M-Pesa Usage Data: OTC Makes Sense, Even for Kenya

Cash-lite and cashless are big buzzwords at the moment in financial inclusion circles. They have become the goal of many business and public policy initiatives especially in the mobile ecosystem. Nonetheless, usage data, shared below, puts into question whether these goals are aligned with the needs of people at the base of the income pyramid. This data puts the spotlight on the core innovation of mobile financial services for the unbanked: not on cashless electronic wallets (i.e. “wallets”), but on highly distributed cash management networks.

Even M-Pesa, the most successful mobile wallet implementation in the developing world, generates evidence that people use it to get a hold of cash, rather than going cashless. Taking a look at M-Pesa’s latest usage data, it is evident that a large majority of the funds are cashed out after transacting only once (single loop).

Figure 1: Examples of one-loop transactions and two-loop transactions

This quick flow of funds through the system suggests that customers use M-Pesa as a set of services for managing and moving cash around: remittances, airtime top-ups, bill payments, etc. In this context, over-the-counter (OTC) transactional models should not be discarded as a sub-par solution given that they might even prove more efficient for a large number of use cases. Crunching some of the (scarce) numbers on the usage of M-Pesa in Kenya during the last semester, October of 2012 through March 2013, a conservative estimate reveals that around two-thirds of the money cashed-in did not transact more than once before being cashed-out (see chart below). This single loop pattern is consistent with other research in this field such as Mbiti & Weil (2013, 2011) or CGAP (2012).

Figure 2: Share of transaction aggregate value for two-loop transactions in two different scenarios, in Ksh H2 FY 2013

As only aggregate usage data is available, this analysis uses a best/worst case scenario methodology to assess the theoretical lower and upper limits for single- and two-loop transactions. In the best case scenario we make all the possible assumptions to maximize the aggregate value of multiple loop transactions, while in the worst case we maximize the aggregate value of single loops transactions.

This does not allow us to equate aggregate value share to their share of transactions. But making use again of the best/worst case scenario method and assuming a homogenous average transaction value among the different use cases, we can infer that, on one extreme, a third of the value of every transaction loops at least twice, which implies cashing in and out the other two thirds.

There are of course some caveats in this analysis. First of all, the large aggregate numbers made public by Safaricom and Kenya’s Central Bank do not allow us to differentiate among use cases and thus broad assumptions are required, such as the homogenous average transaction value described above. Additionally, we do not know important information on use cases such as time gaps between cash-in and transaction, zero-loop transactions, or use for short-term savings or treasury-like services. There are also conflicting studies around the occurrence of these last use cases, as the savings case for example (see Jack & Suri 2010 or Cohen & Stuart 2011).

Overall, although there are some interesting new developments with loans and savings products such as M-Shwari, there is a consensus around the idea that M-Pesa usage consists mostly on domestic transfers and airtime top-ups, while evidence of store of value or savings is scarce. These usage patterns are in general consistent with the case for a majority of single loop transactions.

This high incidence of single loop transactions inside M-Pesa has been known for some years, but its main consequence has been generally overlooked: these cases would have been better served by an OTC service rather than cashless wallets. Why?

Find out more in part II of this post this week.