Charge-offs, Defaults and the Financial Accelerator

We uncover new facts: U.S. banks countercyclically vary the ratio of charge-offs to defaulted loans (COD). The variance of this ratio is roughly 15 times larger than that of GDP. Canonical financial accelerator models cannot explain this variance. We develop an expression for the wedge between charge-offs and defaults in the model and show that introducing stochastic default costs as in Gunn and Johri (2013a) and stochastic risk as in Christiano et al. (2014) into the canonical theoretical model can potentially resolve the discrepancy since both shocks have the ability to move this wedge. Estimating the augmented model using Bayesian techniques reveals that default cost shocks account for most of the variance of COD, while risk shocks account for most of the credit spread. Both shocks also matter for standard U.S. business cycle variables, with the anticipated components of each being most important.