Working Paper: NBER ID: w23629
Authors: Erik Heitfield; Gary Richardson; Shirley Wang
Abstract: The initial banking crisis of the Great Depression has been the subject of debate. Some scholars believe a contagious panic spread among financial institutions. Others argue that suspensions surged because fundamentals, such as losses on loans, drove banks out of business. This paper nests those hypotheses in a single econometric framework, a Bayesian hazard rate model with spatial and network effects. New data on correspondent networks and bank locations enables us to determine which hypothesis fits the data best. The best fitting models are ones incorporating network and geographic effects. The results are consistent with the description of events by depression-era bankers, regulators, and newspapers. Contagion - both interbank and spatial - propelled a panic which healthy banks survived but which forced illiquid and insolvent banks out of operations.
Keywords: banking crisis; Great Depression; contagion; Bayesian hazard model
JEL Codes: C11; C23; C41; E02; N1; N12; N2; N22
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Contagion occurred during the initial banking crisis of the Great Depression (F65) | Weaker banks were forced out of business due to contagion effects (F65) |
| Higher leverage and lower liquidity prior to the crisis (F65) | Banks were more likely to fail (G21) |
| Geographic proximity to failed banks (F65) | Significant source of contagion leading to correlated bank failures (F65) |
| Liquidity and asset shocks interacted (E44) | Clustering of failures in time and space (C38) |