A1512
Title: Bayesian multiple-indicator mixed-frequency model with moving average stochastic volatility
Authors: Boriss Siliverstovs - Bank of Latvia (Latvia) [presenting]
Abstract: A Bayesian mixed-frequency multiple-indicator model is suggested with moving-average stochastic volatility that nests 1) U-MIDAS model of a prior study, 2) U-MIDAS model with MA-component of another study, and 3) a multiple-indicator model with stochastic volatility of a previous study. The general models, as well as their restricted versions, can be efficiently estimated using the precision-based algorithm as introduced by another study. The evidence of the past study is re-examined on the usefulness of including a moving-average component in the mixed-frequency forecasting models. The results are less optimistic on the additional value of the MA-component based on out-of-sample model evaluation, using either point or density forecasts.
