How to rethink credit scoring?
Since 1989, credit scoring has become an integral part of the consumer lending industry, marking one of the most significant innovations in the field. Yet, some data problems underlying the credit scores need to be addressed.
According to Fair Isaac Corporation, creators of the FICO score, the five categories of data that drive their credit score are payment history, amounts owed, length of history, mix of credit and new credit.
Here’s a breakdown of the FICO score components and their respective contribution percentages:
These categories are well-known to consumers and professionals, and they directly influence consumers behavior. For example, consumers may avoid applying for new credit cards to maintain their score.
However, these factors are all being used to measure something else:
- Payment History measures Willingness – to – repay
- Amounts Owed measures Ability-to- repay
- Length of Credit History and Mix of Credit measure Financial Responsibility
- New Credit measures Financial Health
These factors are called proxy variables. A proxy variable is a variable used to measure something that’s difficult to measure. Proxy variables are beneficial in risk modeling, but if the correlation with the underlying behavior is weak, it can lead to significant distortions and unintended consequences. Banks and credit unions need to be aware of these imperfections and the recent advancements in credit attributes to counter these unintended consequences.
Payment history is a proxy for a consumer’s willingness-to-repay. This is a factor in risk modeling because past payments are a strong indication that a consumer intends to continue paying. But credit bureau data doesn’t actually track customer payments, it tracks whether an account is in ‘satisfactory’ standing (e.g. not delinquent). This creates situations that exclude certain groups.
For example, many delinquent customers are unable to repay their entire past due balance upfront. If the consumer is making partial payments, which are applied to late fees and interest before the principal payment, it can take many months before they can become current. Even though they’re making payments they’re still delinquent. Therefore we’re not actually measuring their willingness to repay. Truv addresses this issue by measuring payment behavior directly.
This proxy also doesn’t represent consumers who are new to credit and those who are considered subprime. In both cases credit bureaus struggle to measure their ability to repay because the records may not exist or the records are not traditional credit repayments.
Amounts owed is a proxy for the ability-to-repay. Even if a consumer is willing-to-repay, they need income or cash flow to repay a loan. This is why many lenders verify income and key expense ratios as part of their underwriting.
Ability-to-repay is a function of 1) the customer’s net cashflow after expenses, and 2) the amounts owed. Since income and cash flow are not tracked or reported to the credit bureaus, ability-to-repay is proxied solely with amounts owed.
This doesn’t reflect an individual’s actual income and spending behaviors. For a given level of debt, all else equal, the customer with the higher income should be less risky. Or more precisely, the customer with the higher cash flow should be less risky. Truv addresses this issue by directly measuring all of the components of ability-to-repay, including income, expenses and net cash flow.
Length of Credit History and Types of Credit
These categories proxy financial responsibility. The greater the consumer’s financial responsibility, the more likely they are to maintain their financial health and continue being able to repay their debts over time.
Length of Credit History is clearly biased against consumers with short credit histories that are otherwise financially responsible.
Types of credit refers to certain types of loans that are more correlated with responsibility. For example, credit cards have a lot of flexibility and features that can be abused if not used wisely, so someone who is using them properly is demonstrating financial responsibility. Mortgages indicate that an individual has sufficient savings, financial stability and long term commitment, which are all signs of financial responsibility. These are problematic measurements. Studies have shown that millennials are using less revolving credit and deferring home purchases, which will make them appear riskier using this proxy, regardless of their true financial responsibility.
Truv addresses these issues by taking a more direct approach to measuring financial responsibility. We have attributes which track whether a consumer is able to balance their checkbook, whether they make pro-active transfers to prevent OD/NSF fees, whether they maintain sufficient financial buffers and whether they have consistency in their financial behaviors.
New credit and credit inquiries indicate a customer’s financial health, or potential financial distress. For example, someone who is opening new lines of credit could be trying to offset loss of income or increasing costs which would indicate distress. However, distinguishing new credit arising from distress versus planned activity can be challenging.
Truv addresses this by measuring financial health and distress directly, considering changes in income, expenses, and net cash flow.
Truv is a Better Way to Manage Risk
It’s easy to criticize the current credit scoring method, which is why we’ve also come up with an alternative.
Truv uses various attributes to build a more powerful risk model which results in a more accurate assessment of risk. For one of our clients, a traditional FICO score provided a Kolmogorov-Smirnov (KS) statistic of only 6%. That is not a typo. Traditional credit scores had essentially no bearing on loan charge-off performance for their consumer base. In comparison, an origination risk model built using Truv Financial Behaviors attributes delivered a KS statistic of 45%, or 7.5X, which allowed them to have much greater precision in their underwriting strategy.
The current credit scoring approach sits atop a foundation of faulty data. This data distorts applications for certain consumer groups. Truv is working to level this playing field by providing more accurate data and better modeling. Check out Truv here to learn more about how we’re advancing equality in the lending industry.