Forecasting macroeconomic variables using a structural state space model

摘要

This paper has a twofold purpose; the first is to present asmall macroeconomic model in state space form, the second is to demonstrate that it produces accurate forecasts. The first of these objectives is achieved by fitting two forms of a structural state space macroeconomic model to Australian data. Both forms model short and longrun relationships. Forecasts from these models are subsequently compared to a structural vector autoregressive specification. This comparison fulfills the second objective demonstrating that the state space formulation produces more accurate forecasts for a selection ofmacroeconomic variables.