1

Economics of the Internet A-Karim Lalani

Fragmentation in Mobile Payment Platforms: A Case Study on East Africa

The success of M-Pesa in Kenya has been celebrated as one of Africas greatest success stories. The service was introduced in 2007 and within five years M-Pesa has attracted fourteen million subscribers and in the last year processed $14.2 Billion in transactions1. M-Pesa is a mobile payment platform deployed by Kenyas largest telecom operator, Safaricom. The payment platform functions as a mobile wallet, an individual can deposit cash through an affiliated agent and then use their balance to pay and receive payments from other individuals using the platform. Individuals can also withdraw cash from their electronic account through an agent. M-Pesa is a household name in Kenya, however, the same payment platform has struggled in a neighboring country. The Vodafone group is a multinational firm that owns telecom operators globally. This paper is particularly interested in the fact that the firm owns 40% of Safaricom and has a majority share in Vodacom Tanzania. In 2003, Vodafone contracted a firm to develop what would become the M-Pesa platform and began to implement it in Kenya, Tanzania and Afghanistan post 2007. Despite the many economic and political differences between Kenya and Tanzania, we can argue that they have many similarities that make them comparable target markets. Both countries have similar population sizes (41 to 43 million people), population age structures, economic composition (both have approximately 75% of the labor force in agriculture) and both countries have experienced similar surges in mobile phone penetration rates. It is reasonable to expect the successes of M-Pesa in Kenya to be comparable to the successes of M-Pesa in Tanzania, however this is not the case. In May 2009 Vodacom Tanzanias implementation of M-Pesa is said to have attracted 280,000 users compared to Safaricoms 6.5 Million and had processed 3% the amount of money in transactions that year (IFC)2. The question that remains unanswered is: why is M-Pesa so much more successful in Kenya than it is in Tanzania? Having witnessed the potential of a mobile payment platform, many organizations are keen to replicate the success of M-Pesa in other parts of the world but it seems as though the Kenyan example is an outlier when compared to the deployment of similar platforms in other countries. There have been attempts to answer why M-Pesa has worked better in Kenya than in Tanzania, but these hypotheses have been stated and not verified through empirical data nor have they been based on the predictions of a formal economic framework3. In this paper I present my own hypothesis for M-Pesas user adoption, I describe a model that applies to the M-Pesa in and then compare the predictions of the model to our observations in both Kenya and Tanzania.

1 2

Kenyas GDP is $70.6 Billion, their transaction volume is very high relative to GDP The most recent statistics are less drastic and they will be evaluated later in the paper.

Economics of the Internet A-Karim Lalani

Section 1: Background and Hypothesis

Mobile payment platforms as they have been developed in emerging markets are structurally different from similar platforms in the developing world. Most payment platforms in the United States (this includes credit cards, the Square Card Case and Google Wallet) are tied to an underlying bank account and all payments made on the platform withdraw money from the account of the payer and deposit money into the account of the recipient. In emerging markets, payment platforms differ because their accounts are not connected to an underlying bank account, instead the accounts are registered with the platform operator who takes on the responsibility of providing the user the ability to deposit and withdraw money. Once the user has debited their platform account they are free to transact with other users on the platform using a virtual currency that is backed by actual cash. The operators earn revenue on each transaction on their platform and the individual making the payment is charged according to a tiered payment schedule4. Payment platforms in East Africa employ individual agents to convert physical currency to the platform-specific virtual currency and vice versa. Agents are generally entrepreneurial individuals, shop owners, or even ATMs. These agents allow users to deposit and withdraw money from their accounts and are paid for each transaction5. They are vital to the success of the platform because they collect the money underlying their virtual currency and provide liquidity to the users who can withdraw their money at any time. We expect that as the number of users increases there will be more trust in the virtual currency and transactions with other users on the platform will become vastly more popular than transacting with agents (Mas). In Tanzania there are four different telecom operators and each has their own mobile payment platform that is similar to M-Pesa. Each one of these platforms operates a different virtual currency, also known as an e-money, and each virtual currency is incompatible with the others. For example, if A uses platform 1 and B uses platform 2, A cannot pay B using platform 1 and expect Bs account balance to increase by the transaction amount. As the payment platforms have evolved, they have developed measures for weak compatibility, for example if A pays B who is on platform 2, B can withdraw money from an agent of platform 1 and then deposit it their account. True compatibility doesnt exist among the various payment platforms in East Africa and one virtual currency has to be converted to cash before it can be converted into another virtual currency. The telecom markets in both Kenya and Tanzania are also structured differently. There is a much higher level of fragmentation in the Tanzanian telecom market than there is in Kenya as depicted in Figures 1 and 2. Safaricom, the leading operator in Kenya, has a market share of almost 70% whereas the leading operator in Tanzania (also owned by Vodafone) has a market share of 45%. Competition is also much greater in Tanzania
4 5

Traditionally lower transaction values have had a higher percentage cost With the exception of ATMs that are not capable of letting the users deposit money.

Economics of the Internet A-Karim Lalani

whereas in Kenya, the combined market share of Safaricoms competitors is less than half their market share. The fragmentation of these markets may be a key predictor in the user adoption rates of M-Pesa in the respective countries.
Kenyan Subscribers (Millions of People)
1.63 2.75 4.17 17.95 Safaricom Airtel Orange yuMobile

Figure 1: Kenyan Telecom Subscribers (as of Jan 2012)

Tanzanian Subscribers (Millions of People)

1.524

5.45 11.625 Vodacom Airtel Tigo 6.993 Zantel

Figure 2: Tanzanian Telecom Subscribers (as of March 2012)

The differences in market fragmentation may significantly reduce the value proposition to an M-Pesa subscriber in Tanzania. Assuming a users transactions are mainly domestic, an M-Pesa subscriber in Kenya can transact with 70% of the mobile phone using population whereas an M-Pesa user in Tanzania can transact with less than

Economics of the Internet A-Karim Lalani

half of all telecom subscribers. It is more attractive to use M-Pesa in Kenya because of the existence of network effects. Network effects exist in a platform when the value of using that platform increases with the number of other people using it. This is the case in mobile payment platforms where the intrinsic value of the platform is lower if there is no one to transact with. Higher fragmentation reduces the value for the payment platforms because there are fewer users on any given network. It is the level of market fragmentation that I believe has caused such a large discrepancy in M-Pesas success in the two East African countries. I hypothesize that in a market for platforms, a greater degree of fragmentation will cause lower total user adoption at a given price level and that prices will have to drop significantly in order to attract more users. In the next section I propose a model to test this hypothesis and analyze the extent to which this hypothesis predicts the user adoption rates and relative revenues of M-Pesa in the two countries.

Section 2: A Model of Network Effects

As previously defined, a technology exhibits network effects when the value of that technology to a potential customer increases with the number of other people currently using it6. The technology this paper is concerned with, in network terminology, exhibits direct network effects and is arguably one-sided. Direct network effects refers to a simple increase in the value of a product as the number of people using it increases, and this is present in payment platforms as a larger user base provides customers more users to transact with. However, it is less obvious the extent to which the network is one-sided. In two-sided networks, the increase of users for one technology increases the demand for another. This is true in mobile payment platforms with respect to agents, as a higher number of agents can attract more users to the network. However, if we take the example of credit cards, which are the Western analogous of payment platforms, the two sides of the platforms are cardholders and merchants. Banks are the entities that support the credit card system. Similarly, agents act as supporters to the system and both merchants and cardholders have their own accounts on the platform, making the platform one-sided for financial transactions. In further analysis I consider mobile payment platforms as being one-sided. With regards to the adoption of M-Pesa, a consumer makes the decision whether or not to join the network. This is analogous to buying a good, but with the constraint that
6

An alternate way of viewing network effects is to consider them as positive externalities. For example, the number of users joining a social network is a spillover benefit to current users of that network because there is no mutually defined way of compensating the new users who are contributing to the surplus of old users. However, if the added surplus is compensated for, the network effect ceases to be an externality. I believe this is the case in payment platforms where the providers are able to charge higher prices for larger networks and capture the consumer surplus.

Economics of the Internet A-Karim Lalani

the consumer can purchase one good at most. The consumer decides to join the network only if the value of joining the network exceeds the cost to join it. If n is the number of user using the network, we can say the total value to the user is given by the function !(!), where the value is dependent on n. In the original model of one-sided networks, Katz and Shapiro (1985) consider a model where the good has an intrinsic value independent of the size of the network. In the case of mobile payment platforms, we have to ask whether there is any value to the first adopter of the network given there is no one else to transact with? The platform is provides no utility to a user if he or she cannot transact with anyone else, and hence there is no inherent value in joining the network. This can be restated as: ! 0 = 0 In making the decision to join a network, individuals can have different valuations for the platform for a given network size. If the network size was to be held fixed, a restaurant owner may value the platform more than a civil servant because the restaurant owner appreciates the ability to streamline a large volume of transactions whereas a civil servant who has to be frugal and makes few transactions doesnt value joining the platform as much. In both cases, their valuations differed due to factors that were unrelated to the size of the network. These random valuations suggest the total value function is composed of two separate functions: ! ! = !(!)!(!) The first function ! ! accounts for the random variation in values for each user, similar to distributed reservation prices that users have for goods in a traditional market. The second function !(!) describes the relationship between total value of the platform and the size of the network. Let us consider the nature of the relationship between total value and size of network. Adding an additional person should increase the total value of the platform, but there will be diminishing returns for each additional person. Mathematically we can say: !" ! ! > 0 !" !!! ! ! < 0 !! ! !" ! ! =0 !! !" lim Any value of ! ! where 0 < ! < 1 fulfills such these three conditions7, to choose an unbiased value we choose the midpoint of that range and assign: ! ! = !!
7

Other functions also fulfill the previously stated conditions but we choose nx for simplicity of calculation.

Economics of the Internet A-Karim Lalani

We borrow the functional form of ! ! from work done by Shapiro and Varian (1999) as follows: Lets take the example of the demand for a good that does not have network effects. We index 100 people by ! = 1, , 100 and define !, where ! measures the reservation price for a good by person !. Then if the price of the good is p, the number of people who think the good is worth at least p is 100 !. For example, if ! = 50, there will be 50 people for which ! > 50 and hence 50 people will buy the good. The consumer decides to buy that good if ! < !(!) and if 1 ! 100. In such a scenario there will be a person who is indifferent to buying the good. If this individual is indifferent to buying the good, we know that everyone with a higher value for !(!) will want to buy the good, so we know that: ! = 100 !(!) Rearranging, we can define ! ! as: ! ! = 100 ! Shapiro and Varian apply this same function for reservation prices to a market with network effects to model the different values a population has for a good at a given network size8. So if we consider the market of 100 people for a good that does exhibit network effects, the total value to an individual is defined by: ! ! = ! ! ! ! = (100 !)!! In equilibrium, the platform provider can set a price equal to the total value function such that: ! = (100 !)!! This equation gives us the relationship between price and size of the network. For a mobile payment platform a user can decide to join or not, which is analogous to buying at most one good. Hence we can take the total number of users to be the same as total quantities sold and the above equation represents an inverse demand curve. We see the shape of this curve in Figure 3 to be parabolic.
! !

Easley and Kleinberg (2010) suggest that the general function for r(x) need not look like the way defined by Shapiro and Varian, however, they go on to say that they typically expect to see a function of similar form

Economics of the Internet A-Karim Lalani

Figure 3: Demand for a Mobile Payment Platform

The parabolic shape of the demand curve describes three unique equilibriums when the price is set to 200. The first of which is if ! = 0. If there is currently no one on the network, then all individuals will derive no utility from the joining the network and no users will choose to join the network. This equilibrium exists in all platforms with similar demand curves, in which case we must ask why does the first user ever join a platform? We can answer this if we assume the user is forward looking and makes their valuation decisions based on their expected size of network. So the first user may expect that at a price of 200, the platform will reach equilibrium A and expect !! number of users when making their valuation decision. This is a self-fulfilling expectations equilibrium where if !! number of users expect as many people to join, they all join and fulfill their expectation. Once the platform gets to equilibrium A, the platform will have reached a critical mass. We see on the graph that for the !! + 1 user, his willingness to pay is higher than the price; hence this user will join the network. We see this is true for many users who will continue to join the network until we reach equilibrium B. In that sense, A is not a stable equilibrium because once a platform reaches equilibrium A, there will be upward pressure for the size of the network to reach equilibrium B. It is important to note that the process of moving from A to B is not instantaneous and requires time. In the example we have considered so far, we assume there is only one mobile payment platform in a market of 100 potential consumers. If we were to assume there is fragmentation, the demand curve would look different. Figure 4 describes the demand curves for a single mobile payment platform where there are 1, 2 and 4 total platforms in the market, each with equal market share. We observe that fragmenting the market reduces prices. As most markets, we expect competition to decrease profit margins for firms and increase welfare, however in

Economics of the Internet A-Karim Lalani

a market with network externalities, welfare may decrease with competition for identical platforms9.

Figure 4: Demand Curves in a Fragmented Market

Given the mobile payment platforms we consider in East Africa, the potential market share of each platform is given before the platforms compete. Telecom operators were already selling voice and text message services to their subscribers before they began selling financial services. If we assume there is a significantly high cost of switching between telecom operators because a customer would have to change their phone number, ask their network to update their contact information and potentially miss calls that were directed to their old number, we can assume that mobile payment platforms do not compete with each other with prices. So we can think of each platform provider as a profit-maximizing monopolist given within the subscribers of their telecom services. We assume no costs for simplicity, so the monopolist solves the profit function: ! = !" = ! !! (! !) Where n is the number of users that join the platform, p is the price they are charged and ! is the total size of their telecom subscriber base. We solve the profit function when ! = 100 to find ! = 60. This example corresponds to percentages, so the monopolist would want to charge a price at which 60% of his potential consumers join the platform. This model of network effects has several powerful predictions about the East African market for mobile payment platforms that shed light on the current state of the industry. The predictions are described as follows: 1. In a market with network effects, where there exist multiple firms who do not compete on prices and have no costs, the aggregate number of subscribers will be the same regardless of the number of competitors.
9

We assume the different platforms are identical for this to be true. Farrell and Saloner (1985) show this is not the case for differentiated products and that welfare can potentially increase if the dominant product is inferior to a competitive product.

Economics of the Internet A-Karim Lalani

If ! is the total size of the market, and there are m number of firms, if we assume they all have equal market share then each firm has ! ! potential customers. The firm maximizes ! = ! !! ( We find the maximum of this function when: 0.6 ! =! !
!

! !) !

If all firms m maximize their profits, and they all have equal market share, then the aggregate number of users is equal to: 0.6! ! = 0.6! ! This result still applies when firms dont have equal market share but only in an industry where all the platforms are identical and dont compete on price. The implications for this is that in both Tanzania and Kenya we would expect telecom operators to have 60% of their total subscriber base join their networks if they are profit-maximizing. However, the assumption of all operators being profit maximizing may not be reasonable. Introducing a new product to a new market is challenging for a monopolist who yet to learn the preferences of their potential customers and it takes much learning by doing to learn their what the demand curve looks like a new market. Hence if we assume that platform providers have the same pricing strategy, which means that they set their prices to sell a certain quantity, we would expect they would attract the same proportion of their total subscriber base. We see that this is empirically the case for M-Pesa in both Kenya and Tanzania.

Figure 5: 2009 Price Schedules for M-Pesa in Kenya and Tanzania

10

Economics of the Internet A-Karim Lalani

Figure 5 shows the 2009 price schedules for M-Pesa in both Kenya and Tanzania and we see they are very similar. Due to the differences in market share, we would expect significantly lower users of M-Pesa Tanzania as a proportion of Vodacoms customers. In 2009 Safaricom had converted 50% of their subscriber base to M-Pesa users whereas Vodacom had converted only 5% (IFC). These price schedules have changed over time and in early 2012 Vodacom recently reduced M-Pesa prices by 75%. If we assume over time M-Pesa has adopted a unified pricing strategy to induct the profit maximizing number of users across countries, then we would expect the conversion rate of subscribers to M-Pesa users to be the same in both countries. In December 2011 Safaricom reported that approximately 81% of its 22 Million subscribers at the time had created accounts on M-Pesa. Later in February 2012 Vodacom reported that roughly 9 of their 11.6 Million users had created accounts on M-Pesa (though they did claim only 3 Million were active users). That means Vodacom had attracted close to 78% of their total user base to subscribe to M-Pesa, which is comparable to their Kenyan counterpart10. 2. If there are two firms offering identical platforms and arent competing with prices, then the firm with lower market share will have to charge a lower profitmaximizing price. This can be restated as, in a market for platforms the greater the degree of fragmentation the lower the prices charged will be. In Figure 4 we see that when a firm has lower market share it shifts the firms demand curve downward so the firm is forced to reduce prices. We can see this more simply when: ! = !! (
!

! !) !

As ! is fixed in a market, only m can increase which causes prices to decrease (as long as !/! is greater than !). This shows that as the number of firms increase, which means there is a greater degree of market fragmentation, the prices charged by a platform provider will decrease11. This result is the basis for the next prediction that revenues also fall with fragmentation. 3. In a market where firms offer identical platforms and arent competing with each other through prices, a greater degree of fragmentation will lower the aggregate revenues earned in that market. We can reinterpret a greater degree of fragmentation as a downward shift of the demand curve. However, with a parabolic demand curve the drop in maximum profits is disproportionately greater. In the example so far, if we assume each firm maximizes profits then the total revenues for each firm have been depicted in Figure 6. We see that doubling market share for a firm increases revenues by 566% and quadrupling market
10

There was no credible resource suggesting that the countries have adopted a uniform pricing strategy and this was just a hypothesis. However, it is a reasonable hypothesis that after several years of learning by doing the parent company has issued its subsidiaries to implement the best practices they have learned. I was unable to collect data on the most recent price schedules and platform user statistics for M-Pesa and other mobile payment platforms at the time of writing this paper to make further comparisons. 11 The decrease in prices charged by Vodacom for M-Pesa Tanzania supports this prediction.

11

Economics of the Internet A-Karim Lalani

share increases revenues by 3200%. This general result of disproportionate changes in revenue for a linear increase in quantity sold should hold for any firm with a parabolic demand function. The difference is quite large at an aggregate level, for example a market with four firms with equal market share will earn in aggregate 232412 units of revenue whereas that same market with just a single firm will earn 18590 units of revenue, approximately 8 times as much.

Figure 6: Revenue for Firms in Fragmented Markets

We can begin to compare this prediction to what we observe in the mobile payment markets. Due to a scarcity of comparable revenue data for M-Pesa in both Kenya and Tanzania, we consider transaction amounts processed as a proxy for revenues. In December 2011 M-Pesa in Kenya had reported processing $14.2 Billion in transactions. M-Pesa Tanzania has less reported data but has claimed to process over $400 Million in December 2011, if we assume they had similar transaction volumes for each month then they would have managed to process $4.8 Billion in transactions, almost a third of their transaction volume while they have 25% less market share. This leads to a key prediction for incompatible networks: 4. In a fragmented market where firms offer identical platforms and that arent compatible, the firms would generate higher aggregate revenues if they operated a compatible platform. We observe in Figure 6, aggregate revenues are greater in a market with one payment platform rather than a fragmented market. Similarly, a fragmented market can choose to make their platforms compatible. The effect of compatibility is that consumer will make their decision to join a platform if ! ! !! > ! where ! is the total number of users in the market and it is constant regardless of the number of platforms that arise. The market operates as if there is effectively a single payment platform provider and the various platform owners can share the profits fairly. For example, in a market where there are
12 !

This is because each individual firm earns 581 units and there are four firms, so the aggregate revenue is 581 * 4 = 2324 units

12

Economics of the Internet A-Karim Lalani

four platforms with equal market share, making them compatible with each other will generate 8 times the aggregate revenue and each firm should rationally be willing to switch to such a platform if they receive a share of the increased revenue.

Conclusion and Further Analysis

This paper tried to evaluate the initial hypothesis that a fragmented market for platforms reduces the total user adoption rates for a give price level and that for a platform to induce a higher number of adopters they will have to lower prices. The model of network effects put forward is applied to mobile payment platforms in both Kenya and Tanzania, and we see the claim of lower user adoption rates being invalidated in theory but we still see that firms in fragmented markets have to significantly reduce prices to attract users. The model seems to predict many of the empirical observations in the growth of M-Pesa in both Kenya and Tanzania, however there are many predictions that could have been tested more rigorously with more data. For further analysis, I would like to have specific data on the user growth of each East African mobile payment platform with respect to the total subscriber base of the telecom operator since its inception. Ideally the dataset would include all similar mobile payment platforms globally. With such data we could test whether firms that operated in multiple countries yet had the same pricing policy achieved similar user adoption rates as a percentage of the total subscriber base. If I could also get the price schedules of each firm I could test whether the average price is lower for networks with fewer subscribers in a specific country. And if I could also get the monthly revenue data I would compare the total revenue numbers for firms with larger market share with those with lower market share and see if the former firms have disproportionately larger revenues. One of the most significant predictions of the model was that of compatibility. It suggests that the payment platforms in East Africa, or any other fragmented market, develop a compatible payment platform and share the profits fairly. Such a change in the payment platforms could potentially have a significant impact on the GDP as more users discover the value of using an electronic payment platform instead of just transacting with cash. Several other areas on enquiry would be how exactly mobile payment platforms add value to a consumer. For example, Id like to begin to look at how these platforms provide individuals with a measure for accountability and how it can further enhance accountability in countries with few transaction records.

13

Economics of the Internet A-Karim Lalani

Bibliography
Camner, Gunnar, Caroline Pulver and Emil Sjblom. What makes a Successful Mobile Money Implementation? Learnings from M-PESA in Kenya and Tanzania. White Paper Report. London: GSMA, 2010. Consultative Group to Assist the Poor. Mobile Banking in Tanzania: Can Kenya's success be replicated next door? 21 June 2009. 10 March 2012 <http://www.cgap.org/p/site/c/template.rc/1.26.10908/>. Easley, David and Jon Kleinberg. Networks, Crowds and Markets: Reasoning about a Highly Connected World. New York: Cambridge University Press, 2010. Farrell, Joseph and Garth Saloner. "Standardization, Compatibility and Innovation." The RAND Journal of Economics 16.1 (1985): 70-83. Greenstein, Shane. The Lexicon of Network Economics in Evolving Markets: An Applied Guide. White Paper Report. Evanston: Northwestern University, 2011. iHub. iHub Research. 10 March 2012. 10 March 2012 <http://www.research.ihub.co.ke/pages/resources.php>. International Finance Corporation. M-Money Channel Distribution Case Tanzania. White Paper Report. Washington DC: World Bank, 2009. Jack, William and Tavneet Suri. The Economics of MPESA. Research Report. Cambridge: MIT Sloan, 2010. Kagwe, Winfred. Kenya: Safaricom Loses Eight Percent Voice Market Share. 6 January 2012. 16 March 2012 <http://allafrica.com/stories/201201090118.html>. Katz, Michael and Carl Shapiro. "Network Externalities, Competition, and Compatibility." The American Economic Review 75.3 (1985): 424-440. . "Technology Adoption in the Presence of Network Externalities." Journal of Political Economy 94.4 (1986): 822-841. Mantshantsha, Sikonathi. Vodacom Sales Rise 12% on Record New Customers, Data Usage. 12 February 2012. 15 February 2012 <http://www.bloomberg.com/news/2012-0208/vodacom-revenue-rises-12-in-december-quarter-adds-customers.html>. Mas, Ignacio and Dan Radcliff. " Mobile Payments go Viral: MPESA in Kenya." White Paper Report. 2010. Ng'wanakilala, Fumbuka. Tanzania's Vodacom adds 4 million users. 17 February 2012. 16 March 2012 <http://af.reuters.com/article/investingNews/idAFJOE81G05Y20120217?pageNumber=1 &virtualBrandChannel=0>.

14

Economics of the Internet A-Karim Lalani

Ombok, Eric. Kenyan Mobile-Money Transactions Jump 60%, Business Daily Says. 27 February 2012. 15 March 2012 <http://www.bloomberg.com/news/2010-0526/safaricom-of-kenya-full-year-profit-climbs-44-as-sales-from-data-increase.html>. Oren, Shmuel and Stephen Smith. "Critical Mass and Tariff Structure in Electronic Communications Markets." The Bell Journal of Economics 12.2 (1981): 467-487. Shapiro, Carl and Hal R. Varian. Information Rules. Cambridge: Harvard University Press, 1999. Tanzania Daily News. Tanzania: Vodacom Cuts M-Pesa Transaction Fees By 75 PC. 13 February 2012. 16 March 2012 <http://allafrica.com/stories/201202131394.html>. TelecomPaper. Tanzania ends Q1 with 24 mln telecom subscribers. 31 January 2012. 16 March 2012 <http://www.telecompaper.com/news/tanzania-ends-q1-with-24-mlntelecom-subscribers>. TeleGeography. TCRA reports 25.83m fixed and mobile users at end-2011. 15 March 2012. 16 March 2015 <http://www.telegeography.com/products/commsupdate/articles/2012/03/15/tcra-reports25-83m-fixed-and-mobile-users-at-end-2011/>.