How does the Fed’s use of data analytics in monitoring the banking industry help consumers?

The kind of work our team performs supports the safety and soundness of the banking sector, and ultimately, the safety of consumers’ deposits in banks. In short, we analyze many different types of data to help identify banks that might be engaging in riskier activities that could lead to increased losses, so that we can pay more attention to those firms through monitoring we call supervision. Likewise, we can use data to identify banks that appear to be engaged in low-risk activities, and streamline our supervisory monitoring of these firms.

The Federal Reserve is one of three organizations that monitors banks to help maintain stability for consumers. All 12 Federal Reserve Banks across the country use data and models to tailor supervision to the risk of each bank. This approach—called Bank Exams Tailored to Risk (BETR)—helps to ensure we have effective supervision with the least amount of regulatory burden for institutions. We don’t want to cause undue burden; we want banks to be able to focus on serving their customers in a safe and sound manner.

We’ve been using forward-looking, data-driven models in the bank supervisory process for several years now, but the Federal Reserve Board enhanced and clarified this process in 2019 through the BETR initiative. Prior to BETR, we supervised banks according to risk, but the risk was based more on rules derived from the expert judgment of examiners, or staff who evaluate a bank’s financial statements, management, risk level of loans, and other activities to ensure financial stability.

We use carefully validated statistical models along with examiner judgment to determine how to classify banks according to risk. There are various dimensions of risk for which we examine firms—for example, the risk that banks might not get repaid for loans they made or the risk that they don’t govern themselves appropriately—and we have a model for each of those dimensions.

As part of this process, we look to see if a model would have been successful at predicting which banks had adverse outcomes during unfavorable market conditions in the past. The idea is that if there is some kind of shock in the world—like the COVID-19 pandemic—we should be able to look to our models to see which of our firms are considered “high risk” and are more likely to be affected by the event.

Using data-driven monitoring presents opportunities and challenges. When the pandemic first started spreading in the US and we paused in-person bank examinations, we relied a lot more on data and models to assess risk because examiners weren’t going onsite at banks. But we also had to modify our models and consider how the pandemic would affect them. For instance, some banks that processed a large number of Paycheck Protection Program (PPP) loans exhibited significant loan growth, but this growth didn’t imply the same amount of risk as traditional loan growth. So we had to make sure we considered this activity when interpreting model output. So far, I think the banking sector has been very resilient through the pandemic.

It’s really satisfying to see the use of data and models to predict risk in order to improve the safety and soundness of the financial system in an actionable way. Another exciting part of this work is that there are always opportunities to innovate and improve as data become more plentiful and as mathematical and statistical methods advance, and it’s great to see the Federal Reserve at the forefront of this work.