A New Time-Varying Parameter Autoregressive Model for U.S. Inflation Expectations

摘要

We study the evolution of U.S. inflation by means of a new noncausal autoregressive model with time-varying parameters that outperforms the corresponding causal and constant-parameter noncausal models in terms of fit and forecast accuracy. Our model also beats the unobserved component stochastic volatility (UCSV) model, one of the best-performing univariate inflation forecasting models, in terms of both point and density forecasts. We also show how the new Keynesian Phillips curve can be estimated based on our noncausal model. Both expected and lagged inflation turn out important, but the former dominates in determining the current inflation.