Abstract

The local projection method proposed by Jordà (2005) has been widely used as a promising framework for computing impulse responses. In the previous literature, a time-varying version of the local projection has been proposed, but it does not address the time-varying variance of the error distribution. Ignoring a possible time variation in the error variance could cause a severe bias in the time-varying impulse responses. To overcome it, this paper proposes the time-varying parameter local projections with stochastic volatility. A Bayesian method using an efficient Markov chain Monte Carlo is developed to analyze the proposed model. The application to monetary policy effectiveness is provided using the U.S. macroeconomic variables.