No abstract

No abstract

To forecast the Consumer Price Index, adopting dimension reduction to extract useful predictors from a large set of monthly macroeconomic time series is a way to improve forecasting accuracy. This article compares two methods for extracting predictors, including the well-known classical factor model and the Peña-Box Model, which is a dynamic factor model. Compared with the classical factor model, the Peña-Box Model is more robust with respect to misidentifying models since it captures the time-effect relationship of original variables. Both simulations and empirical studies on forecasting the Consumer Price Index of the Four Asian Tigers confirm the advantages of the Peña-Box Model.