Preserving Predictive Power
How to Mitigate Decline in Predictive Power
As you may recall from my last post, we examined the impact accommodation could have on credit risk scores. We saw that even though the majority of the scores remain unchanged, the loss of information affects the predictive power of certain segments — with the greatest impact to subprime.
So the question becomes, how do we mitigate a decline in predictive power?
The good news is there is significant strength in the credit file. It is unquestionably the primary and best asset for understanding creditworthiness. I want to be very clear about that.
One way to use the credit file to help ensure predictive power in the subprime segment is to optimize credit attributes based on performance.
Based on historical precedent and current research, we know that there are certain attributes that remain very strong in economic downturns, even with a high rate of loan accommodation: credit utilization, collections, certain inquiries and account balances are still very predictive. The same goes for age of trades, credit limits, number of trades and number of recent account openings. All are powerful.
With the high rate of loan accommodation in the current environment, delinquency- and satisfactory-related trade attributes that only measure the last three to six months of consumer behavior lose significant predictive power. Conversely, attributes that measure consumer behavior over the entire credit history or at least the last 24 months lose very little predictive power. “Always satisfactory” attributes are among the best attributes in economic downturns because they reflect consumers’ resilience.
My bottom line and best guidance here is: if you’re going to limit yourself to a credit only view, make sure you’re using the most predictive data in the file.
But these are unprecedented times. Every single day we are seeing something new and unexpected; so we have to think and act differently. I urge you to expand beyond traditional models and habits.
Optimize Predictiveness with Employment and Income Data
As we dive into approaches that will optimize credit predictiveness in this economy, we see great opportunity with alternative data – specifically employment and income data.
By leveraging income and employment data in addition to credit data, our customers are able to increase the predictive power of risk solutions.
For instance, a data-driven approach leveraging the credit file and supplementing it with income and employment data can recapture 42% of performance loss caused by loan accommodation and forbearance in the subprime segment.
Traditional credit data typically reflects a consumer’s willingness to pay as well as a consumer’s ability to pay. However, when consumers go through a major life event, like a change in job status or income, credit data might not accurately reflect ability to pay.
Given the unprecedented number of Americans who have become unemployed, changed jobs, or had changes in income, supplementing credit data with income and employment data becomes even more important in this economic cycle.
Income and employment data are highly predictive of a consumer’s ability to repay debts. In a recent study of retail credit cards, we found that consumers with minimum job tenures of 18 months or more across their entire work histories are half as likely to become delinquent as consumers with minimum tenures less than 18 months.
We also found that consumers with annual incomes of $36,000 or less are 2.5 times as likely to become delinquent as consumers with incomes above $36,000.
Many lenders use income and employment data to approve more consumers without taking on more risk. In one scenario, we were able to identify a large segment of higher-income consumers with subprime credit scores who perform more like near-prime consumers than subprime. We were also able to identify a large segment of near-prime consumers with higher incomes who perform more like prime consumers than near prime.
So what do these employment and income insights mean for lenders? Put simply, they remind us that while the credit file is a powerful tool, income and employment data can be used to increase the predictive power of credit data, even in normal times, but especially now.
Understand the Individual Stories
All of us are looking at the macroeconomic trends ̶ monthly, weekly, maybe even daily in some cases. We absolutely need to understand the big picture.
We also need to understand the individual stories. To build a resilient business and help people, I suggest leveraging credit data and alternative data to better understand each individual segment and each individual credit story. I believe it will help companies to better serve their customers while managing risk appropriately.
The approach will also help consumers get the credit they need when they need it based on their unique situation. Not an aggregate “surface” view. I think that is a win-win.
Don’t miss the other posts in my series:
If you have any questions, please visit our COVID-19 website, and reach out and let us know what you’d like to hear about in future blog installments.
Recommended For You
A More Inclusive Credit Scoring Model Economists refer to the U.S. recovery as ‘K-shaped’, meaning that one segment of the […]
In this recap of our June 24 webinar, “Market Pulse: Accelerating Access to Credit,” Chris Wheat, Co-President at JPMorgan Chase […]
In this recap of our June 3 webinar, “Market Pulse: The Post-Pandemic Small Business Economy,” Evan Leaphart, Founder and CEO […]
As we emerge from the pandemic, small businesses are integral to the health of the post-COVID economy. For our June […]