Estimating a Banking-Macro Model Using a Multi-regime VAR

Abstract

This paper indroduces a Banking-Macro Model and estimates the linkages using a Multi-Regime Vector Auto Regression (MRVAR). The model of the banking-macro link is a simplified version of the Brunnermeier and Sannikov (Am. Econ. Rev., 2014) model. The banking sector is represented as a wealth fund that accumulates capital assets, can heavily borrow and pays bonuses. We presume that the banking sector faces not only loan losses but is also exposed to a deterioration of its balances sheets due to adverse movements in asset prices. In contrast to previous studies that use the financial accelerator—which is locally amplifying but globally stable and mean reverting—our model shows local instability and globally multiple regimes. Whereas the financial accelerator leads, in terms of econometrics, to a one-regime VAR, we demonstrate the usefulness of the MRVAR approach. We estimate our model for the U.S. with a MRVAR using data on a constructed financial stress index and industrial production. We also conduct impulse-response analyses which allowing us to explore regime dependent shocks. We show that the shock profiles depend on the regime the economy is in and the size of the shocks. As to the recently discussed unconventional monetary policy of quantitative easing, we find that the relative effects of monetary shocks depend on the size of the shocks.