First-of-its-kind data on millions of loans in East Africa suggest it is time for funders to rethink how they support the development of digital credit markets. The data show that there needs to be a greater emphasis on consumer protection.
In recent years, many in the financial inclusion community have supported digital credit because they see its potential to help unbanked or underbanked customers meet their short-term household or business liquidity needs. Others have cautioned that digital credit may be just a new iteration of consumer credit that could lead to risky credit booms. For years the data didn’t exist to give us a clear picture of market dynamics and risks. But CGAP has now gathered and analyzed phone survey data from over 1,100 digital borrowers from Kenya and 1,000 borrowers from Tanzania. We have also reviewed transactional and demographic data associated with over 20 million digital loans (with an average loan size below $15) disbursed over a 23-month period in Tanzania.
Both the demand- and supply-side data show that. The data suggest a market slowdown and a greater focus on consumer protection would be prudent to avoid a credit bubble and to ensure digital credit markets develop in a way that improves the lives of low-income consumers.
High delinquency and default rates, especially among the poor
About 12 percent and 31 percent, respectively, say they have defaulted. Additionally, supply-side data of digital credit transactions from Tanzania show that 17 percent of the loans granted in the sample period were in default, and that at the end of the sample period, 85 percent of active loans had not been paid within 90 days. These would be high percentages in any market, but they are more concerning in a market that targets unserved and underserved customers. Indeed, the transactional data show that Tanzania’s poorest and most rural regions have the highest late repayment and default rates.
Who’s at greatest risk of repaying late or defaulting? The survey data from Kenya and Tanzania and provider data from Tanzania show that men and women repay at similar rates, but most people struggling to repay are men simply because most borrowers are men. The transaction data show that borrowers under the age of 25 have higher-than-average default rates even though they take smaller loans.
Interestingly, the transactional data from Tanzania also show that early morning borrowers are the most likely to repay on time. These may be informal traders who stock up in the morning and turn over inventory quickly at high margin, as observed in Kenya.
Borrowers who take out loans after business hours, especially at 1 or 2 a.m., are the most likely to default — likely indicating late-night consumption purposes. These data reveal a worrisome side of digital credit that, at best, may help borrowers to smooth consumption but at a high cost and, at worst, may tempt borrowers with easy-to-access credit that they struggle to repay.
Further, the transaction data show that first-time borrowers are much more likely to default, which may reflect lax credit screening procedures. This can have potentially long-lasting negative repercussions when these borrowers are reported to the credit bureau.
Most borrowers are using digital credit for consumption
Many in the financial inclusion community have looked to digital credit as a means of helping small, often informal, enterprises manage daily cash-flow needs or as a way for households to obtain emergency liquidity for things like medical emergencies. However, our, including ordinary household needs (about 36 percent in both countries), airtime (15 percent in Kenya, 37 percent in Tanzania) and personal or household goods (10 percent in Kenya, 22 percent in Tanzania). These are discretionary consumption activities, not the business or emergency needs many had hoped digital credit would be used for.
Only about 33 percent of borrowers report using digital credit for business purposes, and less than 10 percent use it for emergencies (though because money is fungible, loans taken for one purpose, such as consumption, could have additional effects, such as freeing up money for a business expense). Wage employees are among the most likely to use digital credit to meet day-to-day household needs, which could indicate a payday loan type of function in which digital credit provides funds while borrowers are waiting for their next paycheck. Given the evidence from other markets of the high consumer risks of payday loans, this should give pause to donors that are funding digital credit.
Further, the phone surveys show that. Any benefits to consumption smoothing could be counteracted when the borrower reduces consumption to repay.
The survey data also show that 16 percent of digital borrowers in Kenya and 4 percent in Tanzania had to borrow more money to pay off an existing loan. Similarly, the transactional data in Tanzania show high rates of debt cycling, in which persistently late payers go back to a lender for high-cost, short-term loans with high penalty fees that they continue to have difficulty repaying.
Confusing loan terms and conditions are associated with difficulties repaying
Lack of transparency in loan terms and conditions appears to be one factor contributing to these borrowing patterns and high rates of late repayment and default. A significant percentage of digital borrowers in Kenya (19 percent) and Tanzania (27 percent) say they did not fully understand the costs and fees associated with their loans, incurred unexpected fees or had a lender unexpectedly withdraw money from their accounts. Lack of transparency makes it harder for customers to make good borrowing decisions, which in turn affects their ability to repay debts. In the survey, poor transparency was correlated with higher delinquency and default rates (though correlation does not indicate causation).
What does this mean for funders?
Even though digital loans are low value, they may represent a significant share of a poor customer’s income, and repayment struggles may harm consumers. Overall, the use of high-cost, short-term credit primarily for consumption coupled with high rates of late repayments and defaults suggest that funders should take a more cautious approach to the development of digital credit markets — and perhaps stop providing grants or concessional funding terms for this segment of products.
More specifically, the free and subsidized funding currently used to expand digital credit products to unserved and underserved customer segments would be better used helping regulators monitor their markets, identify opportunities and risk and promote responsible market development. One way to do this would be to fund and assist regulators with gathering and analyzing data on digital credit at the customer, provider and market levels. More comprehensive and granular data would help regulators — as well as providers and funders — better assess the opportunities and consumer risks in digital credit.
Improved data gathering need not be cost prohibitive. CGAP’s research in Tanzania shows that affordable phone surveys can provide useful data that are remarkably consistent with provider data. Digital lenders’ transactional and demographic data should be collectable since lenders regularly assess them when calculating and reporting on key performance indicators. However, additional investment may be needed to ensure the consistency, integrity and reliability of the data.
At a market level, it will be important to strengthen credit reporting systems and require information reporting from all sources of credit, including digital lenders, to improve the accuracy of credit assessments. These efforts should consider whether prevailing digital credit screening models are strong enough and whether rules are needed to ensure first-time borrowers are not unfairly listed. This could include rules on reckless lending or suitability requirements for digital lenders.
Donors and investors can play an important role in the next phase of digital credit’s market development. This phase should see greater emphasis on assisting regulators to regularly gather and analyze data and act to address key warning signs that are already emerging around transparency, suitability and responsible lending practices.
Excellent point. You get to the heart of the issue here: why are these models doing such a poor job of filtering on first loans, and should greater scrutiny be placed on these models? Some lenders have lower fees because of better data to score or subsidized channel use via their MNO partners. Is opening those up more a first step—if not the only step—to reduce the high fees that come with the high defaults? It would also be better if lenders didn’t charge the same high fee on your fifth loan after paying back the first four, as many still do in this sector.
The obvious question that arises from this is how any financial services company can survive on ~50% on-time payment, and ~20% chronic delinquency. As you know, any regulated financial institution would have been in receivership well before getting to anywhere near here.
The large part of the answer is in the monthly "service charge," which are 10-15% a month, and easily add up to 100%+ in APR terms. Good borrowers are cross-subsidizing and keeping these portfolios in near-perpetual-distress afloat. Which adds even more weight to the need for stringent consumer protection you rightly call for.
ZA, that is an important question to consider indeed. In mature, traditional credit markets we have already seen business models that relied on a combination of high delinquency rates and high fees/charges --leaving vulnerable borrowers in financial problems, and feeding the fire of consumer credit crises. These previous experiences remind us that once a credit market leaves its early development stage, and providers have more information available and better understanding of the market, then credit screening, assessment and provision models shall be reassessed and refined to ensure responsible long-term market developments (that do not rely on high fees/charges). Recently, the Basel Committee guidance on financial inclusion indicated that authorities shall pay close attention to these types of developments. Authorities shall make sure that all needed adjustments in business practices are made for the benefit of all borrowers. This is a key reason why gathering granular data is so important for market monitoring from a consumer protection standpoint, especially for relatively new products like digital credit. Rich data and adequate monitoring can enable more effective consumer protection frameworks that aim to generate positive customer outcomes and minimize customer harm.
Most of the digital credit providers have been using what they call 'alternative data' as a basis for determining risk of a borrower. However, from the study it is clear that more traditional risk profiling based on credit history and debt exposure is crucial in mitigating default and over indebtedness. As pointed out, it is imperative that credit information is shared by the digital lenders through the already existing national credit information reporting mechanisms in these countries. It may require some policy and regulatory changes to bring about this next evolution of digital credit and avoid the predatory lending that is now starting to emerge.
It is too late to mourn the delinquency of digital credit . This is the consequence of blind digitalized financial inclusion of the poor focusing on credit delivery without talking cognizance of the real needs of the poor and their level of digital numeracy in the informal sector.
It is time to seriously think of preparedness for inclusion of the poor through oral financial practices and provision of integrated micro financial products and services & capacity building besides credit . Putting the last (credit ) first is harmful in poverty sector.
Two main reasons among others for credit delinquency include mis utilisation of the credit and using for closing earlier loan . In both cases the very purposes of credit be it digitised or non digitalized ie, income generation is absent .
Mere sequestered micro credit either digitised or non digitised without integrating other needs of the poor like insurance, savings, pension, capacity building etc is harmful.
Researchers in Micro finance and CGAP need to focus on the above conundrums being faced in Micro finance industry
Thanks for both of your comments, Dr. Rengarajan,
You may have read our follow-up blog post on suggestions for policy makers (https://www.cgap.org/blog/how-regulators-can-foster-more-responsible-di…), including the importance of assessing financial product governance processes and procedures, which should consider client needs and constraints and customer outcomes. It's concerning to see the type of debt traps that some clients seem to enter into with digital credit, as the slide deck with provider data from Tanzania shows (https://www.cgap.org/research/slide-deck/digital-credit-market-monitori…). At the same time, that slide deck shows that some clients are continuously improving their use of digital credit, with better conditions in each subsequent loan, and with justifiable loan purposes... We fully agree with the premise that credit, let alone digital credit, is not that be-all and end-all of financial inclusion, and it's crucial to see how financial products and services can actually generate positive customer outcomes.
Hi - I really enjoyed the article. It is the most useful piece of analysis I have seen, when it comes to digital credit. I had a few questions, if you don't mind me bothering you:
Is the data available to be crunched some more?
If the data isn't public, could I ask you - I didn't see in the article whether there was a question/discussion about their total combined outstanding debt, both in terms of an amount and the number of loans and what are the sources of the other debt. I would have also loved to see, if possible, some debt to income ratios - ie were those surveyed asked about their income levels? Lastly, how do these results square with the credit bureau data? And, just a curiosity - are the late night borrowers more likely to be male or female? Lastly, how willing did they perceive their lenders to be to discuss their problems with repayment and were they aware of the risk of potentially being blacklisted? And, last, I promise - were they asked about any benefits of the various financial literacy initiatives that might be taking place around them?
Thank you, Natasa!
For different reasons, we have not made the data publicly available. Actually, we committed to erase the supply-side data (that also included demographic data) for confidentiality purposes.
In terms of your specific questions, some answers can be found in our working paper (https://www.cgap.org/research/publication/digital-credit-revolution-ins… ), like the number of sources of digital and conventional loans that digital borrowers had at the time of the SMS surveys in Kenya and Tanzania (p. 28), which sources were used less when clients gained access to digital credit (p. 29), and the percentage of digital borrowers who contacted customer service to complaint on credit bureau information -the main complain reason in Tanzania (p.26). Another very important finding was that the survey data showed consistent results with the transactional data obtained from digital lenders. You can see that the levels of delinquency and default observed in provider data (https://www.cgap.org/research/slide-deck/digital-credit-market-monitori…) were at roughly the same high levels as those self-reported by consumers. In terms of late night borrowers, who are more likely to not repay on time, they follow the same pattern of all the Tanzanian digital borrowers --they are more likely to be male.