A state-space approach to time-varying reduced-rank regression

Abstract

We propose a new approach to reduced-rank regression that allows for time-variation in the regression coefficients. The Kalman filter based estimation allows for usage of standard methods and easy implementation of our procedure. The EM-algorithm ensures convergence to a local maximum of the likelihood. Our estimation approach in time-varying reduced-rank regression performs well in simulations, with amplified competitive advantage in time series that experience large structural changes. We illustrate the performance of our approach with a simulation study and two applications to stock index and Covid-19 case data.