To Limit Losses, Focus on Links Between Big Banks

Regulating large institutions with overlapping assets is a more efficient way to make the whole financial system safer

Based on the research of Daniel Mitchell and Stathis Tompaidis

As the 2008 financial crisis demonstrated, banks tend to rise and fall together. Before the crisis, their back-and-forth lending and borrowing boosted both assets and profits. But as housing prices tanked, one bank’s losses led to losses at others, dragging the economy into a painful global recession.

To prevent future financial contagions, regulators have tried to better understand the systemic risks posed by interconnected banks. New research from Texas McCombs offers a new approach for understanding — and reducing — these risks.

A standard approach has been to model banks’ financial links as a network. Regulators can then consider how shocks to the network might cascade from one bank to another.

But there’s a problem with the network model: limited data. Regulators don’t know how much each bank borrows and lends to each other bank. They have only overall figures for how much each bank lends to and borrows from all the others.

Daniel Mitchell, clinical assistant professor of information, risk, and operations management (IROM), proposes a workaround for the missing data. Rather than trying to unravel all the links between banks, he looks for a worst-case scenario for the whole system.

“In the face of imperfect data, we still want to understand how the economy reacts to interbank lending,” Mitchell says. “The goal of our paper is to see what types of borrowing and lending behaviors are the worst for the economy as a whole.”

Although it sounds counterintuitive, he calls such a scenario a “robust” network. The reason, he explains, is that being robust means being able to withstand adverse conditions.

“We want them to regulate to a network that is particularly adverse,” he says. “If they can regulate the worst-case network, then that regulation will also be effective on all networks that aren’t as bad.”

Biggest Banks, Biggest Risks

With Stathis Tompaidis, IROM professor, and McCombs Ph.D. student Feihong Hu, Mitchell looked at data from the fourth quarter of 2019, just before the financial shock of the COVID-19 pandemic.

With Federal Reserve data on the 353 largest bank holding companies — with assets over $3 billion — the researchers used an algorithm to identify the network structure likely to experience the worst losses. Then, they identified which links between banks were most critical and the amounts of the losses.

They found that a relatively small number of large banks with overlapping assets posed the greatest overall risk to the financial system. The 29 core banks held 79% of the assets loaned from one bank to another.

In the worst-case scenario, the researchers calculated, such interconnections could have led to $2.6 billion in overall losses during the second quarter of 2020, when the pandemic hit.

Focus for Regulators

Although worst-case scenarios sound ominous, the findings are reassuring and promise to improve regulatory efficiency, Mitchell says. They show that by focusing on large institutions, regulators can effectively monitor the total levels of risk in the financial system and increase its overall safety.

The framework can also guide specific regulatory strategies, Mitchell says, by helping regulators choose rules to rein in risk-taking by financial institutions. It can impose different standards on different banks, based on size, position in the network, or nature of their holdings.

Regulators don’t need detailed information on interbank lending to effectively regulate banks, he says. Rather, they can focus on the links that lead to the largest expected financial losses.

“You don’t need to know who owes whom what,” Mitchell says. “You just keep on regulating the bigger banks more, no matter what the interdependencies are between banks. If you have limited resources, the lesson is to weight them more toward the big banks.”

“Robust Financial Networks” is published online in Operations Research.

Story by Deborah Lynn Blumberg