.
“What Makes Consumers Use VoIP over Mobile Phones? Free Riding or Consumerization of New Service.”
Telecommunications Policy
36
, no.
4
(
2012
):
311
23
. doi:.
Stork, Christoph, Steve Esselaar, and Chenai Chair.
“OTT - Threat or Opportunity for African Telcos?”
Telecommunications Policy
41
(
2017
):
600
16
. doi:.
Taale, Francis, and Christian Kyeremeh.
“Households Willingness to Pay for Reliable Electricity Services in Ghana.”
Renewable and Sustainable Energy Reviews
62
(
2016
):
280
88
. doi:.
Tayo, Omolara, Randall Thompson, and Elizabeth Thompson.
“Impact of the Digital Divide on Computer Use and Internet Access on the Poor in Nigeria.”
Journal of Education and Learning
5
, no.
1
(
2015
):
1
6
. doi:.
Toubia, Olivier, John R. Hauser, and Duncan I. Simester.
“Polyhedral Methods for Adaptive Choice-Based Conjoint Analysis.”
Journal of Marketing Research
41
, no.
1
(
2004
):
116
31
. doi:.
Ueda, Masashi.
“An Analysis on Smartphone Adoption for University Students by Using Choice-Based Conjoint Analysis.”
Paper presented at Proceedings of the the 3rd Multidisciplinary International Social Networks Conference on Social Informatics 2016, Data Science 2016, Union
,
New Jersey
,
2016
. doi:.
Venkatesh, Viswanath, Michael G. Morris, Gordon B. Davis, and Fred D. Davis.
“User Acceptance of Information Technology: Toward a Unified View.”
MIS Quarterly
27
(
2003
):
425
78
. Accessed February 23, 2019, https://www-jstor-org.proxy.library.cmu.edu/stable/pdf/30036540.pdf?refreqid=excelsior%3A9f550ffa6b23c366317234629b2b9a1e.
Wellmann, Nicolas.
“Are OTT Messaging and Mobile Telecommunication an Interrelated Market? An Empirical Analysis.”
Telecommunications Policy
43
, no.
101831
(
2019
):
1
17
. doi:.
Wilson, Leslie S., Traci M. Blonquist, Fangxin Hong, Barbara Halpenny, Seth Wolpin, Peter Chang, Christopher P. Filson, Viraj A. Master, Martin G. Sanda, Gary W. Chien, Randy A. Jones, Tracey L. Krupski, and Donna L. Berry.
“Assigning Value to Preparation for Prostate Cancer Decision Making: A Willingness to Pay Analysis.”
BMC Medical Informatics and Decision Making
4
(
2019
):
1
9
. doi:.
Zhang, Xiaoqun.
“Exploring the Patterns and Determinants of the Global Mobile Divide.”
Telematics and Informatics
34
, no.
1
(
2017
):
438
49
. doi:.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.