Forecasting Economic Time Series Using Unobserved Components Time Series Models

Abstract

AbstractThis article discusses the forecasting of economic time series on the basis of a general class of unobserved components time series models, and is organized as follows. Section 2 provides a comprehensive review of unobserved components time series models. Section 3 discusses the methodology of state-space analysis. Section 4 discusses how forecasts can be generated as part of a state-space time series analysis and how observation weights of the forecast function are computed. Various multivariate extensions of the unobserved components time series model are discussed in Section 5. Specifically, it presents multivariate time series models with common trends and cycles, and discusses how a dynamic factor analysis based on maximum likelihood can be carried out in a computationally efficient way. To illustrate the methodology, Section 6 presents an empirical analysis for daily electricity spot prices based on a univariate and a bivariate model. It presents some interesting features of this analysis, focusing primarily on the forecasting of daily spot prices. Section 7 concludes.