What causes bank failures?
In this paper, Correia et al. analyze the question of whether bank failures are predictable based on fundamentals or does the random nature of self-fulfilling panics make most bank failures unpredictable. Toward this goal the authors construct a database with balance sheet information for all U.S.-based banks from 1863 to 1941 and a modern sample that covers all commercial banks from 1959 to 2023. This data set includes the balance sheets of 38,630 distinct banks, of which 6,401 failed. They construct a new measure, Banks-at-Risk, that captures the share of banks with an elevated failure probability in t + 1 using information up to year t. Intuitively, this measure captures the thickness in the right tail of the distribution of the predicted failure probabilities. The authors find that the Banks-at-Risk measure forecasts the major waves of bank failures in both the historical sample and the modern sample. They conclude that:
- Bank failures are almost always and everywhere primarily related to weak fundamentals. Failing banks are characterized by a realization of credit losses that slowly erodes bank solvency several years before failure.
- The primacy of fundamental solvency issues for failures implies that bank failures are highly predictable.
Thus, our findings challenge the empirical relevance of theories that suggest bank failures are driven by non-fundamental, self-fulfilling panic runs. The central role of fundamentals emphasizes the importance of ex-ante interventions that increase the resilience of the financial system and reduce excessive risk-taking by banks.
Failing Banks
Authors: Sergio Correia, Stephan Luck, Emil Verner
From: Board of Governors of the Federal Reserve System, Federal Reserve Bank of New York, MIT
