Using time-varying volatility for identification in Vector Autoregressions: An application to endogenous uncertainty

Abstract

We develop a structural vector autoregression with stochastic volatility in which one of the variables can impact both the mean and the variance of the other variables. We provide conditional posterior distributions for this model, develop an MCMC algorithm for estimation, and show how stochastic volatility can be used to provide useful restrictions for the identification of structural shocks. We then use the model with US data to show that some variables have a significant contemporaneous feedback effect on macroeconomic uncertainty, and overlooking this channel can lead to distortions in the estimated effects of uncertainty on the economy.