Dynamic Comovement Among Banks' Returns and Chargeoffs in the U.S.

Abstract

This paper develops a new measure of comovement in the banking sector that takes into account the dynamic nature of interlinkages among different bank holding corporations at different stages of business cycles. For this purpose, we use a dynamic factor model with time-varying parameters and stochastic volatility that decomposes the panel data for the return on assets (ROA) and net chargeoffs (NCO) into a common component and an institution-specific idiosyncratic components. We find that the relative contribution of the common factor in explaining the variation in ROA and NCO peaked during the financial crisis, suggesting a significant increase in systemic stress in the banking sector. Using the least absolute shrinkage and selection operator (LASSO) approach, we show that the estimated common components and their stochastic volatilities from our approach perform well when compared to other widely used measures of systemic risk in explaining real economic activity. Furthermore, we find that these measures have better in-sample fit with real economic activity measures than the industry averages of ROA and NCO that are frequently used in the banking literature. Finally, we provide economic interpretation for the idiosyncratic components as banking balance sheet characteristics.