Credit spreads and systematic risk in the U.S. banking industry - A neural network model approach

Abstract

This paper shows that systematic risk in the U.S. banking industry displayed historical responsiveness to variations in the AAA-Baa credit spread. Critically, through the development of a series of single hidden layer perceptron neural network models, the principal credit spreads in the fixed income market catalyzed a defined regime shift in systematic risk proximate the financial crisis, and was more influential to the quantification of realized systematic risk than the statistical specifications of beta. As an intriguing result of the learned model simulations, the beta slope coefficients for the largest banks in the study exhibited significant acceleration in the statistical dependence on credit spread variations.