Forecasts with Single-Equation Markov-Switching Model: An Application to the Gross Domestic Product of Latvia

Abstract:     
The paper compares one-period ahead forecasting performance of linear vector-autoregressive (VAR) models and single-equation Markov-switching (MS) models for two cases: when leading information is available and when it is not. The results show that single-equation MS models tend to perform slightly better than linear VAR models when no leading information is available. However, if reliable leading information is available, single-equation MS models tend to give somewhat less precise forecasts than linear VAR models.

Number of Pages in PDF File: 13

Keywords: Markov-switching, VAR, forecasting, leading information

JEL Classification: C13, C22, C32, C51, C52, C53

working papers series