How Do Kenyans Really Use M-PESA?

A row of stores in Kenya that accept M-PESA. Photo: Darwin Wanjiru, 2018 CGAP Photo Contest
Photo: Darwin Wanjiru, 2018 CGAP Photo Contest

How do low-income Kenyans really use M-PESA? CGAP recently partnered with the Busara Center for Behavioural Economics to attempt to answer this question. A group of over 400 low-income users living in Nairobi consented to share anonymized transaction data and participated in a survey in the Busara lab. While not nationally representative, the data provides insight into M-PESA use. We encourage others to see what they can find in the full, anonymized data set, but below are some preliminary findings.

It’s no surprise – M-PESA remains primarily a payments tool 

The average monthly balance among participants’ wallets was low — less than Ksh 1,000 ($10) for 64 percent of the sample. The average net balance (difference between inflows and outflows) ranged between +/- Ksh 250 ($2.50). These balances were the same across employment types and income levels.

Net balances were also similar whether or not the user had a bank account. This suggests that while banked M-PESA users might be saving money in the bank and using their mobile money accounts for transactions, those without bank accounts are finding ways to save outside of their wallet accounts (to the extent they are saving).

M-PESA: Average Net Balance (Inflows - Outflows)

Average net balance M-PESA
Source: M-PESA user transaction data. n=400

This finding will be nothing new to most readers. As noted in the 2017 CGAP paper “Banking in the M-PESA Age,” Safaricom is responding to these behaviors with products like M-Shwari, allowing M-PESA users to open bank accounts from their wallets to further promote savings among M-PESA users.

M-PESA: Distribution of Average Monthly Balance

M-PESA: Distribution of Average Monthly Balance
Source: M-PESA user transaction data. n=400

Users also tend to have a biased perception of their average monthly balance. Comparing survey data with actual use of mobile money revealed that participants with an average balance under Ksh 500 ($5) on their wallets overestimate their balance by 41 percent, while users with higher balances underestimate their average balance by 59 percent. Similarly, respondents who make fewer than four withdrawals per month overestimated the frequency of their withdrawals, whereas those who take out cash more frequently underestimated their withdrawals.

Mobile money agents are more like petrol stations than barber shops 

Respondents did not appear loyal to individual agents. Only 5 percent of users always went to the same agent to make a deposit, and just 3 percent always went to the same agent to make a withdrawal. In this way, the use of a mobile money agent has more in common with a petrol station (or an ATM) — a place to stop quickly and transact — than a barber shop — a place that has many repeat customers. Transaction data showed that Kenyan customers are more loyal to retail stores like grocers than they are to agents. In a sense, this is not surprising since deposit and withdrawal services are largely commoditized.

However, M-PESA users did tend to use fewer agents for deposits than for withdrawals. The average user saw two to five agents for deposits compared to 10 to 20 for withdrawals.

M-PESA: Number of Agents Used

M-PESA number of agents used
Source: M-PESA user transaction data. n=400

Liquidity stock-outs were common from the customer’s perspective, with 41 percent of respondents saying they experienced at least two cash-in/cash-out issues per month. However, customers did not see these as challenges; 86 percent simply went to another agent to conduct the same transaction.

The number of monthly deposits and withdrawals was high across all income levels, but customers did tend to withdraw money slightly less often than they deposited it, and they withdrew higher values. This suggests that users wait to withdraw larger amounts at one time.

Digital credit does not stay digital for long

Just over one-third of participants had taken a loan from a digital credit provider, with an average ticket size of Ksh 2,420 ($24). This finding corroborates earlier survey research conducted by FSD Kenya, which found that digital borrowers who own mobile phones tend to take out loans of a similar size. Study participants withdrew 46 percent of loan funds out of their wallets within the first day. The largest percentage of respondents cashed out the funds (56 percent of users). The second highest use case was a personal transfer to another customer (25 percent of users). The average borrower took 4.5 loans, and 73 percent were customers of only one credit provider.

What other insights can you find?

This study is not nationally representative, nor should it be taken to explain behavior in other markets. But it does provide a detailed glimpse into how a certain group of customers is using mobile financial services in Kenya. The data is available online, and we would love to hear what others can learn.



This Working Paper explores three approaches banks in Kenya have used to respond to mobile money. While nonbank mobile financial services can fundamentally reshape the financial sector in a developing market, as they have clearly done in Kenya, mobile services need not represent an existential threat to the traditional banking industry.

Add new comment