How do events within war affect financial market assessments of how the war will end? This question is both intrinsically interesting and provides a method for researchers of war to assess how battles influence war termination in an environment where there is little quality quantitative intra-war data. Since bond prices are a function of expected future payments and war results can have a large influence on the probability of receiving those payments, sovereign bonds can act like a pseudo-prediction market of the war result. Due to their high frequency, financial markets can be used to assess which events affected the expected war result, and the magnitude of those effects.
This paper focuses on the American Civil War, and estimates the effects of major battles in the American Civil War on the price of bonds issued by the Union. To do so, it introduces two datasets: one of the battles of the American Civil War and another of the financial market prices of issued by the state and federal governments during that era. The effect of battles on prices is estimated within the Bayesian dynamic linear model framework. This paper shows how Bayesian dynamic linear models can easily and efficiently be estimated with the software package, Stan.
This is done by combining Stan's Hamiltonian Monte Carlo algorithm with Forward Filter-Backwards Sampling on the latent states.
