Dr. Wayne Tarrant (2020, 2021, 2022)
When people think of financial innovation, Kenya is not often at the top of the list. However, Kenya was one of the first countries to fully embrace the potential for mobile payment systems. Despite the fact that only 19.2% of the country had access to electricity in 2010, Safaricom started the M-PESA (M is for Mobile, and pesa is the Swahili word for money) payment system in 2007, allowing customers to transfer funds from mobile phone to mobile phone with a text message.
Much of Kenya was unbanked as of just a few years ago, with over 60% of the country not having a bank account in 2011. This has changed since the advent of M-PESA. M-PESA exists as an alternative to carrying cash, with franchises at every Safaricom authorized dealer, at many gas stations, and at larger distributors. At present there are over 40,000 M-PESA agents in Kenya, giving M-PESA users access to rudimentary financial services virtually everywhere in the country. In fact, today more Kenyans are registered with M-PESA than have bank accounts, and less than 20% of the country lacks a bank account. Clearly M-PESA has made a difference.
In 2012 Safaricom launched a new product called M-Shwari (Shwari is the Swahili word for calm) in conjunction with the Commercial Bank of Africa (CBA). M-PESA customers can hold their funds in an M-Shwari account and earn interest of 2-5% per year, depending on balances they carry. Customers can also take out 30 day loans, with a facilitation fee of 7.5%, and pay those loans back electronically. Critically, the CBA is responsible for any losses on loan products. Thus, Safaricom customers do not lose their mobile phone service for failing to pay back their loans, and they do not have any airtime credit taken from them. Rather, the CBA charges late fees and puts a strike on their credit if they fall 120 days in arrears. That strike lasts for five years. CBA reports about 2% of M-Shwari loans are non-performing. (The most recent World Bank data is from 2016 and shows 3.917% of loan amounts worldwide as non-performing, and US data show a 50% default rate on similar types of payday loans.)
M-Shwari has been successful in employing data from M-PESA customer use to calculate a form of a credit score for potential borrowers. CBA forms a customer’s quasi-credit score based upon M-PESA transaction history. Though many companies have been using data analytics to offer enticing products to customers, CBA seems to be the first company to use big data to reach out to those in poverty. Clearly their algorithms are working, as their non-performing loan statistics are better than worldwide averages.
Overall, we want to pursue several questions related to electronic transactions. We would investigate if the conclusion of Jack and Suri about M-PESA lifting 2% of Kenyans from poverty is reproducible. Then, is M-Shwari giving Kenyans a chance to save, or are Kenyans ending up in greater debt? It seems clear that more Kenyans have access to a wider range of banking services because of M-PESA and M-Shwari and that there is more microfinance available to Kenyans. We would seek to show whether there is statistical significance in the numbers that are being briefly reported in the headlines.
We would be interested in the algorithms that M-Shwari is using to assess credit worthiness for the unbanked. Can such algorithms be altered to work in the other countries where M-PESA has a presence? Could new distributed ledger technologies (DLTs) like blockchain and directed acyclic graphs help in the storage and authentication of transaction data? Could DLT help with the CBA quasi-credit algorithms? Can such ideas be transferred to other countries, whether less developed or more developed? FDIC statistics show that about 7% of the United States is unbanked, for instance. Typical payday loans in the US incur a 521% APR, while M-Shwari’s rate is less than half of this. Do default rates correlate with quasi-credit scores? While questions about applicability to other countries would be mostly speculative, multiple regression may show which parameters are most relevant for success of programs like M-PESA and M-Shwari. We could then use data from other countries to determine if success is predicted.
Because there are a multitude of problems in myriad areas, we seek students at all levels of mathematical preparation. At a minimum an applicant should have completed single variable differential and integral calculus. Having succeeded in courses in statistics or probability would be helpful, and exposure to logic, differential equations, programming, linear algebra or multivariable calculus would likewise be useful.