A recent study of digital lending in Kenya reveals the benefits of relying on mobile phone data to expand credit access—improving financial outcomes and overall well-being even for individuals borrowing as little as $40.
Tala, a fintech lender that uses information gleaned from mobile phones to determine creditworthiness, randomly approved potential borrowers who would have otherwise been rejected by its own algorithm, creating an opportunity for researchers to study how expanded credit can affect outcomes for households and businesses outside the traditional banking system.
A new paper forthcoming in The Accounting Review, “Digital Lending and Financial Well-Being: Through the Lens of Mobile Phone Data,” finds that these borrowers who are randomly approved for a digital loan saw bigger mobile money balances, higher income, better employment prospects, and even an expansion in their own social networks, says Harvard Business School Assistant Professor Jung Koo Kang, a coauthor of the new report.
“Conventional wisdom would suggest that these borrowers would be more likely to misuse credit, but instead, we found that they experienced significant improvements in financial well-being,” Kang explains. “This may suggest that the traditional cutoffs for creditworthiness can potentially be conservative in some emerging market settings, where access to even small amounts of credit can unlock economic opportunities for these borrowers.”
The analysis reveals how financial flexibility—enabled by alternative data—can unlock prosperity in countries where people and businesses tend to have lower incomes and may have previously been considered too risky by traditional lenders. To be sure, Kang says, some digital lenders and other high-interest rate creditors, like payday lenders, prey on poorer communities with excessive late fees and other tactics, making the sector one to regard with caution.
Kang coauthored the paper with AJ Chen of the University of British Columbia; Omri Even-Rov of the University of California, Berkeley; and Regina Wittenberg-Moerman of Northwestern University.
Markets emerge for new kinds of loans
Some 3.5 million of Tala’s 10 million global borrowers are based in Kenya, where about a quarter of adults had secured a digital loan by 2018. These digital lenders make credit decisions using nontraditional data from borrowers’ mobile devices, like transaction histories and travel patterns. In many poorer countries, households lack formal credit scores, bank accounts, and tax returns on which lenders often rely.
“Entrepreneur opportunities in these markets are very different from developed markets—for example, people open food stands on the street. It’s a very informal kind of economy. They may not necessarily have tax filings and formal documents to prove their income sources and business,” Kang explains. “Fintech lenders like Tala see a huge opportunity if they can distinguish between the credit profiles of those ‘invisible’ borrowers.”
The authors chose Kenya in part because it has one of the “most developed mobile money ecosystems in emerging markets, with a high adoption of digital financial services,” Kang notes. Tala discloses fees and repayment incentives to help borrowers, in keeping with regulatory guidelines, the paper explains.
Approving loans for applicants rejected by algorithm
Tala created a proprietary database of digital loans in Kenya, which it administers via a mobile app where borrowers apply for loans and grant the company access to their mobile phone data.
The database contained several important features for the researchers to study:
Random approvals for some applications. Starting in 2018, Tala randomly approved some applications that would have otherwise been rejected based on its credit scoring model as part of an internal test of its approach. In the final study sample of just over 20,000 applications submitted between April 2018 and January 2022, roughly 5,000 borrowers were randomly approved for a digital loan and while nearly 4,400 comparable applicants were rejected.
This type of experiment, known as reject inference, is a well-established practice among credit risk modelers in the consumer credit industry, and can be used to determine if credit can responsibly be extended to a high-risk population.
Applicants’ mobile phone use data. Tala captured information like borrowers’ mobility—measured by the number of cell towers they passed and the cities they visited—and financial transactions, based on the amounts of money mentioned in their SMS messages, and the size and strength of their social networks, inferred from contact lists and texting behavior. This allowed researchers to build on computer science literature showing how mobile phone data can be used to measure financial well-being.
Borrowers’ loan amounts and interest rates. On average, approved borrowers received $36 for a roughly 28-day term loan with a one-time fee of about 15 percent.
Measuring social life and prosperity through mobile phones
Even with what may seem like high fees, researchers found borrowers' overall well-being improved—especially when the loans were used for business purposes. They can isolate the effect because both groups started the study ineligible for a loan—that is, the approved group was not initially better off or qualified enough to receive a loan in the first place.
Once granted the loan, the authors found that borrowers:
Were nearly 24 percent more likely to be employed or self-employed than the control group of rejected applicants.
Traveled to 9.4 percent more cities, suggesting boosted economic activity.
Sent 27 percent more text messages, an indicator of social connections and expanding networks.
Earned 21 percent more income each month, on a self-reported basis.
Spent 15 percent more per financial transaction.
What’s more, the default rate for the small loans—defined as the percentage of loans unpaid for one year after their due date—is “relatively low at 5 percent,” the authors write. In contrast, other large digital lenders in Africa have default rates as high as 27 percent.
Safeguards and building trust
Predatory lending remains a concern with any loan, but especially the relatively new digital sector in emerging countries with lower incomes. Tala used some safeguards to help ensure a low default rate—and to gain borrowers’ trust. These included:
Refraining from aggressive collection tactics.
Limiting borrowers to one loan at a time; new loans were only granted once the first is paid off.
Potential to earn larger loans the next time around by establishing a good track record with Tala.
Implications for any market
Kang’s findings may still apply in the wealthier countries, such as the US, where “unbanked” customers in rural areas may lack the credit histories and data needed to secure loans. However, applying a similar fintech model would require further research and careful consideration of regulatory and consumer protection frameworks, Kang says.
In the US, privacy and consumer protection laws that Kang calls both stringent and essential also make it more challenging to implement and study similar fintech models.
“Ultimately, understanding how to responsibly integrate alternative credit models into developed economies like the US could have profound implications for financial inclusion, helping millions of individuals access fair and affordable credit while maintaining the integrity and safety of the financial system,” Kang says.
Image by Ariana Cohen-Halberstam for HBSWK.
