A0539
Title: Assessing partial exogeneity of shock proxies: A Bayesian proxy-DSGE framework
Authors: Ryohei Oishi - University College London (United Kingdom) [presenting]
Abstract: The proxy DSGE framework is introduced, a Bayesian econometric methodology for evaluating the validity of externally identified structural shock proxies using dynamic stochastic general equilibrium (DSGE) models. The framework assesses proxies based on relevance and partial exogeneity conditions, which is a weaker but more empirically tractable notion of exogeneity that requires proxies to be uncorrelated with non-targeted structural shocks within a DSGE model. These conditions are econometrically implemented through a spike-and-slab prior, which allows posterior loadings on shocks to be exactly zero with positive probability. The framework further provides a coherent estimation strategy for DSGE models with potentially weak proxies by extracting informative signals from their joint distribution. Applying the framework to a medium-scale DSGE model with US monetary policy proxies shows that the Romer-Romer narrative proxy and several high-frequency identified proxies are valid, although the latter are weak. Incorporating these proxies yields a more inertial Taylor rule and more precise estimates of monetary policy shocks.
