Forecasting Non-Stationary Economic Time Series

Abstract

Addresses the problems confronting forecasting in economies subject to structural breaks. Discusses economic forecasting and provides a taxonomy of the sources of forecast failure. Explores the role of unmodeled shifts in parameters as a source of misprediction and considers other potential sources of forecast failure, such as model misspecification, parameter estimation uncertainty, a lack of parsimony in model selection, and forecast origin mismeasurement. Analyzes the likely effects of each source of forecast failure via Monte Carlo simulation and empirical examples. Investigates analytically whether time-series models in differences can outperform econometric models in forecasting when a structural break has occurred. Evaluates whether intercept corrections help to robustify forecasts in the face of change. Illustrates the effects of structural breaks and the performance of the various forecasting strategies in a model of consumer expenditure in the United Kingdom and in a model of U.K. money demand. Addresses the removal of deterministic shifts in systems of equations using linear combinations of variables. Considers the forecasting performance of some regime-switching autoregressive models relative to linear autoregressive alternatives. Illustrates a number of the forecasting approaches in a model of wages, prices, and unemployment. Clements is Research Fellow in Economics at the University of Warwick. Hendry is the Leverhulme Personal Research Professor of Economics and Fellow of Nuffield College at Oxford University. Glossary; author and subject indexes.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 419 citations exist upstream.