One of the reasons why practitioners find the Theories of Disruptive Innovation to be so effective is their causal nature. Unlike correlation, which measures a set of attributes related to an action, causality is a much more powerful tool as it explains why something happens in a given circumstance. To illustrate that point, consider an important module of the banking system—credit models.
A majority of credit models today rely on data from the past to predict the future. For instance, when these models ask the question “What is the creditworthiness of the individual?” they assume the individual’s credit history is correlated with future credit behavior. Most of the time, however, consumers’ actions are a reflection of their present circumstance and have little correlation with past choices. What most credit models ignore are the circumstances that drive consumer behavior. In other words, they fail to consider causality.
As an example, consider a person who has filed for bankruptcy in the past and is trying to recover. Regardless of the circumstance, past data would effectively financially handicap the individual moving forward, in addition to causing a host of other problems. Other than having limited or no access to credit products, people with poor credit or no credit history pay higher insurance premiums. While renting an apartment or taking a utility connection, such consumers often need to provide a security deposit or letter of guarantee. Further, some employers are hesitant to hire employees with poor credit scores, believing that individuals with high amounts of debt could pose a higher risk of theft.
I had an opportunity to interview an individual who was in a similar situation. What I discovered about that individual’s creditworthiness was just the opposite of what correlated data would predict. Alex* dropped out of college in 2001 in search of a job. At that stage, he already had a student loan that he had to pay upon finding employment. However, due to the recession in 2001, he was unable to find a job for the next two years and ended up paying for his living expenses using credit card. By the time he found a job in 2003, his debt had reached a level where he could “barely keep his head above water”. So he consulted a bankruptcy lawyer who advised him that his best bet was to file for bankruptcy at that stage and move on from there.
As a result of the bankruptcy, Alex’s credit history was completely destroyed and he lost all access to any form of credit. Life was difficult after bankruptcy. He shared, “I was knocked down, I was trying to pick myself up and nobody was helping out”. At one point he feared it might even impact his employment when his bankruptcy appeared in the background check for his job. Fortunately, he was able to assuage his employers by explaining the circumstances that lead to his bankruptcy. Over time, Alex was able to pull himself out of debt and from then on displayed very good financial behavior. Since then he has never defaulted on his payments, never had to pay any interest on any credit cards, and he now has enough savings for an emergency and a sound financial plan for his future.
Looking at this case, a credit model would paint a picture of high risk (at least until he builds his credit back). On the contrary, when you understand the circumstance under which Alex had to file for bankruptcy, and the steps he’s taken to pull himself out of debt, he appears to be of least risk because he has had the opportunity to learn from his mistakes.
The challenge with current credit models is that they rely heavily on the notion that “success in the past guarantees success in the future”. While this theory is highly flawed, doing the causal analysis for the masses is not easy. It would mean looking beyond past data and spending time, effort and resources to uncover the real intent of borrowing for every consumer. Most financial institutions have lost the ability to tweak their models in order to better understand consumers. However, new online lenders are emerging with their own proprietary credit models, providing an enticing alternative to people with a range of credit history.
In Banking on Disruption: The rise of online lenders and the FinTech Fallout, I describe how many online lenders are using their proprietary credit models to look beyond what current credit scores have to say about a borrower. Because there are multiple data points taken into account when underwriting risk for a loan request other than the credit score, online lenders are able to better assess risk and offer more competitive rates to consumers along with comparable customer service to banks.
However, even online lenders have room for improvement with regards to understanding causality. For example, in The rise of online lenders and the FinTech Fallout, I provide an example of a borrower who sought help from an online lender in order to consolidate her debt and make progress towards a mortgage. By assessing her credit risk better and providing comparable customer service, the online lender succeeded in obtaining her business. However, it failed to fully identify progress the borrower was trying to make in that specific circumstance. Rather than seeing her mortgage as the larger opportunity, the lender only identified immediate debt consolidation as her goal, resulting in a missed opportunity for both the borrower and the lender.
Causality is a far more powerful tool than correlation, but it needs investment in terms of resources. That’s one reason why it’s difficult to create a causal model for the masses to predict creditworthiness. However, those who are able to invest in causality benefit from the returns that it has to offer. Because it leads to better understanding of the customer, customers can be targeted with the right products in the right context. If lenders are looking to differentiate themselves from the competition, they should invest in understanding causality.
*Alex’s name has been changed to protect his privacy.