Forecasting in the Presence of Recent Structural Breaks

• Activity Number: 385
• Type: Invited
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Simon Price*+
• Companies: Bank of England

Abstract

We ask how to forecast after a recent break. We consider: monitoring for change and combining forecasts from models that include and exclude data before the change; and methods robust to structural change. The latter include rolling regressions, forecast averaging over different windows and exponentially weighted moving average (EWMA) forecasting. We derive some analytical results, and assess using Monte Carlo methods. Break monitoring improves performance at low cost; rolling regressions are effective for large breaks but may not be in other cases; the EWMA is often a poor performer; forecast averaging is effective. We then evaluate performance for a large number of UK and US macroeconomic series.