Panic and propagation in 1873: a computational network approach

We assess systematic risk in the U.S. banking system before and after the Panic of 1873 using a combination of linear programming and computational optimization to estimate the interbank network based upon total gross and net positions of national banks a week before the crisis. We impose various liquidity shocks resembling those of 1873, and find the network can capture the distribution of interbank deposits a year later. The network may be used to predict banks likely to panic (i.e., change reserve agent) in the crisis. The identified banks see their balance sheets weaken in the year after the crisis more than other banks. The results shed light on the nature and regional pattern of withdrawals that may have occurred in a classic 19th century U.S. financial crisis.