A shuttered First Republic Bank branch in San Francisco on June 6, 2023, about a month after the FDIC announced its closure and sale to JPMorgan Chase. (Credit: JHVEPhoto)
BLOOMINGTON, Ind. – Mixtures of high interest rates and major losses led to the sudden collapses of Silicon Valley Bank and First Republic Bank two years ago, and some experts worry that rising inflation and slow economic growth could be a concern for other banks.
While stress tests may assess bank performance under some predicted future shock, new research from the Indiana University Kelley School of Business proposes policies that could resemble more of an early warning system that also could help avert panic among borrowers and investors.
The International Monetary Fund and World Bank regularly conduct forward-looking debt sustainability analysis, aiming to provide the market with an early warning of sovereign debt distress. After the Great Recession, bank supervisors adopted similarly focused stress tests and publicly disclosed whether a bank could sustain that adverse shock.
Deepal Basak
“Our proposed timely disaster alert resembles an early warning system used in practice,” said Deepal Basak, assistant professor of business economics and public policy at the Kelley School. “When designed appropriately, they can be remarkably effective under a reasonably general setup.”
Basak and Zhen Zhou, associate professor of financial economics at Tsinghua University, are co-authors of the accepted paper, “Panics and Warnings,” forthcoming in the Journal of Political Economy, published by the University of Chicago Press.
Their paper studies how to design an information disclosure policy that averts such panic.
“No policy can eliminate panic under adversarial selection, which occurs when economic agents play the worst equilibrium,” Basak said. “The core challenge is that if a creditor or investor believes others might run, he will run too.”
A common assumption in this literature is that the problem is static. Introducing a ‘time dimension’ to the issue, which is natural in many applications, enables the regulator to design dynamic information disclosure policies that can eliminate panic even when the agents play the worst equilibrium.
Basak, who has been at the Kelley School since 2020, researches how strategic decisions create market inefficiencies and how practical solutions can lead to improved conditions, including reduced market panic. He teaches game theory to undergraduates and MBAs at Kelley.
In the paper, he and Zhou studied optimal adversarial information design in a dynamic regime change game. Agents decide when and whether to attack in several scenarios such as creditors deciding when to withdraw funds from a bank, or investors choosing when to move their money elsewhere.
As they point out in their paper, a disaster may happen even when conditions do not suggest one is warranted. For example, creditors may anticipate a bank will face a negative future shock, such as borrowers in a certain industry defaulting on loans, and they may begin withdrawing funds. This action can weaken the bank and cause a bank run – an example of a panic — even though the shock may not be severe when it arrives.
The authors introduce a disclosure policy called “timely disaster alert.” This policy promptly informs stakeholders about any impending disasters. Essentially, the regulator communicates to the agents: “Do not panic, just wait a moment, and I will notify you if a disaster occurs.” A “timely disaster alert” makes waiting for and following the alert the unique rationalizable strategy, which eliminates panic, they found.
In their paper, Basak and Zhou demonstrate that no rational agent will wait and then ignore the alert; they will either wait and heed the alert or opt not to wait at all. Since agents who choose to wait will always follow the alert, waiting for it becomes valuable for any agent who doubts the correctness of their actions. If the cost of delay is continuous and the alert is timely, even in the worst-case scenario, an agent will disregard their private signals and choose to wait for and follow the alert. This eliminates panic.