Fed Opens Up Alternative Data – More Credit, More Algorithms, More Regulation

A Dec. 4 joint statement released by the Federal Reserve Board, the Consumer Financial Protection Bureau (CFPB), the Federal Deposit Insurance Corporation (FDIC), the National Credit Union Administration (NCUA) and the Comptroller of the Currency (OCC), highlighted the importance of consumer protections in using alternative data (such as cash flow etc.) across a wide range of banking operations like credit underwriting, fraud detection, marketing, pricing, servicing, and account management.

The agencies acknowledged that modeling approaches using these alternative data sources will both enhance the credit decision process bringing in underserved consumers and unlock pricing, offerings and repayment benefits for existing consumers. 

Despite the potential benefits, the agencies also caution against using this new data in ways that are inconsistent with the current regulatory consumer protection framework of fair lending and fair credit reporting laws.

What is “Alternative Data”?

An example of alternative data that can be used is cash flow information calculated from borrowers’ income and expenses. This improves predictions based solely on traditional data points on the ability of the borrower to repay the loan. However, consumers have to give permission to the underwriter for use of this data and be able to request disclosures on its use. In this case, using alternative data allows consumers with irregular incomes like gig economy workers to better access credit services. 

While the agencies did not provide guidance on examples of alternative data that should be avoided (e.g. social media etc.), they strongly advocated a ‘responsible use’ of any new data being considered. 

Companies should thoroughly assess all alternative data against the existing regulations. This requires a sound compliance management process that appropriately factors the sensitivity of the data to protect consumers against risks.

What does this mean for businesses?

These guidelines create a potential boon for financial services who have been competing for the same traditional credit restricted pool of consumers by unlocking access to new, often proprietary sources. On the other hand, leveraging new data sources at scale will likely warrant new techniques and algorithms for processing that data. Machine learning algorithms are an obvious choice as the size and variety of data scales. Indeed, technology forward financial services enterprises have already been adopting machine learning practices to solve these challenges and to better compete for a new pool of consumers. This joint statement empowers the rest of the financial services firms to use similar approaches. 

Enterprises scaling their machine learning operations to incorporate alternative data should address associated AI risks (e.g. explaining adverse action notices, bias, unfairness, etc). A robust AI governance framework will ensure they are in compliance with the spirit of the statement. 

When explanations are integrated in the AI workflow from data selection, model development, and validation to compliance and monitoring, it addresses the potential gaps enterprises will face to ensure consumers are protected and treated fairly.

The opening of new data sources for use by lenders is a great step forward towards democratizing access to credit  to more consumers while empowering the entire financial services and broader underwriting industries to build better solutions.