Recent research on the state of the digital consumer credit market in Kenya has documented breakneck growth in mobile lenders and alarming levels of defaults and delinquency. These findings, coupled with anecdotal reports of aggressive sales tactics, abusive debt collections, and unacceptable levels of debt stress, raise an enormous red flag regarding digital consumer credit’s potential for financial inclusion.
The promise of digital credit is to make people with no documented financial history visible to the formal financial system, helping low-income people gain access to loans and other financial products. Microcredit started out with a similar promise to help people move out of poverty, but the microfinance industry learned the hard way that good intentions are insufficient. After microcredit-fueled, overindebtedness crises spiraled out of control, the industry introduced protective measures and formal regulation followed in many countries. The red flags appearing in Kenya’s digital credit market are familiar, but a host of new players beyond traditional banks and microfinance providers may not recognize the signs of a looming crisis.
New Entrants, New Credit Models
New tech players are flooding into markets like Kenya to offer scalable models for faster and more accurate loan products backed by big data. Mobile app-based lenders use nontraditional data to provide near-instant credit decisions for loans ranging from the equivalent of $5 to $100. The algorithms at the core of these products rely on unverified, self-reported user information and data scraped from users’ mobile phones, including call logs, contacts, location, social media activity, and text messages. An inherent flaw in these models is that they are built on data reflecting a consumer’s willingness to repay rather than their ability to repay. This is a crucial distinction representing the difference between a client that is willing to skip meals or borrow from other lenders to repay the original loan and a client that can afford the loan without experiencing such debt stress. Repayment rates may appear acceptable, but this flaw in the model masks a brewing overindebtedness crisis as repayment rates don’t account for loan stacking or abusive debt collection practices.
By the Numbers: Kenya’s Digital Credit Challenges
86
2.2
1/2
26.4
Providers have an important role to play in mitigating an overindebtedness crisis in Kenya by instituting a structure for algorithm governance and internal processes that embed consumer protection in their products and culture. The scalability of app-based lenders depends on “low touch” or “no touch” models that use experimental lending phases to collect data to refine their credit decision algorithms. (Experimental lending is the learning phase of the algorithm when loans are distributed to refine the model based on real repayment behavior.) They don’t verify client-reported data or other gathered data against actual income or cash flows. The algorithms are optimized to predict repayment accurately, but don’t factor debt stress into the model ex ante.
Algorithms are optimized to predict repayment accurately, but don’t factor in debt stress ex ante.
Yet there are many internal measures that app-based lenders can take to better protect clients without fundamentally changing their business model. Providers should have written policies and procedures to assess a client’s repayment capacity. Ex ante, they should document the rationale for their models, including the types of data and variables used and a justification for relying on those factors. They should also create an institutional definition of overindebtedness and collect data to do ex post analysis on client outcomes or debt stress. An internal algorithm governance structure doesn’t mean that companies reveal the IP at the core of their business, but that they have internal checks. The structure should bring together representatives of different departments, other than the unit that developed the algorithm, for a regular review of the model’s performance in the context of broader market developments. The Smart Campaign’s Consumer Protection Standards for Digital Credit document a number of these practices, as well as more traditional measures such as setting policies against aggressive sales and debt collection.
With any algorithm we should be asking what is the target variable of the model? What is it optimized to do? There are many talented data scientists and product developers working at these companies that can design sticky product experiences and models that accurately predict whether the company will be repaid. Let’s encourage them to factor client stress into that model. The data collected by these lenders could likely be sufficient to detect patterns of loan stacking and even develop other proxies for debt stress.
Market-Level Action
Beyond improving provider practices, regulatory action is necessary to improve monitoring and data sharing for the Kenyan credit market. Some app-based lenders report data to credit bureaus, but they often don’t use credit bureau information in their loan decisions, sometimes because their clients lack credit histories but also due to technology limitations. The technology systems and speed of data processing are misaligned between providers and credit bureaus, with mobile credit decisions occurring in minutes while data requests from the credit bureau can take a lengthier period. Credit bureaus must be upgraded to accommodate real-time data reporting by introducing a common API that can automatically retrieve data on clients’ outstanding debt obligations, and regulation must mandate that providers do so. A limit for blacklisting individuals should also be set so that hundreds of thousands of defaulting customers aren’t further barred from accessing formal financial products.
Bringing mobile providers under the same regulatory framework will promote greater harmonization of practices, transparency of pricing and interest rates, and useful data reporting.
The Kenyan government should also introduce national information campaigns to warn consumers about risks and inform them of their rights. MSC analysis (in partnership with CFI) showed that many Kenyan borrowers had a low understanding of pricing, terms and conditions, and use of client data in digital loan products. The government should support awareness campaigns to help customers understand the terms and conditions of their loans and discern between good and bad actors in the market. Regulatory requirements on clear and concise disclosure principles are needed to ensure providers use simple language and provide information that can be compared consistently across channels and companies in a timely manner. Bringing mobile providers under the same regulatory framework will promote greater harmonization of practices, transparency of pricing and interest rates, as well as useful data reporting.
Today Kenya, Tomorrow…Everywhere
What we’re seeing in Kenya’s digital credit market reflects a global trend of consumer rights under threat from unchecked automation. A misunderstanding of debt stress as microfinance took off in developing countries trapped poor customers in damaging debt cycles and threatened market stability. Algorithm-based, consumer credit products that aren’t carefully evaluated from the perspective of debt stress could replicate the earlier mistakes of the microfinance industry at a greater pace and scale before the problem is even visible. As these products continue to explode in Kenya and rapidly expand to new markets like the Philippines, India, and Indonesia, it’s time to wake up to the consumer protection risks and take action.
Authors
