A 2016 study states that as of 2015, about 40% of the US population above 18 years of age was excluded from the formal financial system. An important reason for this glaring fact is that more people than before are opting for alternative sources of lending, as they are unable to fulfill the requirements laid down by the traditional banking system. Most first-time borrowers dont have adequate financial information to back their case, which renders banks incapable of assessing their credit-worthiness accurately. And this is where the traditional credit scoring model fails.
So how do financial institutions lend to customers without any traditional credit information? Information gleaned from unconventional sources like telecom and utilities billing records, social media sites, and e-commerce sites could be the answer. In addition, financial institutions can use government records such as census information and other data available in the public domain, to enrich the data files of customers from the underbanked segment (also known as the thin credit file segment). This approach has been pioneered by fintech companies. Take the example of SoFi, a fintech company that started off by refinancing student loans of high-earning graduates with no traditional credit history, and now offers mortgages and personal loans to them. Such companies target millennials that not only lack a strong credit history, but also do not want to step into a bank to avail loans. A win-win situation for both parties.
Is alternate data enough?
To make up for the lack of relevant financial information, lenders need to ascertain the behavioral and financial patterns of their customers using various data sources. However, alternate data cannot be termed as a standalone solution since different regions across the globe are at different stages of evolution with respect to these unconventional data sources.
While designing a credit decision model, we need to consider the depth (volume of data, accuracy, percentage of acceptance, and so on) of the alternate data, and then determine appropriate weightages for the parameters selected for the credit model based on statistical analysis. To ensure accuracy of results, such a model must undergo continuous fine-tuning. Since the usability of alternate data has not been completely tested over a full credit cycle, it makes sense for lenders to leverage alternative data analysis in conjunction with traditional credit assessment, especially for customers who have inadequate credit history.
Risks of using alternate data
The 2016 white paper published by the US Treasury Department on the Opportunities and Challenges in Online Marketplace Lending highlights some challenges of using alternate data for credit assessment. Financial institutions can mistakenly deny credit to deserving customers (the ones with valid credit scores derived from traditional scoring methods) based on these aspects. The white paper points out that in certain cases lenders could use aspects such as gender and race as factors in their credit assessment models. In certain other instances, potential borrowers may not be aware of which factors will be analyzed to arrive at a credit decision. For instance, a borrower using multiple mobile numbers could provide a less frequently used number as the primary mobile number, without realizing that it could drive down the final credit score. This scenario can prove to be unfavorable for lenders as well. Incorrect use of alternate data will lead to inaccurate credit assessment of worthy borrowers, and the consequent denial of credit may make them switch to a competitor.
Moving toward greater transparency
We believe that educating customers on how alternate data is used for credit assessment can make the entire process fair and transparent. In cases where credit is denied based on this data, customers must be given an opportunity to provide clarifications. For example, a customer with multiple mobile connections should be given the option to review and change the primary mobile number to one that would possibly yield a higher score.
Alternate data sources are still evolving, and some of them may become less relevant or difficult to leverage over time. In the US, regulatory requirements have led alternate lenders to rethink their strategy on using social media as a data source. But this should not discourage lenders from using alternate data, as new and more effective sources will soon be available. Lenders are slowly, and certainly, waking up to the reality and potential of using alternate data to assess borrower credit-worthiness. According to this survey report, two-thirds of the traditional has helped them reach out to more customers. We believe that alternate data will become a useful addition to lenders’ credit assessment toolkit. What do you think?