B0369
Title: Optimal forecasts for multivariate Markov switching autoregressive models
Authors: Maddalena Cavicchioli - University of Modena and Reggio Emilia (Italy) [presenting]
Abstract: The optimal forecasts for multivariate autoregressive time series processes are derived subject to Markov switching in regime. Optimality means that the trace of the mean square forecast error matrix is minimized by using suitable weighting observations. Then, neat analytic expressions for the optimal weights are provided in terms of the matrices involved in adequate state space representations of the considered processes. The matrix expressions in closed form improve computational performance since they are readily programmable. Numerical simulations and empirical applications illustrate the feasibility of the proposed approach. Evidence is provided that the forecasts using optimal weights increase forecast precision, and are more accurate than the linear alternatives.
