Inflation Forecasts Using a Nelson-Siegel Representation of Nominal and Real Yield Curves

Abstract

he aim of this paper is to evaluate the performance of inflation forecasts backed out from the nominal and real yield curves in the United Kingdom. We use the Nelson-Siegel (NS) framework to model the break-even inflation term structure, and we also consider the one-day break-even inflation derived from NS parameters. In both cases we fit (vector) autoregression models, some of which are augmented by nominal and/or real Cochrane-Piazzesi factors. The paper provides three main results. First, parsimonious models based on the one-day break-even inflation outperform other models in forecasting inflation out-of-sample. Second, we quantify the parameter uncertainty, showing that it may have considerable impact on inflation forecasts, in particular after the Lehman Brothers event. Third, model-free forecasts directly obtained from break-even inflation rates are typically hard to beat by parametric alternatives.