Bank runs have long terrified depositors and policymakers alike. Yet fresh analysis from the Federal Reserve Bank of New York shows that the true danger lies not in the panic itself but in the condition of the institutions under siege. When fundamentals weaken, runs turn deadly. Otherwise, they often fade without widespread damage.
Researchers at the New York Fed built a massive historical database using large language models. They scoured millions of digitized newspaper pages to catalog U.S. bank runs like never before. The work, released Tuesday, challenges old assumptions. Small shocks rarely ignite broad panics. Poor bank fundamentals prove necessary for runs to cause failures and economic pain.
“Poor bank fundamentals are necessary for bank runs to translate into failure and for bank distress to generate severe economic distress,” the researchers wrote. They added that “although runs can occur in both weak and strong banks, poor fundamentals are necessary for runs to result in bank failures.” The findings appeared in a Liberty Street Economics blog post and a companion academic paper.
This matters for monetary policy. Central banks adjust interest rates expecting their moves to flow through banks to the real economy. Healthy banks transmit those signals cleanly. Troubled ones distort them. Rate hikes can exacerbate weaknesses. Cuts may fail to stimulate if institutions hoard liquidity instead of lending. The New York Fed work adds historical depth to this dynamic.
Consider recent events. The 2023 failures of Silicon Valley Bank and Signature Bank showed how quickly confidence can evaporate in the digital age. Yet those banks faced real solvency questions tied to unrealized losses on bond portfolios. The New York Fed’s historical lens suggests such underlying issues often separate contained incidents from systemic collapse. A separate New York Fed staff report from 2024, revised in 2026, used high-frequency payment data to trace 22 runs during that spring episode. Only a fraction ended in failure.
The AI-powered database marks a breakthrough. Previous efforts relied on limited records or manual review. Here, models sift through the Library of Congress’s Chronicling America collection. They identify runs, classify triggers like macroeconomic news or bank-specific problems, and distinguish distress types. A companion site at finhist.com lets users explore individual episodes and original clippings.
And the results speak volumes. Bank runs appear throughout history. But their economic fallout varies sharply with bank health. Strong banks absorb the outflow. Weak ones fail, dragging credit and confidence down with them. This pattern holds across eras, from the 19th century through the Great Depression and beyond.
Policy implications extend far. Central bankers at the Federal Reserve, European Central Bank and others watch bank balance sheets closely during tightening cycles. The ECB noted in a 2024 speech that strong bank capital helped transmission proceed smoothly during recent rate hikes. Lending rates rose, credit standards tightened, yet the system avoided disorder. “The robust state of bank balance sheets meant that the transmission of rate hikes to banks could proceed in an orderly manner,” ECB officials observed.
But heterogeneity persists. Smaller banks, regional players and those with concentrated exposures transmit policy differently. A March 2026 paper from the Bank for International Settlements examined regional bank health from 2008 to 2023. Local commercial real estate conditions, especially industrial properties, heavily influenced supervisory ratings. Management quality stood out among CAMELS components.
Nonbanks complicate matters further. As they grow, traditional bank-centric transmission weakens in spots. An older Wall Street Journal analysis from 2017 found nonbanks often amplify rather than mute policy effects, contradicting some fears. Yet gaps remain in oversight. Macroprudential tools target banks more effectively than shadow finance or households.
Recent easing cycles reveal another side. A December 2025 SUERF policy note found monetary transmission progressing but uneven. Risk-taking and balance sheet channels operated less forcefully for smaller firms and lower-income households. Mortgage demand shifted toward higher-quality borrowers. Banks showed muted appetite for risk. Such patterns may blunt stimulus precisely when consumption and investment matter most.
Meta-studies reinforce the nuance. A 2026 analysis in Comparative Economic Studies reviewed interest-rate pass-through across countries. Speed and strength vary. Post-crisis periods often see slower adjustment. Factors like bank capitalization, competition and regulatory regime shape outcomes.
So what does this mean for today’s policymakers? They cannot ignore bank health when setting rates. Stress tests help, as the Fed’s 2026 exercise confirmed large banks could weather severe downturns. Yet tests miss real-time runs and behavioral shifts. The New York Fed’s real-time payment data work from 2024 offers one tool. The new historical database supplies another, letting history inform models of crisis propagation.
Critics once blamed mega-banks for impairing policy transmission. A 2009 Dallas Fed speech highlighted how too-big-to-fail institutions distorted incentives. Today’s mix of big banks, regionals and nonbanks creates different fragilities. Commercial real estate exposure lingers as a vulnerability for many mid-sized lenders.
The researchers caution against overinterpreting isolated runs. Data offer “little support” for the notion that minor shocks spawn widespread banking panics. Insolvency or deep weakness typically precedes catastrophe. Runs act as triggers, not root causes.
This reframing carries weight. Regulators might focus more on early detection of fundamental weakness than on run prevention alone. Deposit insurance and liquidity facilities buy time. But solvency remains the decisive factor. History, now readable at scale through AI, confirms it.
Monetary policy works best when banks stay sound. Rate changes then influence borrowing costs, investment and inflation as intended. When health falters, transmission frays. Credit contracts unevenly. Economic pain concentrates. The New York Fed’s latest contribution sharpens that understanding with evidence spanning centuries.
Future research will test the database further. Geographic patterns, run durations and links to broader crises await deeper examination. For now, the core message stands clear. Bank health doesn’t just matter. It determines whether panic stays contained or cascades. Central banks ignore that lesson at their peril.
WebProNews is an iEntry Publication