Title: A subspace estimator for mixed frequency VAR systems
Authors: Manfred Deistler - Vienna University of Technology (Austria)
Philipp Gersing - Vienna University of Technology (Austria) [presenting]
Abstract: A novel estimation procedure is proposed for mixed frequency data where the underlying high frequency process is a VAR process. We derive an explicit minimal stable and miniphase state space representation for a blocked process which second moments correspond to all those second moments observable under mixed frequency. Based on this representation a relatively simple subspace estimator for the high frequency parameters is derived which makes full use of the information available and is consistent and efficient. The computational burden of the procedure is much lower compared to methods involving Kalman filtering. We aim to extend the estimator to the case of mixed frequency generalized dynamic factor models where the unobserved component has rational spectral density.