CMStatistics 2021: Start Registration
View Submission - CFE
Title: Bayesian assessment of identifying restrictions for heteroskedastic structural VARs Authors:  Tomasz Wozniak - University of Melbourne (Australia) [presenting]
Abstract: A flexible Bayesian structural vector autoregressive model is introduced identified through heteroskedasticity, encompassing a range of volatility processes and allowing for additional identifying restrictions. Consequently, it enables comparisons across structural models with alternative sets of restrictions that just identify homoskedastic specifications. The structural model exhibits a new standardization that sets the sum of each structural shock variances to one. This solution facilitates the development of a complete toolset for Bayesian inference, including a reference prior, an efficient estimation algorithm, and an unbiased marginal data density estimator for locally identified models. Applying this apparatus to three U.S. monetary policy models, we document the empirical outperformance of models making use of two policy variables over those with a single one.