BankCaR (Bank Capital-at-Risk): A Credit Risk Model for US Commercial Bank Charge-Offs

Abstract

BankCaR is a credit risk model that forecasts the distribution of a commercial bank's charge-offs. The distribution depends only on systematic factors; BankCaR takes each bank and projects its expected charge-off across a distribution of good years and bad years. Since most bank failures occur in bad years, this analysis has promise for both banks and bank supervisors. In BankCaR, charge-offs depend on the bank's loan balances and the charge-off rates of twelve categories of lending. A joint distribution of the twelve charge-off rates is calibrated to a long history of regulatory reporting data. Applied to the US banking system, BankCaR finds that credit risk is rising and is concentrated most significantly in construction lending. Applied to individual banks, BankCaR efficiently identifies those that have an adverse combination of credit risk and capital.

BankCaR uses publicly available regulatory reporting data, the most common credit portfolio model, and standard quantitative techniques. These generic qualities can provide a standard of comparison between banks. They also can provide an individual commercial bank with a benchmark for more elaborate vended credit models.

Keywords: credit risk, risk screening, loan charge-offs, validation

JEL Classification: G32, G21, G38