For some consumers, the use of unconventional sources of information, or “alternative data,” to evaluate creditworthiness may be a way to increase access to credit or decrease the cost of credit. Alternative data includes information not typically found in core credit files of nationwide consumer reporting agencies and may indicate a likelihood of meeting obligations on time that a traditional credit history may not reflect.
In addition to the use of alternative data, increased computing power and the expanded use of machine learning can potentially identify relationships not otherwise discoverable through methods that have been traditionally used in credit scoring. As a result of these innovations, some consumers who now cannot obtain favorably priced credit may see increased credit access or lower borrowing costs.
In 2017, the Bureau issued a . The RFI noted that while the use of alternative data and models has significant potential benefits for consumers, it also presents certain significant potential risks. Commenters provided RFI responses about a range of topics, including current uses of alternative data, potential positive and negative consequences associated with its use, and the application of specific statutes and regulations. The RFI also indicated that the Bureau would begin to consider future activity to encourage responsible use of alternative data and lower unnecessary barriers impeding its use.
Later in 2017, the Bureau announced a . (the “NAL Recipient”), a company that uses alternative data and machine learning in making credit underwriting and pricing decisions. The NAL Recipient’s underwriting model uses traditional underwriting data and various categories of alternative data, including information related to borrowers’ education and employment history. The No-Action Letter references the application of the Equal Credit Opportunity Act (ECOA) and its implementing regulation, Regulation B, to the NAL Recipient’s use of alternative data and machine learning for its underwriting and pricing model. This No-Action Letter is specific to the facts and circumstances of the NAL Recipient and does not serve as an endorsement of the use of any particular variables or modeling techniques in credit underwriting and pricing. In addition, neither the No-Action Letter nor this blog post serve as an endorsement of the NAL Recipient or the products or services it offers.
As a condition for receiving its No-Action Letter, the NAL Recipient agreed to a model risk management and compliance plan that requires it to analyze and appropriately address risks to consumers, as well as assess the real-world impact of alternative data and machine learning. Pursuant to the No-Action Letter, the NAL Recipient provides the Bureau with information comparing outcomes from its underwriting and pricing model (tested model) against outcomes from a hypothetical model that uses traditional application and credit file variables and does not employ machine learning (traditional model). The NAL Recipient independently validated the traditional model through fair lending testing to ensure that it did not violate antidiscrimination laws. Over the last 22 months, the NAL Recipient worked to answer several key questions, including:
- Access to credit: whether the tested model’s use of alternative data and machine learning expands access to credit, including lower-priced credit, overall and for various applicant segments, compared to the traditional model
- Fair lending: whether the tested model’s underwriting or pricing outcomes result in greater disparities than the traditional model with respect to race, ethnicity, sex, or age, and if so, whether applicants in different protected class groups with similar model-predicted default risk actually default at the same rate
The NAL Recipient has agreed to allow the Bureau to share key highlights from simulations and analyses that it conducted pursuant to its model risk management and compliance plan; the simulations and analyses were not separately replicated by the Bureau. The following results provided by the NAL Recipient reflect the net effect of both the alternative data and the machine learning methodology used in the lender’s model as applied to the lender’s applicant pool.
The results provided from the access-to-credit comparisons show that the tested model approves 27% more applicants than the traditional model, and yields 16% lower average APRs for approved loans.
This reported expansion of credit access reflected in the results provided occurs across all tested race, ethnicity, and sex segments resulting in the tested model increasing acceptance rates by 23-29% and decreasing average APRs by 15-17%.
In many consumer segments, the results provided show that the tested model significantly expands access to credit compared to the traditional model. In particular, under the tested model, the results provided reflect that:
- "Near prime" consumers with FICO scores from 620 to 660 are approved approximately twice as frequently.
- Applicants under 25 years of age are 32% more likely to be approved.
- Consumers with incomes under $50,000 are 13% more likely to be approved.
With regard to fair lending testing, which compared the tested model with the traditional model, the approval rate and APR analysis results provided for minority, female, and 62 and older applicants show no disparities that require further fair lending analysis under the compliance plan.
For those concerned about access to affordable credit, more work remains to be done. The Bureau estimates that 26 million Americans are credit invisible, meaning they have no credit history with a nationwide consumer reporting agency. Another estimated 19 million consumers have a credit history that has gone stale, or is insufficient to produce a credit score under most scoring models. Without a sufficient credit history, consumers face barriers to accessing credit, or pay more for credit.
The Bureau encourages lenders to develop innovative means of increasing fair, equitable, and nondiscriminatory access to credit, particularly for credit invisibles and those whose credit history or lack thereof limits their credit access or increases their cost of credit, while maintaining a compliance management program that appropriately identifies and addresses risks of legal violations. Effective compliance management, as well as fair lending analysis results, may vary depending on a number of factors including the applicant pool and the lender’s specific use of data and models.
The Bureau remains committed to using all of the tools at its disposal under the Dodd-Frank Act to help address these important issues around access to credit. Toward that goal, the Bureau is currently reviewing comments to its proposed No-Action Letter, Trial Disclosure, and Product Sandbox policies.