Panic and Propagation in 1873

Understanding the role of interbank credit relationships to systemic risk is challenging in that the total gross and net positions of banks can usually be observed from balance sheets but detailed information about particular correspondents is usually confidential. This was less the case for the United States in the early 1870s when the vast majority of banks were organized under national charters. The simpler portfolios held by banks at the time also render their balance sheets, which survive as recorded in the Annual Reports of the Comptroller of the Currency, more informative statements of condition than their modern-day counterparts. We use these advantages together with the regulatory constraints of the time to assess systemic risk in the U.S. banking system before, during, and after the Panic of 1873. We do this through a combination of linear programming and computational optimization that allows us to estimate the interbank network and utility parameters simultaneously using balance sheet data reported a week before the crisis. We show that, consistent with historical observation, direct counter party risk was small, i.e., an unrealistically large shock would be needed for a significant number of banks to fail. Comparisons of the network for 1873, after imposing various liquidity shocks, with actual post-crisis balance sheets from the fall of 1874 show that the network can predict which banks became subject to panics.