Reading the Panic: How Investors Perceived Bank Risk During the 2023 Bank Run

Natalia Fischl-Lanzoni, Martin Hiti, and Asani Sarkar

The bank run that started in March 2023 in the U.S. occurred at an unusually rapid pace, suggesting that depositors were surprised by these events. Given that public data revealed bank vulnerabilities as early as 2022:Q1, were other market participants also surprised? In this post, based on a recent paper, we develop a new, high-frequency measure of bank balance sheet risk to examine how stock market investors’ risk sensitivity evolved around the run. We find that stock market investors only became attentive to bank risk after the run and only to the risk of a limited number (less than one-third) of publicly traded banks. Surprisingly, investors seem to have mostly focused on media exposure and not fundamentals when evaluating bank risk. In a companion post, we examine how the Federal Reserve’s liquidity support affected investor risk perceptions.

How Risky Were Banks Before the Bank Run?

We emphasize two balance sheet features that turned out to be particularly problematic during the bank run: the share in total assets of uninsured deposits (denoted UID) and the share in total assets of unrealized losses on securities held in accounts not intended for trading (denoted Losses). High values of UID proved to be risky as they were concentrated in certain sectors, which heightened the risk of rapid withdrawals. When Losses are high (typically when interest rates are increasing, as in 2022) and ultimately realized, bank capital is more likely to be eroded below regulatory limits.

To benchmark bank risk, we construct four groups of publicly traded banks that were differentially affected during the bank run: distressed banks that were downgraded in April 2023, large regional banks (those included in the Regional Banking Index, or KRX) that were at the heart of the crisis, small regional banks (those with assets greater than $10 billion that are not included in any bank index), and stress-tested banks, or STBs (the large banks that participated in the Federal Reserve stress tests of 2022 and that are included in the broad bank index, or KBW).

The chart below shows the median values of UID (left panel) and Losses (right panel) in 2022 by bank group. For reference, we also include the three banks that failed in March 2023 (Silicon Valley Bank, Signature Bank of New York, and Silvergate Bank), although they are not part of our analysis. Bank vulnerabilities were apparent as far back as 2022:Q1. Banks that would later fail or be distressed stood out with the highest levels of UID and Losses. Notably, Losses increased for all bank groups as the Federal Reserve raised rates, peaking in Q3. While large regional banks did not have atypical levels of UID, they had more Losses than smaller regionals and STBs.

Bank Vulnerabilities Were Apparent in 2022

Sources: Federal Reserve Board, Consolidated Financial Statements of Bank Holding Companies (FR Y-9C data); Federal Financial Institutions Examination Council, Consolidated Reports of Condition and Income (Call Reports).
Notes: The chart shows the median asset shares of uninsured deposits (UID) and unrealized securities losses (Losses) in 2022 by groups of publicly traded banks. Failed banks are those that were liquidated or failed in March 2023. Distressed banks are those that were downgraded in April 2023. Large regional banks consist of non-downgraded regional banks in the Regional Banking Index (KRX). Small regional banks are those with assets of at least $10 billion that were not included in any bank index. Stress-tested banks were part of the Federal Reserve’s stress tests in 2022 and part of the broad bank index KBW.

Did Stock Market Investors Monitor Risky Banks?

Our novel measure of bank balance sheet risk is constructed as follows. First, we calculate the average stock returns of a portfolio of banks with a particular balance sheet characteristic (such as UID) in 2022:Q3. For example, the UID portfolio return is the average difference in stock returns of banks with the highest UID minus those with the lowest UID in 2022:Q3. If banks with higher UID are riskier, this factor is expected to have a positive return on average (to compensate stock market investors for the greater risk).

Next, to measure how investors perceived the UID risk of a bank, we estimate the co-movement of a bank’s excess stock returns with the UID portfolio return, after accounting for other types of risk (for example, size, value, and stock market risk). We denote this co-movement as the UID beta. If a bank’s UID beta increases, this indicates that investors are more sensitive to the bank’s systematic UID risk. We construct the Losses beta using an identical procedure. (Our paper also constructs betas with respect to cash and regulatory capital.)

The chart below shows estimates (indicated by the dots) of the average UID beta and Losses beta for all banks before the run (January–February 2023; Pre in the chart) and during the run (March 1–May 5, 2023; Post in the chart). The lines through the dots indicate confidence intervals. The chart shows that, before the run, the betas were not statistically different from zero (since the confidence interval straddles zero). But after the run the betas become positive and statistically significant. In other words, investors mostly ignored risks from high levels of uninsured deposits and unrealized losses on securities—until the crisis actually hit.

Investors Mostly Ignored Bank Risk Until the Bank Run Hit

Which Banks Did Investors Run On?

At the individual bank level, we find that the betas increased during the run for about a third of all publicly traded banks. Thus, investor concerns about bank risk were seemingly not broad-based. What characterized this limited set of banks? Surprisingly, we find that balance sheet variables as of 2022:Q3 or Q4 fail to predict which banks had significantly higher betas during the run. In other words, banks perceived as riskier by investors during the run were seemingly not the ones with worse fundamentals in 2022.

How News Drove Risk Perceptions

If fundamentals did not drive investor attention, then what did? We consider whether news coverage facilitated the coordination of investor attention on certain banks. Such a possibility has previously been found in the context of bank failures. We define a bank’s news coverage as the number of articles about a bank on a given day divided by the bank’s assets, to account for larger banks having more publications. We denote this variable Pubcount. In the chart below, Pubcount (measured as deviations from zero, and in standard deviation units) shows considerable daily variation, implying that even banks with low average media coverage experience periods of relatively intense publicity. Through March 8, just before the crisis, Pubcount was negative (i.e., below average) for all groups. When some distressed banks were put on a downgrade watch by Moody’s after the markets closed on March 13, Pubcount spiked for all distressed banks. Media interest surged again when the distressed banks were downgraded starting on April 14. In general, it appears that increases in Pubcount are associated with risk events.

Bank Publications Increase Around Risk Events

Does news coverage affect the balance sheet betas? In the chart below, we show estimates of news betas (or the component of betas that vary with Pubcount) around the bank run. These news betas are significantly negative before the run (i.e., publications were associated with lower investor risk sensitivity) but became significantly positive during the run (i.e., publications were associated with higher investor risk sensitivity). These effects are economically meaningful as the news betas are at least as large as the non-news betas. The effect of news on the betas persisted for days after publication, suggesting that investors paid attention to news only when it became salient to them.

When a bank appeared in the news during the crisis, investors became much more sensitive to its risks—regardless of whether the bank had a riskier balance sheet than its peers. This result reinforces the notion that news coverage coordinated investors’ actions (and thereby their perceptions of bank risk), either because it reflected latent risk events not captured by balance sheet data, or because investors overreacted to the news.

Final Words

During the bank run of 2023, news flows were at least as important as underlying bank fundamentals in driving investor perceptions of bank risk. News coverage, even when stale, appeared to have served as a coordination device, helping investors focus collectively on certain banks. These results imply that investors may be unable to quickly process information in a crisis, potentially making market price dynamics noisier, to the detriment of market participants and policymakers. However, as investor attention was focused on a few banks rather than a broad swathe of the banking sector, the contagion was contained. Liquidity support by the Federal Reserve may also have limited contagion, a topic we examine in our companion post.

Natalia Fischl-Lanzoni, a former research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group, is pursuing a master’s in computer science at NYU Courant.

Martin Hiti, a former research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group, is a Ph.D. student in finance at the MIT Sloan School of Management.

Asani Sarkar is a financial research advisor in the Federal Reserve Bank of New York’s Research and Statistics Group.

 @article{
        Natalia Fischl-Lanzoni, Martin Hiti, and Asani Sarkar2025,
        author={Natalia Fischl-Lanzoni, Martin Hiti, and Asani Sarkar},
        title={Reading the Panic: How Investors Perceived Bank Risk During the 2023 Bank Run},
        journal={Liberty Street Economics},
        note={Liberty Street Economics Blog},
        number={September 30},
        year={2025},
        url={https://doi.org/10.59576/lse.20250930a}
    }