Margaret Pearce






The invention of the Internet has had a tremendous impact on society. It has changed nearly every aspect of modern life, including the way individuals communicate, do business, access information, and make decisions. Having the ability to go online can mean access to an education via online classes and distance learning programs, increased access to breaking world and local news, and online or work-from-home job opportunities. Developing countries building their citizens’ access to the Internet boosts their chances of achieving sustainable economic growth, according to a senior United Nations official [7]. On an individual level, a recent study has suggested that people with limited access to the Internet consider it as important as medicine and are more likely to agree that Internet access is necessary to survival [8].

In this project, Internet usage, affordability, and quality are explored for countries in Southeast Asia. Countries in this region of the world are diverse in terms of economics, culture, geography, and history. Southeast Asia has a population of over 600 million people and is the 8th largest economy in the world, collectively [10]. Internet penetration in the region is lower than the global average average, but is experiencing a rapid growth of Internet and social media usage [11]. According to one study, the number of Facebook and Twitter users for Southeast Asia (collectively) will surpass the number in the United States in 2016 [12]. Furthermore, the region is considered by some to be a hotspot for new mobile business models [13].

On the other hand, many countries in Southeast Asia are considered low income: the median gross national income per capita in 2014 was $3776. In 2012, only 31.1% of Cambodians had access to electricity. During the same year, 25.2% of the population in the Philippines lived below the national poverty line [14]. This information paints two different pictures of how the region is faring economically and what role technology plays in the everyday lives of citizens. In this project, we seek answers to questions clarifying these points, such as: what percentage of the country uses the internet? How affordable are internet services? What is the quality of internet options like, and are there speed limitations or data caps?

For the purposes of this report, the following countries are considered to be in the Southeast Asia region: Brunei Darussalam, Cambodia, Indonesia, Lao PDR, Malaysia, Philippines, Singapore, Thailand, Timor-Leste, and Vietnam.

Internet usage

We begin by exploring how many people have access to the Internet in each country in Southeast Asia, and we compare it to the number of mobile cellular subscribers. Internet users in this context are defined as individuals who have used the Internet in the last year from any location. Users could access the Internet through a variety of devices, such as a computer, a mobile phone, a PDA device, mobile broadband card, and so forth.


For many countries, we observe the number of Internet users (per 100 people) increasing rapidly since the early 2000s and onward. Singapore, Brunei, and Malaysia show particularly high growth in the number of Internet users, surpassing the next highest country, Vietnam, by 20% (20 people per 100 people). However, countries with fewer users show slower growth over time. Myanmar, Cambodia, and Lao PDR remain below 15 Internet users per 100 people as of 2014.

By contrast, we observe less of a divide among countries when comparing the number mobile cellular subscriptions per 100 people. Mobile cellular subscriptions are defined as a subscription to a mobile phone service that provide access to public switched telephone network (PSTN) via cellular technology. This includes postpaid subscriptions and active prepaid subscriptions, where active is defined as used within the last three months. Subscriptions via data cards, USB modems (VOIP), and subscriptions to public mobile data services are not included in this measure.

Mapping the number of mobile cellular subscriptions (per 100 people), we can identify several countries that are lagging behind the rest of the region on the number of Internet users but surging ahead on the number of mobile subscribers. According to The World Bank:


“The International Telecommunication Union (ITU) estimates that there were about 6 billion mobile subscriptions globally in the early 2010s. No technology has ever spread faster around the world. Mobile communications have a particularly important impact in rural areas. The mobility, ease of use, flexible deployment, and relatively low and declining rollout costs of wireless technologies enable them to reach rural populations with low levels of income and literacy. The next billion mobile subscribers will consist mainly of the rural poor…. If the service is not affordable to most people, goals of universal usage will not be met.”

Interestingly, for many countries, the number of mobile cellular subscriptions per 100 people exceeds 100. In other words, some individuals have more than one mobile cellular subscription throughout the year. Many cell phone users have prepaid or “pay-as-you-go” plans rather than phones leased on contracts and paid for at the end of each months, as is common in the United States [15].

It is important to note that the measure for Internet users does include individuals who access the internet from a mobile device. Thus, the difference in the number of Internet users and the number of mobile cellular subscriptions suggests that many people either can’t or don’t access the Internet from their mobile devices. Looking at 2014 (the newest data available), we see stark differences in Internet usage versus mobile cellular subscriptions: for instance, Timor-Leste has nearly 53.4 times the number of people with mobile cellular subscriptions than people using the Internet.

The gap between the percentage of Internet users and the percentage of Mobile Subscribers hints at the strong potential for increased Internet access. A report by Ericsson notes that smartphone adoption is predicted to increase dramatically between 2013 and 2019 with smartphone subscriptions growing nearly fivefold [16]. Smartphones significantly lower the barrier to entry for convenient access to the Internet compared to the early 2000’s where users would need a desktop computer, monitor, mouse, and keyboard, or a laptop computer. Increased smartphone adoption means more people will have mobile devices capable of connecting to the Internet, particularly on faster 4G/ LTE networks. However, as noted by the World Bank, services must be affordable in order for achieve broader access and usage.

Affordability

Next, we investigate the affordability of broadband and mobile cellular service in each country. Intuitively, costs can be a limiting factor in technology adoption. To put the costs in context, we must first explore the economic status of the region. Countries in Southeast Asia range greatly in terms of per capita income. Using data from 2014, the country with the lowest Gross National Income (GNI) per capita was Cambodia ($949 USD per person). Compared to the country with the highest GNI per capita, Singapore ($53,986 USD per person), GNI per capita in Cambodia was over 98% lower. The average GNI per capita in the region was $11,767.20 USD (median: $3776 USD). Countries are classified as low income, lower-middle income, upper-middle income, or high income using cutoffs established by The World Bank [9].


It is important to distinguish between postpaid and prepaid mobile broadband services for pricing and data limits. According to ITU, prepaid prices are included in the data because they are often the only payment method available to low-income users, who may not have a regular income and thus won’t quality for a postpaid subscription. Broadband prices are further separated into three core groups:

  1. Fixed broadband (e.g. home internet)
  2. Mobile broadband, computer-based (e.g. a mobile hotspot card)
  3. Mobile broadband, handset-based (e.g. a smart phone)

There are rules that qualify or disqualify individual plans from being included in the country averages. More details can be found in the ITU’S Measuring the Information Society Report 2015, Annex 3.

As hinted by The World Bank, usage is linked to the affordability of broadband services. In fact, we can see a strong link between the cost of each service for the prepaid and postpaid plans and the number of Internet users per 100 people (p < 0.0001). Looking only at mobile broadband, we can see similar trends between the cost of mobile broadband services (prepaid and postpaid) and the number of mobile cellular subscriptions per 100 people (p = 0.044). The pattern is unsurprising: the cheaper that broadband and mobile services are, there tend to be more users/ subscribers regardless of whether the service is paid for in advance or not.


On the graph of “Internet Users vs. Service Cost”, we can identify trends based on the service type. For several countries, including Timor-Leste, Cambodia, and Lao PDR, fixed broadband (purple) is costlier than mobile broadband options. How significant are these differences, and what does it mean in terms of the amount of access that users might have? To answer this question, first we look at relative prices for each service, and then we consider how “usable” each service is in terms of its data limits per month.

Country 1. Fixed broadband 2. Mobile broadband, computer based (postpaid) 2. Mobile broadband, computer based (prepaid) 3. Mobile broadband, handset based (postpaid) 3. Mobile broadband, handset based (prepaid)
Brunei Darussalam 1.87 0.80 0.57 0.80 0.57
Cambodia 12.64 3.15 3.16 2.53 2.53
Indonesia 3.11 1.56 2.12 1.56 1.13
Lao PDR 11.84 5.15 5.15 4.12 4.12
Malaysia 3.10 1.69 1.69 0.99 0.99
Philippines 8.27 8.27 4.13 2.47 2.47
Singapore 0.70 0.35 0.49 0.35 0.26
Thailand 3.63 2.49 2.76 1.38 1.38
Timor-Leste 14.79 3.77 4.53 3.02 3.02
Vietnam 2.00 3.92 3.92 1.96 7.31

Table: Prices for broadband as % of GNI p.c. Highest priced options are in orange, lowest priced options are in green.

Costs for each type of service and each payment method vary from one country to the next. In general, handset based mobile broadband is the cheapest option in terms of the % of GNI per capita. Fixed broadband services tend to be the highest priced option. However, data caps and speeds for each plan type are not equal: as illustrated in the visualization below, not all options give the same “bang for your buck”. Although mobile broadband services may be cheaper, this comes at the cost of being able to do less on the Internet over the course of a month.


Fixed broadband often advertise “unlimited” data, which means there is no cap on the amount of data usage in a month. Thus, fixed broadband plans can be the value best in terms of data for the money spent, particularly when the cost is not significantly higher than mobile options. Among mobile options, handset-based mobile broadband services tend to have higher data caps, often at lower prices compared to computer-based options.

Quality

What we have shown from the data so far is that the number of mobile cellular subscriptions far exceeds the number of Internet users. Among Internet service options, mobile broadband tends to be a less expensive option compared to fixed broadband. Aside from data caps, there is a big caveat to relying on a mobile phone for Internet connectivity: network quality. After all, even the most affordable Internet-capable mobile phone is of little use if there is no cellular service, if the service is spotty, or if speeds are unusably slow. In this final section, we explore the quality of mobile broadband services in terms of their download and upload speeds, reliability, and signal strength, and we comment on the usability implications.

For this section, we evaluate crowd sourced network statistics from each country using OpenSignal’s NetworkStats API. OpenSignal gets data from users opting into their app, which constantly monitors the coverage and performance of mobile networks on physical devices around the world. The resulting performance metrics are sent to OpenSignal, who then generates maps and aggregate coverage statistics for a given geographical area. For each country, we focus on network statistics in the capital city district. See notes in the Methodology section for more details on this approach.


First, we look at the variety of network providers captured from each capital city district. While the total number of network providers varies from one capital city to the next, within a capital city, networks also vary by type: 2G, 3G, or 4G, where “G” stands for generation. The 2G network type is the slowest (imagine using the Internet on a cell phone before smart phones) whereas 4G is the newest and fastest type. More 4G networks means that there is an increased likelihood of having faster service within the capital city. Furthermore, we can compare download and upload speed ranges by network type to highlight speed improvements for different network types.



Looking at the graphs, we can see sharp differences between network types in terms of speed. 2G networks are mere fractions of the download and upload speeds for 4G and even 3G networks. Significant speed improvements can be seen for 4G networks. For comparison purposes, note that Charter Spectrum Internet services in the Madison, Wisconsin area advertise a download speed of 60 Mbps and an upload speed of 4 Mbps.

Download speeds in particular can have stark consequences for everyday Internet activity. To illustrate this point, consider the time it would take to perform common Internet tasks using the average speed among network providers on each network type in each country. Watching an hour-long HD video on Netflix can range from taking a matter of minutes to stream all of the data (4G) to hours upon hours (2G). Furthermore, activities that take too long may time out and disconnect before they are completed, in which case users may need to start the activity over from the beginning in the worst case.


However, speed is only one aspect of network quality. A network with fantastic speeds that drops all the time won’t let the user actually realize the top speeds. Therefore, we must also consider how reliable networks are as well as the strength of network signals.



Overall, though 4G networks tend to be more reliable, their signal strength is lower than 2G and 3G options. This means that 4G networks are high quality, but they do not have as much coverage as older network types. This is not surprising given that 4G is the newest network type of the three, and therefore it may not be available as broadly yet. Overall, 3G tends to have the best service coverage. Finally, 2G tends to rank last in terms of reliability, but because of its status as an older generation network technology, it can outperform coverage of 4G.

Conclusion

Southeast Asia is a region with vast differences from one country to another in terms of income, Internet usage and costs, and mobile cellular service quality. Overall, we see that it is far more likely that someone in this region has a mobile phone than they access the Internet. However, as smartphones become cheaper as they are predicted to in this region, more mobile cellular subscribers may have the capability to access the Internet. Indeed, the barrier to entry for Internet access is lower in terms of mobile broadband costs compared to fixed (home) broadband. That is not to say that mobile broadband is an equal substitute for fixed broadband, however: monthly limits on data, varying download/ upload speeds, and low signal strengths for faster networks such as 4G can severely constrain practical usage of mobile broadband networks. Network quality will need to continue improving in order for the rapid growth of Internet and smartphone users to have the biggest impact.

Methodology

Data Sources

Data used in this project came from several distinct sources:

  1. The World Bank
    The World Bank is an organization that offers support to developing countries through policy advice, research, and technical assistance. As part of this effort, The World Bank provides free access to comprehensive data sets about development and indicators on countries worldwide [1]. Data is available by country, by development indicators, by topic, or a la carte.
  2. ITU’S Measuring the Information Society Report 2015
    ITU is a specialized agency within the United Nations for information and communication technologies (ICTs). One of the three sectors that ITU operates is the Telecommunication Development Sector, which publishes an annual report measuring the information society. The report includes data for 167 countries worldwide with data on pricing of mobile and broadband technologies, service levels, insights on overall development in each country pertaining to telecommunications, and comments on trends such as the Internet of Things. The appendix of the report details the methodology used to collect the data as well as criteria for inclusion/ exclusion for certain metrics.

    Note: Several metrics used in The World Bank’s development indicators are sourced from ITU, including the number of Internet users (per 100 people), and The World Bank advises individuals utilizing these data to cite ITU. The World Bank data is used in this project for two reasons: 1) The World Bank provides data in a readily-consumable format (.csv, .xslx) rather than needing to extract it from a PDF report, and 2) The World Bank includes additional metrics not covered in ITU’s report, such as information on access to electricity.

  3. OpenSignal
    OpenSignal publishes data on coverage and performance of mobile networks worldwide [3]. The data is obtained in a crowdsourced fashion using the OpenSignal app, which monitors the coverage and performance of mobile networks on physical devices around the world. In exchange for providing signal statistics to OpenSignal, app users are given a toolkit to help them locate better connections in real time along with Wifi hotspots. The organization provides a free developer API [4], including a NetworkStats API that provides aggregate signal information by network name and type for a given area.
    One advantage of this source is that the data is obtained directly from users, increasing its integrity. However, the API does have several limitations: first, for non-commercial/ non-subscription users, only 5 requests are allowed per minute maxing out at 2000 requests per month. Second, each request must include a latitude and longitude center point and a bounding box distance (km). The API will return the average signal information in the area enclosed by the bounding box based on the center point. The bounding box size is limited to 10km x 10km. Thus, combined with the restricted number of calls, only a certain amount of data can be reasonably obtained from the OpenSignal API without a paid plan, which starts at $10,000 for academic purposes. To address this issue, the project only considers capital cities for each country and uses city boundaries to determine how many API calls to make and which latitude/ longitude coordinates should be provided.
  4. MapZen
    Leveraging OpenStreetMap, MapZen provides geojson files for each country. The files contain country, region, and city boundaries and are updated on a monthly basis as needed. City boundaries for each country in this project are used in connection with the OpenSignal API to obtain cell signal statistics for capital cities/ districts.

Data Wrangling

After obtaining the data, several processing steps were needed to arrange the data in a usable way. Steps varied depending on the source of the data:

  1. World Bank Data: Data downloaded from The World Bank website is saved with columns such as: Country Name, Series Name, 1999, 2000, …, 2014, 2015. To get the value for Mobile Cellular Subscriptions in 2010 in Brunei, you would look for a row where Country Name = Brunei, Series Name = Mobile Cellular Subscriptions, and then find the value under the column 2010. To make each series more accessible, particularly within Tableau, the data was split into separate files per series with the following column structure: Country Name, Country Code, Year, Series (e.g. Mobile Cellular Subscriptions). Missing series values, indicated by “..” in the original file, were excluded from the new files. Each newly generated file was saved according to the series name.
  2. ITU’S Measuring the Information Society Report 2015: Because the report was released solely as a PDF, obtaining all data from the ITU report was done manually. Statistics used in the report (e.g. average broadband costs by country) were included as tables within the PDF, and the values were manually copied to an Excel file with one column for each statistic and one row for each country.
  3. OpenSignal API: Data retrieved from the API is sent in the JSON format (JavaScript Object Notation). The JSON responses for each API query were stored the parsed into an Excel format for each of use with Tableau.
  4. MapZen: Within the set of geojson files for each country, the boundaries for each capital city district were located and used with the Google Maps JavaScript API to outline on a map. This generated a visualization which helped identify the boundaries of latitude and longitude coordinates to use with the OpenSignal API. Using the drawn boundary as a guide, a simplistic polygon was fit along the city outline. When searching for coordinates, the polygon simplifies the mathematical checks to determine whether or not the point lies inside the region.

The process of determining which API calls to make against OpenSignal API took advange of the fact that city boundaries were known from MapZen. The methodology is as follows:

  1. Draw the boundaries of capital city districts using data from MapZen
  2. Draw a simplified polygon representing the boundary on the same map
  3. Identify the search region as the smallest rectangular area containing the simplified polygon
  4. Starting from the top left corner, sketch bounding boxes of size 10km x 10km within the rectangular area.
  5. For each bounding box, determine if the center point lies within the simplified polygon boundary. If so, save the center point to a list of coordinates to pass to the OpenSignal API.

To ensure that each data point returned from OpenSignal has an equal weight in terms of the square kilometers that it corresponds to, the same size bounding box was used for each country’s capital city (10km x 10km). As a result, the number of API calls for a given capital city varied from one country to the next and is proportionate to the size (in square kilometers) of the city. For instance, there was only one API request made for Manila in the Philippines, but there were 37 made for Vientiane (Laos).


Above: Bounding boxes representing coverage achieved with OpenSignal API data, Laos

Links to country maps:

There is a margin of error for each city: some bounding boxes may be partially outside of the city boundaries, and some areas within the city are not covered by any bounding boxes. There is a tradeoff between the size of the bounding boxes and the margin of error. For this project, the bounding box size was chosen such that the number of API calls was minimized while obtaining sufficient city coverage. Bounding box coverage maps are available with the project data files [6].

Appendix

References

[1] http://www.worldbank.org/en/about/what-we-do
[2] http://www.itu.int/en/ITU-D/Statistics/Pages/publications/mis2015.aspx
[3] http://opensignal.com/about/
[4] http://developer.opensignal.com/
[5] https://mapzen.com/data/borders/
[6] https://github.com/margaretpearce/lis677-project
[7] http://www.un.org/apps/news/story.asp?newsid=43459
[8] http://www.itproportal.com/2015/01/20/internet-access-necessary-survival-developing-countries/
[9] http://data.worldbank.org/news/2015-country-classifications
[10] http://www.internetsociety.org/what-we-do/where-we-work/asia/southeast-asia
[11] http://aseanup.com/southeast-asia-digital-social-mobile-2015/
[12] http://qz.com/560326/southeast-asia-is-about-to-pass-the-us-in-facebook-and-twitter-users/
[13] http://techcrunch.com/2015/09/08/why-southeast-asia-is-leading-the-worlds-most-disruptive-mobile-business-models/
[14] http://databank.worldbank.org/data/home.aspx
[15] Case, W. (2015). Routledge Handbook of Southeast Asian Democratization. Routledge.
[16] http://www.ericsson.com/res/docs/2014/regional-appendices-sea-final-screen.pdf