Mobile Money: Issue 2

As of 2021, 10% of adults in developing economies owned a mobile money account, a six percentage point increase on 2014 – mobile money penetration has been most prominent in the Sub-Saharan Africa region, where 33% of adults now have a mobile money account (Demirguc-Kunt et al. 2022). One of the most successful deployments of mobile money has been M-PESA in Kenya. M-PESA has been widely adopted, with 97% of households having an account as of 2014 (see Jack & Suri 2016). Although other countries are now catching up: for example, in Uganda, 51% of individuals older than 15 years have an account, and in Tanzania, 39% do (Demirguc-Kunt et al. 2018 ), there are still many unsuccessful deployments of mobile money. Although it is hard to causally identify the reasons for the success (or lack thereof) of mobile money deployments, it is worth discussing some of the hypotheses for why mobile money has been a success in some economies but not others. It is important to understand how the business models and implementations of the various services may differ across countries and what has correlated with success.

As a summary example, Vaughan et al. (2013), some of the actual implementers of M-PESA in Kenya[1]1. Pauline Vaughan ran M-PESA from 2007 to 2009, and Michael Joseph was the CEO of Safaricom from 2000 to 2010. , describe their pilot, which started in October 2005 with a grant from the UK Department for International Development’s innovation fund and with microfinance clients. The product was then changed and rebranded based on consumer feedback as an internal remittance product to send money to friends and family – this experimentation was important to the success of the product. Vaughan et al. (2013) also highlight some additional key factors that allowed mobile money to reach scale in Kenya, in particular, developing a strong network of agents, removing entry barriers for customers, investing in the infrastructure for scale at the very outset, and regulating the system after the innovation. Of course, the fact that Safaricom had a large market share also likely played a role.

The success of mobile money systems is certainly underpinned by the rapid deployment and growth of the agent network, i.e. the end distributors of the service. This growth and reliability are associated with a network that is trustworthy, efficient, liquid, and profitable for the agents. As an example, Figure 5 shows the rollout of mobile money agents in Kenya during the success of MPESA, displaying the growth in access to agents in 2007 (just as M-PESA launched), 2011 and 2015. Note that there were fewer than 1,000 bank branches, just over 1,000 ATMs, and 3,000 M-PESA agents across Kenya in early 2008 (Camner et al. 2009). At the time of writing, there are 141,542 agents serving both M-PESA and other mobile money customers.[2]2. Communications Commission of Kenya, Quarterly Sector Statistics Report for Quarter 2, 2015–2016 (October to December, 2015).   Table 1 shows data on the number of agents in some of the countries with mobile money deployments.[3]3. The year of the data varies across countries because of the difficulty in finding recent data on the number of agents in each country.  It is striking that the number of agents is at least triple the number of bank branches in Kenya, Uganda, Tanzania, and Bangladesh.[4]4. These numbers come from the FSP Maps data, which was collected in Kenya in 2015, in Nigeria in 2015, in Tanzania in 2014, in Bangladesh in 2013, and in Uganda (date unknown). The FSP maps data is available at http://fspmaps.org.  Adoption rates in Sub-Saharan Africa range from 67% in Kenya to 9% in Nigeria (Demirguc-Kunt et al. 2022), where mobile money is both poorly deployed and poorly adopted.

Table 1: Number of agents in selected markets

Notes: [a] Data taken from Groupe Speciale Mobile Association data from the Agent Management toolkit, 2012. [b] Data taken from the Financial Services for the Poor maps data for Kenya from 2015, Nigeria from 2015, Tanzania from 2014, Bangladesh from 2013, and Uganda (date unknown). Data available at http://fspmaps.org

As these agent networks grew and became denser, the distance between households and agents shrank. For example, Table 2 shows how the average distance to a mobile money agent changed in Kenya between 2007 and 2015, and how it compares to the average distance to a bank branch. These averages mask a lot of heterogeneity: in 2007, 32% of households lived more than 10 km away from a bank branch, and 19% lived more than 20 km away, whereas 46% of households lived within 1 km of an agent, a number that rose to 68% by 2015. In addition to a dense network of agents, successful deployments of mobile money had a network of agents that were efficient at managing their e-money and cash inventories, helped by consistent monitoring and liquidity management by the service operator. Eijkman et al. (2010) show that agents rebalanced their accounts almost daily, more frequently in urban areas. In addition, agents faced a lot of competition, as consumers favoured agents with better service and trading volumes.

Table 2: Average distance to the closest financial institution, Kenya

Data taken from Finaccess Geospatial Mapping 2016 (https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/SG589T)

Similarly, in a later study, Balasubramanian & Drake (2015) look at how the demand for mobile money in Kenya and Uganda is affected by agent quality (measured in terms of pricing transparency and expertise) and agent competition.They combine a survey of 3,000 mobile money agents with location data on 68,000 financial access and transportation points, spatial census data, and population and poverty estimates. They find that greater agent competition is associated with a higher inventory of both cash and e-money, that more transparency in pricing and greater agent expertise are associated with higher demand, and that the return to expertise increases with competition.

Aside from the agent network, there are a number of other factors that may have driven the successful adoption of M-PESA, and have been described qualitatively. Mas & Morawczynski (2009) attribute some of the success of M-PESA to strong branding, an easy-to-use product, simple and transparent retail pricing, free deposit and no minimum balance feature, ability to send money to nonusers, and ability to perform ATM withdrawals. Heyer & Mas (2009) highlight the importance of volume, momentum, and coverage, as well as the regulatory environment, the quality of the retail infrastructure, and the high telecom penetration. Mas & Ng’weno (2010) highlight brand management, channel management, and pricing as the major contributing factors behind M-PESA’s massive success. Similarly, Mas & Radcliffe (2010) discuss the clever and easy-to-use design, and Safaricom’s business model. The authors suggest that the differentially wide spread of mobile money across countries could be partially attributed to differential regulations. M-PESA, in particular, benefited from a good working relationship between Safaricom and the Central Bank. Mas & Radcliffe (2011) also highlight how important network effects and trust are in scaling up a retail payment system. More recently, Lal & Sachdev (2015) compare five successful mobile money deployments[5]5. These are Telesom ZAAD in Somaliland, Dialog eZ Cash in Sri Lanka, Econet EcoCash in Zimbabwe, SMART Communications SMART Money in the Philippines, and Globe Telecom GCASH in the Philippines.  to five less successful ones.[6]6. These are Vodacom M-PESA in South Africa, MTN money in Uganda, Eko Financial Services in India, and the broader situations in Nigeria and Brazil, although MTN money in Uganda is now growing rapidly.  In addition to the relationship with regulators and the agent networks mentioned above, they suggest that adoption is also driven by an underlying reliable mobile network with a successful and trusted brand and business.

One of the earliest quantitative studies of mobile money started in 2008 in Kenya around M-PESA. Jack & Suri (2011), in this and later work, document the patterns of adoption of M-PESA over 2008–2014[7]7. The surveys were conducted in 2008, 2009, 2010, 2011, and 2014. The sample was not national—it excluded sections of the North of the country, where households are semi nomadic and therefore hard to track over time. The part of the country excluded from the sample covered about 8% of the Kenyan population at the time. The sample was also weighted toward urban areas, as the initial sample drawn in 2008 was weighted by the number of agents in the administrative location (there were about 2,500 locations in Kenya at the time). The reader is referred to Jack & Suri (2011, 2014) for more information on the sample.  using household surveys conducted across a large part of the country. They tracked the fast adoption of M-PESA in Kenya and traced the rollout of the agents in Kenya until 2010, recording the characteristics of the adopters and collecting data on access to the service. As expected, the initial users were richer and more educated; however, adoption of the product did reach down the income spectrum in the country, with over 90% of households in their sample having an account by the time of their last survey in 2014.[8]8. Mbiti & Weil (2011) also document the characteristics of the users.  Khan & Blumenstock (2016) study the adoption of mobile money more carefully in Ghana, Zambia, and Pakistan. They build a supervised machine-learning model of adoption using call record data and find this model does not distinguish very effectively between active and registered mobile money users, contrary to expectations that active users should be quite distinct in their patterns of phone use. Across countries, it is unlikely that any single set of characteristics will consistently predict mobile money adoption and use.

Beyond the household level, Ortigao et al. (2015) survey firms in Maputo and Matola in Mozambique and show that financial illiteracy on the part of both the seller and the buyer restricts the use of financial services such as point of sale (POS) devices. Similarly, lack of trust and knowledge, coupled with technological issues, is a hindrance to business owners using mobile payments. Finally, they show that the adoption of POS devices is positively correlated with the size of business and the volume of transactions, whereas the use of mobile phone technologies for payments is related to the owner’s age and whether they are a frequent cell phone user. 

Cruces et al. (2020) design an RCT in the Gambia to understand the barriers to the use of mobile money. They focus on a sample of individuals who had mobile money but had never used it, and find that offering discounts on withdrawals, and making these salient, created more awareness of mobile money, but did not increase the use of mobile money. They also find that the treated individuals were more likely to perceive the service as expensive. Clearly, there are still cost barriers to making mobile money useful in such environments. Karra et al (2022) also study the adoption of mobile money in underserved populations. Working with M-PESA in Mozambique, they study the impact of the gender of the telephonic sales representatives on take up. They find that female representatives had lower adoption rates of SIM cards, relative to male representatives, but they outperformed their male counterparts in terms of M-PESA enrollments.

Overcharging by mobile money agents is also a potential barrier to mobile money adoption, particularly for women. In Ghana, Annan (2022a) shows that over-charging by mobile money agents is common, taking place in 27% of transactions, with a mean overcharge rate of 54-82% of the official charge. Female vendors are more likely to overcharge and female customers are more likely to experience overcharging. Building on this, Annan (2022b) uses an RCT of a programme providing information on prices and the ability to report overcharging to show that information significantly reduces the incidence and severity of overcharging. Treated consumers trust mobile money agents more and are more likely to trust agents who are not engaging in overcharging. Treated consumers also increase their use of mobile money services, both for transactions and for saving. In a similar vein, Kubilay et al. (2023) study scams on mobile phones (inclusive of, but not limited to, mobile money related scams) in Kenya, and highlight the challenges of improving consumers ability to identify these scams.

Although mobile money was a technological innovation, it was enabled by some creative regulation and, more importantly, a network to efficiently distribute and manage cash across vast distances. As we describe in Section 5, one of the most important uses of mobile money has been P2P remittances. Therefore, having a widespread agent network whose cash and e-money inventories are well managed is crucial to the success of the product. Of course, once adoption starts, there will be strong network effects, even stronger than for mobile phones themselves given that there is little interoperability in these markets. There has been surprisingly little work documenting network effects in the adoption of mobile money. An exception is Batista and Vicente (2020) who conducted a randomised experiment with 200 primary farmer subjects in the Manica province of Mozambique, and 400 of their farming network members. All primary subjects were newly given access to mobile money accounts. In the treatment group, the two closest farming friends of the primary subjects were provided with a mobile money account. The results suggest that the network intervention increased the general use of mobile money by primary subjects and their network members, and reduced household expenditures and lending to social networks. These patterns are consistent with lower social pressure to share resources induced by the network treatment, although the mechanisms underlying this effect cannot be precisely distinguished in the context of this experiment. Further work examining this and related issues is required.[9]9. An early exception is Fafchamps et al. (2016), who study a precursor to mobile money, the P2P transfer of prepaid airtime credit in Rwanda. More recently, Batista, Fafchamps and Vicente (2022) describe sharing behavior of information about P2P mobile money transfers within exogenously created networks of mobile money users in rural Mozambique. The results from their lab-in-the-field experiment show that simpler interventions within these networks are likely to produce strongest information sharing: anonymous communication, for example, yields higher sharing than homophily within networks.

References

Annan, F (2022a), “Misconduct and reputation under imperfect information.” Working paper available at SSRN 3691376.

Annan, F (2022b), “Gender and financial misconduct: a field experiment on mobile money.” Working paper available at SSRN 3534762.

Balasubramanian, K and D Drake (2015), “Service quality, inventory and competition: An empirical analysis of mobile money agents in Africa”, Harvard Business School Technology & Operations Mgt. Unit Working Paper, (15-059).

Batista, C and P C Vicente (2020), “Improving access to savings through mobile money: Experimental evidence from African smallholder farmers”, World Development, 129, 104905.

Batista, C, M Fafchamps and P C Vicente (2022), “Keep It Simple: A field experiment on information sharing in social networks” (No. w24908), National Bureau of Economic Research.

Camner, G, C Pulver and E Sjöblom (2009), “What makes a successful mobile money implementation? Learnings from M-PESA in Kenya and Tanzania”, London: GMSA.

Cruces, G, J Hamidou, A Touray, and F Singhateh (2020), “Information, Price, and Barriers to Adoption and Usage of Mobile Money Evidence from a Field Experiment in the Gambia.” Partnership for Economic Policy Working Paper.

Demirgüç-Kunt, A, L Klapper, D Singer, and S Ansar (2018), “The Global Findex Database 2017: Measuring financial inclusion and the fintech revolution.” World Bank Publications.

Demirgüç-Kunt, A, L Klapper, D Singer, and S Ansar (2022), “The global findex database 2021: Financial inclusion, digital payments, and resilience in the Age of COVID-19.” World Bank Publications.

Eijkman, F, J Kendall and I Mas (2010), “Bridges to cash: The retail end of M-PESA”, Savings and Development, 219-252.

Fafchamps, M and M Soderbom (2016), “Adoption with social learning and network externalities” (No. w22282), National Bureau of Economic Research.

Heyer A and I Mas (2009), “Seeking fertile grounds for mobile money.” Unpublished manuscript.

Jack, W and T Suri (2011), “Mobile money: The economics of M-PESA” (No. w16721), National Bureau of Economic Research.

Jack, W and T Suri (2014), “Risk sharing and transactions costs: Evidence from Kenya’s mobile money revolution”, American Economic Review, 104(1), 183-223.

Jack, W and T Suri (2016), “The long-run poverty and gender impacts of mobile money,” Science, 354(6317), 1288-1292.

Karra, M, M Hernandez, C Brennan, and M McConnell (2022), “Supply-Side Innovations to Increase Equitable Access to Digital Financial Services: Experimental Evidence from Mozambique.” Boston University- Department of Economics.

Khan, M R and J E Blumenstock (2016), “Predictors without borders: behavioral modeling of product adoption in three developing countries”, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 145-154.

Kubilay, E, E Raiber, L Spantig, J Cahlikova, L Kaaria (2023), “Can you spot a scam? Measuring and improving scam identification ability”, Working paper.

Lal, R and I Sachdev (2015), Mobile money services: Design and development for financial inclusion, Harvard Business School (pp. 15-083).

Mas, I and O Morawczynski (2009), “Designing mobile money services lessons from M-PESA”, Innovations: Technology, Governance, Globalization, 4(2), 77-91.

Mas, I and A Ng’Weno (2010), “Three keys to M-PESA’s success: Branding, channel management and pricing”, Journal of Payments Strategy & Systems, 4(4), 352-370.

Mas, I and D Radcliffe (2010), “Mobile Payments Go Viral M-PESA in Kenya”, Yes Africa Can, 353.

Mas, I and D Radcliffe (2011), “Scaling mobile money”, Journal of Payments Strategy & Systems, 5(3), 298-315.

Mbiti, I and D N Weil (2011), “Mobile banking: the impact of M-Pesa in Kenya” (No. w17129), National Bureau of Economic Research.

Ortigao, M, E Macome, P Vicente (2015), “Electronic payments in Mozambique: a baseline on their adoption in Maputo and Matola.” NOVAFRICA Working Paper 1503, Univ. Nova Lisboa, Lisbon.

Vaughan, P, W Fengler and M Joseph (2013), “Scaling-up through disruptive business models. The inside story of mobile money in Kenya”, Getting to scale: How to bring development solutions to millions of poor people, 189-219.