Credit underwriting has undergone a quiet revolution in recent years. We’re seeing expansion of data sources beyond the ubiquitous credit score:

What’s happening is that industry participants are expanding the toolkit for determining which applicants are creditworthy. The impetus for this was the significant portion of American consumers who are thin file/no file, estimated at 45 million in the past decade (Consumer Financial Protection Bureau 2015). These consumers lack sufficient credit transaction history to generate a credit score. In high volume, mass decisioning environments, these consumers are out of luck. They are shut out of credit.

Separately, another cohort of consumers have poor credit scores of under 670, estimated at 57 million (35% of Americans with a credit score). That represents a lot of consumers who are interested in credit but are outside the risk appetite of mainstream lenders.

Combined, we’re talking about 102 million consumers falling outside traditional credit score-fueled underwriting standards. That is a sizable market! You can understand industry interest in adding new ways to identify good credit risk customers.

Brief overview of new data sources

The two big advances following credit score can be categorized as: “more financial facts”.

Recurring payments: Most of us have regular, monthly bills we have to pay. These are recurring obligations with a due date and dollar amount. Sounds like a credit obligation, no? There are a number of obligations that could be considered for credit underwriting, including: mobile phone, cable bill, gas and electric, water, Disney +, etc.

Bank data: Our checking and savings account are a wealth of interesting information. Spending patterns and savings levels can be analyzed. Insight into cash in-flows is there.

Work continues in the development of and access to those additional financial facts. Even so, attention has turned toward data that is predictive of repayment, but is not financial in nature.

Non financial data: The history of underwriting data is based on the concept of, “like data results predict like data performance”. You want to predict a financial indicator? You use financial data! But that’s not an ironclad rule.

In the insurance industry, people’s driving behavior is correlated with their credit reports? What? Here’s how leading auto insurer Nationwide describes their usage of credit data:

For the credit portion of your insurance score, these are the key factors:

• Payment history, including delinquencies or late payments
• Length of credit history
• Types of credit, such as credit cards and loans

Your likelihood of (a) having an accident, is reflected in your (b) credit data. As Nationwide notes, 92% of all insurers use credit data to calculate insurance premiums. Here is a real life example of data from one realm being used to predict behaviors in a different realm.

While financial facts have a finite set of possibilities, non financial data is relatively wide open. The journey to find good, relevant data that can predict credit payment behavior is underway.