EcoSta 2018: Registration
View Submission - EcoSta2018
Title: Forecasting macroeconomic series by unobserved component models with ARMA-SV errors Authors:  Bo Zhang - Australian National University, Research school of Economics (Australia) [presenting]
Abstract: An autoregressive moving average component with stochastic volatility is introduced into the unobserved component model. A transformation in a stacked matrix form of the model is conducted for the posterior fast simulation, and a recently developed precision-based algorithm, particularly for the unobserved component model, is adopted for analyzing the serially dependent errors. The proposed model is then used to study macroeconomic time series in the United States. It is found that the proposed new model provides good full sample simulation for a large part of the macroeconomic variables, and it can improve both point forecast and interval forecast performance of these variables across different horizons.