Title: Efficient estimation of mixed-frequency state-space VARs: A precision-based approach
Authors: Dan Zhu - Monash University (Australia) [presenting]
Joshua Chan - Purdue University (United States)
Aubrey Poon - University of Strathclyde (United Kingdom)
Abstract: Mixed-frequency vector autoregression state-space models are now widely used for forecasting and nowcasting applications. However, despite their popularity, estimating such models can be computationally intensive. We propose a novel precision-based sampler to draw the missing observations of the low-frequency variables in these models. The newly proposed method builds on the recent advances in banded and sparse matrix algorithms for state-space models. In the simulation study, we find our new proposed method delivers superior accuracy and is computationally more efficient compared to standard filtering methods, which are commonly employed in these models. We also illustrate how our new proposed method can be applied to two popular empirical macroeconomic applications. The key insight from these two empirical applications highlights the importance of incorporating high-frequency indicators in macroeconomic models.