A central bank's main concern is the general direction of future inflation, and not transitory fluctuations of the inflation rate. As a result, this paper is concerned with forecasting a simple measure of the trend of inflation, the eight-quarter CPI-inflation rate. The primary objective is to improve the M1-based vector-error-correction model (VECM) developed by Hendry (1995), by imposing a set of equilibrium conditions to better anchor the long-run behaviour of interest rates, the exchange rate and the output gap in the model. These changes provide for greater confidence in the dynamic properties of the model, especially over a longer time horizon. This extended-VECM is shown to provide considerable leading information about inflation, forecasting the eight-quarter inflation rate with relatively small errors. The authors also stress that, to be most useful for monetary policy, inflation forecasts should explicitly indicate the range of uncertainty inherent in forecasting inflation with a long lead. For example, forecasts should explicitly consider confidence bands around forecasted outcomes, which is illustrated with the extended VECM developed in this paper. Finally, the paper emphasizes that monetary policy is probably best-served by an eclectic approach in which policy judgements are based on input from models that summarize different paradigms of the transmission mechanism, or that use different technical approaches.
