Affordable Access

Has U.S. Inflation Really Become Harder to Forecast?

Authors
Publication Date
Keywords
  • C53 - Forecasting And Prediction Methods
  • Simulation Methods
  • E31 - Price Level
  • Inflation
  • Deflation
  • C23 - Models With Panel Data
  • Longitudinal Data
  • Spatial Time Series

Abstract

Recently Stock and Watson (2007) showed that since the mid-1980s it has been hard for backward-looking Phillips curve models to improve on simple univariate models in forecasting U.S. inflation. While this indeed is the case when the benchmark is a causal autoregression, little change in forecast accuracy is detected when a noncausal autoregression is taken as the benchmark. In this note, we argue that a noncausal autoregression indeed provides a better characterization of U.S. inflation dynamics than the conventional causal autoregression and it is, therefore, the appropriate univariate benchmark model.

There are no comments yet on this publication. Be the first to share your thoughts.