Tracing Bank Runs in Real Time: Insights from High-Frequency Data and Implications for Policy
At the 2024 Carey Finance Conference, I had the pleasure of discussing an insightful paper by Cipriani, Eisenbach, and Kovner, titled “Tracing Bank Runs in Real Time.” The paper leverages a unique dataset of high-frequency deposit flows to uncover key details of the banking turmoil following the failure of Silicon Valley Bank (SVB) in March 2023. It introduces a new measure to identify banks experiencing runs and provides significant evidence on the behavior of depositors, particularly at smaller banks.
Here, I want to summarize the key aspects of the paper and expand on the implications for financial stability and policy, particularly regarding liquidity regulation and the coordination challenges that arise during bank runs.
Key Insights from the Paper
The paper focuses on understanding what triggered bank runs in 2023 using a novel high-frequency data approach. This allows the authors to trace liquidity flows at the transaction level across U.S. banks during the banking stress that followed SVB’s collapse. The paper highlights the following key points:
Identification of Bank Runs: The authors introduce a new measure for bank runs, based on deposit outflows and represented by a z-score. This approach allowed them to identify 22 banks that experienced significant outflows on March 10 or March 13, 2023, but only two of these banks ultimately failed (SVB and Signature Bank).
Wholesale vs. Retail Depositors: The data shows that the bank runs in 2023 were largely driven by wholesale depositors, primarily larger institutional players, rather than retail depositors. The wholesale runs were concentrated over just a few days, and most of the fleeing deposits went to larger banks perceived as safer, highlighting a flight-to-quality dynamic.
Liquidity Borrowing: Banks under stress responded by borrowing heavily from Federal Home Loan Banks (FHLBs) and the Federal Reserve's discount window, but interestingly, they did not sell securities at a loss. Borrowing allowed these banks to survive the acute liquidity outflows.
Public Banks More Vulnerable: Publicly traded banks were more likely to experience runs, even when controlling for fundamental factors. This suggests that public information and signals, such as stock price movements, played a role in coordinating depositor behavior.
My Discussion Points
1. Narrative Framework
While the paper provides many insights, it would benefit from a clear framework to interpret the results, especially the distinction between solvency-based runs and coordination problems. Most of the runs in 2023, excluding SVB, appear to have been coordination-driven, not fundamentally solvency-based. This suggests that depositor behavior was influenced by public signals and perceived risks rather than the actual balance sheet vulnerabilities of the banks.
Question to consider: How do the liquidity outflows in 2023 compare to historical episodes, like the Global Financial Crisis (GFC)? How large were these outflows relative to liquidity coverage ratios (LCRs), and why were smaller banks disproportionately affected?
2. Why Were Small Banks Run?
The paper emphasizes that small and mid-sized banks were hit hardest. My argument here is that these banks were vulnerable due to their liquidity risk. They had larger claims on liquidity through uninsured deposits and credit lines, yet they lacked access to adequate liquidity buffers. During the period of quantitative easing (QE), many smaller banks took on more tail-risk, including through leveraged carry trades in U.S. Treasuries. When the Federal Reserve shifted to quantitative tightening (QT), these claims on liquidity did not shrink, and the smaller banks became prime targets for runs.
3. Policy Implications
The findings in the paper have clear policy implications. High-frequency deposit flow data can be a powerful tool for regulators to monitor bank liquidity in real time, both in periods of calm and crisis. Additionally, regulators may need to rethink the current liquidity coverage ratio (LCR) framework. Smaller and mid-sized banks, which are not currently subject to LCR requirements, may benefit from having similar liquidity standards to reduce their vulnerability to runs.
Another area of focus should be mark-to-market accounting for capital and more stringent stress testing of liquidity under extreme conditions, especially for smaller banks. Enhancing transparency and liquidity resilience is essential for avoiding future coordination problems that lead to unnecessary bank runs.
Conclusion and Path Forward
The paper by Cipriani, Eisenbach, and Kovner provides a rich dataset and new insights into the mechanics of bank runs in real time. Future research should focus on comparing the intensity of the wholesale runs in 2023 to earlier crises, such as the GFC, and delineating coordination-driven runs from solvency-driven ones.
Moreover, the role of public signals in exacerbating runs at publicly traded banks needs further exploration. An important aspect will be to identify who the depositors are to understand the intensity and speed of bank runs these days. How concentrated are depositors? How are the connected? By integrating these insights, policymakers can better address the challenges of bank runs and enhance financial stability.
This blog entry draws from my discussion at the 2024 Carey Finance Conference on October 17-18, where I had the opportunity to present my thoughts on the timely and important work of Cipriani, Eisenbach, and Kovner.
You can download my discussion slides here.